Leading SSD controller supplier Silicon Motion (SIMO) is forecasting record revenue in 2026, yet the growth story still appears driven primarily by its more established markets across mobile, PC and automotive rather than by a major breakout in AI or data center.  

In the near-term, SIMO is guiding sequential revenue growth throughout the year, primarily driven by the mobile revenue segment with management stating: “We expect continued strength across nearly all our product segments with a particular emphasis on mobile where we expect significant outperformance due to continued market share gains.” 

That said, management has made it clear execution is dependent in part on securing NAND supply on a stable basis, with higher input costs passed through to customers.  

SIMO has a handful of notable customer wins, aside from working with the leading NAND flash and SSD suppliers. The company is supplying controllers for Nvidia’s BlueField DPUs in the second half of the year, along with several NVLink and Ethernet switch solutions. Analysts also implied that the company has won a role in Google’s ASICs project although management did not directly confirm the customer name. 

Today, SIMO appears to be more adjacent to the action around NAND prices surging rather than being in the center of it. While NAND suppliers that own the flash content are seeing the benefits of higher average sales price per gigabyte and higher bit shipments, instead, SIMO sells the controller silicon and firmware that manages the NAND which does not benefit directly from pricing upside. 

Over time, that could change with key products such as enterprise boot drive storage, MonTitan enterprise SSD controllers and PCIe6 MonTitan controllers. Below, we discuss key products that could help SIMO pivot toward the lucrative and highly supply-constrained AI memory market. 

Brief Product Overview 

Silicon Motion supplies a range of high-performance NAND flash controllers, and is the leading global supplier of PCIe Gen4 and Gen5 SSD controllers used in PCs, client devices, and enterprise/data center applications, along with eMMC and UFS controllers for smartphones and IoT devices. The company’s controller chips and firmware help manage how data is stored and accessed across these devices.  

SIMO is heavily exposed to the consumer/client markets, with eMMC/UFS controllers driving 40-45% of revenue in fiscal 2025, with likely a large portion of SSD controller revenue (45-50% of revenue) coming from the two end markets.  

However, Silicon Motion is looking to make solid inroads into the data center heading into 2027, serving merchant GPUs with Nvidia’s new BlueField DPU on its Vera Rubin generation with boot-storage, which helps to start and manage the system. Although this is not nearly as critical (or as valuable) as SSD like what SanDisk offers, which is the main data storage, supplying the boot-storage marks a shift for SIMO as their products will now be in AI servers.  

There are also some other AI-driven tailwinds within compute and enterprise SSDs, being served by its MonTitan platforms in the second half of 2026 for QLC and TLC NAND for AI storage applications. These SSD controllers will help to solve the data storage challenges that AI servers face by helping them feed data and access data efficiently.  

The company supplies the leading flash manufacturers including Kioxia, Micron, SK Hynix/Solidigm, Samsung, SanDisk, and China’s YMTC. Outside of memory suppliers, Amazon is a key customer and Nvidia.  

Moving into Data Center with MonTitan and Enterprise SSDs for AI Servers 

For its AI-oriented product portfolio, SIMO offers a handful of high-performance enterprise SSD controllers in 8-channel and 16-channel configurations, as well as its MonTitan PCIe Gen5 SSD development platform. MonTitan targets data center and enterprise boot drive storage needs, featuring an SSD controller ASIC, reference design kits and enterprise firmware, allowing customers to optimize designs to meet their performance needs.  

SIMO began qualification of high-performance TLC compute SSDs on MonTitan at multiple customers in Q4, with qualification expected to progress through the first half of calendar 2026 and commercial ramp occurring in 2H. Additionally, SIMO expects to begin qualifications with multiple customers this year for QLC storage SSDs. 

MonTitan is expected to be a core growth driver for SIMO once the ramp begins in 2H, with management already guiding for the platform to reach 5-10% of revenue exiting 2026, or ~$60 to $120 million run rate based on current estimates for $1.27 billion annual and $335 million revenue in Q4.  

SIMO is also planning to tape out its first 4nm PCI Gen6 version of MonTitan (as current controllers are 6nm) this year, targeting hyperscalers, NAND flash manufacturers, storage system providers and CSPs. The new 4nm solution is expected to drive growth in 2027 and 2028, with management noting they already have secured a design win expecting to ramp significantly in 2028. 

Ramping for Vera Rubin’s BlueField DPU in 2H 2026 

Silicon Motion is expecting to see tailwinds from merchant GPUs arise in the second half of 2026, from its involvement within Nvidia’s Vera Rubin on DPU boot drives, as well as its NVLink/Ethernet switch roadmap. Management discussed they were involved with other potential customers for enterprise boot drive solutions including a leading search engine firm, with analyst commentary implying that Google may have already been won as a customer, though this was not confirmed. However, SIMO faces two main headwinds at present that could push this growth story to the back burner – timing of the ramp and PC/mobile headwinds, and challenges procuring NAND.  

SIMO explained in Q4’s call that they kicked off volume shipments to Nvidia in the quarter for its current DPU generation, adding that they are also working with the GPU leader “to qualify the next-generation version of their DPU as well as for several NVLink and Ethernet switches of their new GPU/CPU platform that are expected to launch in the second half of 2026.” Management added that the next-gen DPU and switch both require higher capacities, with much higher unit volumes and ASPs, unlocking a new growth opportunity in the back half of the year through 2027. 

This likely corresponds to Nvidia’s upcoming BlueField4 and BlueField4 STX platform announced at GTC, as well as its Spectrum-6 switch family. The STX rack in particular underpins Nvidia’s in-house ‘Inference Context Memory Storage’ platform, an Ethernet-attached flash SSD tier optimized for KV cache at the pod level to accelerate large-context inference. For more on the ICMS platform, refer to our SanDisk analysis here: SanDisk: Shares Up 559% In 2025 On NAND Flash, Enterprise SSD Tailwinds.SanDisk: Shares Up 559% In 2025 On NAND Flash, Enterprise SSD Tailwinds. 

SIMO provided a brief view on the products it is supplying for Nvidia and their role at GTC: MonTitan SSDs for the ICMS platform and KV cache extension, as well as for near-GPU high performance storage; enterprise SSD controllers for compute-optimized, nearline SSD and warm data storage with TLC and QLC SSD support; and PCIe NVMe boot drives and boot drive controllers.  

Given the new nature of this growth opportunity, analysts questioned about the revenue opportunities from BlueField DPUs/boot drives, and how it will play out this year and into next. Management explained that they expect the DPU/boot drive volume to be “very meaningful in 2026” with revenue around $50 million, but next year to be “much higher” as the NVLink and Ethernet switches see “more volume in 2027.”  

This would imply DPU revenue contribution remaining around <5% of revenue in 2026, based on current estimates for $1.27 billion in annual revenue, though commentary for much higher growth with product ramps weighted next year suggests management has already has solid visibility into the business potentially becoming a much more meaningful topline driver by the end of next year.  

However, the discussions also shed light on some primary headwinds SIMO faces. First, the main challenge comes down to timing, with the ramp more weighted in the back half of 2026 and into 2027, whereas SIMO must currently navigate challenging PC and mobile markets. Second, growth depends on NAND procurement which must be bought at market prices, margins must be passed through to customers, and it is not the sole controller supplier for DPUs.  

Headwind #1: Timing 

SIMO did set the stage for a strong year, expecting record revenue and sequential growth in each quarter with Q1 being the lowest for the year, a strong statement considering growth headwinds facing the mobile and PC markets from rising memory costs, with volumes expected to decline YoY in both end markets.  

This is why timing may be the most challenging aspect to SIMO’s thesis, as it is working to ramp its presence in the data center, yet these initial contributions may not be material enough to overcome PC and mobile challenges as the year progresses.  

As noted above, SIMO is forecasting Q1 to be the lowest quarter of the year, primarily impacted by typical seasonal weakness in client SSDs with mobile expected to significantly outperform. Guidance was $292 million to $306 million, or $299 million at midpoint for roughly 7.4% QoQ growth, counter to typical seasonality but decelerating from 15.1% QoQ in Q4.  

Current estimates point to SIMO’s sequential growth decelerating further to 2.7% in Q2 and remaining in the low/mid-single digit range through year-end, overall marking a high teens deceleration from Q3’s peak on current estimates.  

A key factor behind the soft forward QoQ growth is likely the PC/mobile headwinds given the significance of eMMC/UFS controller and client SSDs in terms of dollar growth and revenue contribution.  

Recent projections from Gartner and IDC point to a worsening environment for PCs. IDC had originally forecast an (8.9%) decline in shipments in 2026 in its pessimistic scenario from December 2025, yet it now projects an (11.3%) decrease in shipments this year as rising memory costs force price hikes that bite into demand. Gartner sees a similarly woeful year for PCs, forecasting a (10.4%) decline, with analysts there saying this would represent the “steepest contraction in device shipments witnessed in over a decade.” 

For smartphones, both of the two groups forecast a similarly challenging year. IDC is forecasting a (12.9%) decline in the smartphone market in 2026, a rather substantial revision lower from its December projection for just a (5.2%) decline. For comparison, Gartner projects an (8.4%) decline. This follows low-single digit growth in 2025. 

SIMO does expect to outperform both markets and deliver growth as it benefits from market share gains and improved ASPs, explaining that they will “get the benefit of both higher share and higher ASPs this year in spite of any sort of macro issues around PC unit volumes.” 

However, it is still worth pointing out that there is no guarantee that it will be able to flawlessly navigate what could shape up to be some of the worst growth in PCs more than a decade and potential double-digit unit declines. If these supply chain difficulties persist and cause inventory buildups at OEMs, consumer/client NAND demand could be impacted, and SIMO’s limited data center opportunities (where NAND demand is likely to remain relatively well insulated from rising storage demand) may not be enough to offset growth and margin headwinds.   

Headwind #2: NAND Procurement 

While the PC and mobile headwinds are two challenging external headwinds, perhaps the most challenging internal headwind SIMO will face comes down to NAND supply. This is because the company has to procure NAND itself at market price, and then work to pass the costs on to customers, without benefitting from some of the end-unit pricing power that its customers are seeing: 

“We need to procure the NAND and NAND price at the market price. So we have to work out with the customer, we can pass through the cost increase to the end customer. So it is challenging but ongoing process quarter-by-quarter. It definitely will impact some of our gross margin but we manage the margin pass-through. So I think because even the customers, they have at least 2 to 3 supplier, so they're based on the price and based on the supply and depends the percentage.” 

The challenge here is that NAND prices reportedly rose ~90% QoQ in Q1, and were estimated to increase ~20% QoQ in Q2 in early February, yet the latest data suggests prices could rise 70-75% QoQ in Q2. Recent reports from Korea also suggest Samsung is doubling prices sequentially in the second quarter, implying that pricing is not yet slowing down. Simply put, when input costs rise 70% or 90%, prices must follow at a similar or faster rate in order to preserve margins.  

Considering the boot drive solution is a new product line with minimal revenue contribution until 2H, outsized pricing headwinds this early in its ramp phase could weigh substantially on gross and operating margins, which are already quite thin in the mid-40% and 13% range. This compares to its customers like Micron who are pushing above 80% and 76% next quarter as major beneficiaries of the price increases. 

Management also flagged that growth will be dependent on its ability to secure NAND supply ‘stably’, another major challenge considering SIMO has two NAND suppliers – stated as Kioxia and WDC (though likely to now be SanDisk). Management explained that one of these suppliers is secure, yet the other is not, which could impact their ability to secure stable supply at favorable prices, especially if the second prioritizes allocations to the highest bidder. To note, Kioxia announced in January that its NAND output for 2026 was essentially fully sold out. 

Financials  

Revenue Growth Accelerates to 45.7% in Q4 2025  

Silicon Motion’s Q4 2025 revenue grew by 45.7% YoY and 15.1% QoQ to $278.5 million. Revenue also beat estimates by 6.7%, reflecting the continued strength in mobile business and strong growth in the PCIe 5 SSD business. Revenue growth accelerated by 31.8 percentage points from 13.9% YoY in Q3 2025. Though sequential growth slowed from 21.8% in Q3, it was better than the decline of (10%) in the same period last year.   

Management also provided a strong Q1 revenue guide of $292 million to $306 million, implying a YoY growth of 79.6% and 7.4% QoQ at the midpoint. Management expects continued strength across its products with a particular emphasis on mobile end markets, where it expects significant outperformance driven by continued market share gains.

2025 revenue grew by 10.2% YoY to $885.6 million. Analysts expect revenue growth to accelerate to 43.1% YoY to $1.27 billion in 2026. Management expects 2026 to be a record year, with revenue growing each quarter sequentially, supported by first-half momentum in eMMC and UFS products. To briefly recap, MonTitan enterprise SSD products are expected to scale in the second half of 2026 and contribute 5% to 10% of revenue by the end of the year, with strong growth anticipated in the years ahead. 

Key Segments  

SSD Controllers  

SSD Controller sales increased by 25% to 30% QoQ and 35% to 40% YoY in Q4 2025, compared to an increase of 20% to 25% QoQ and a decrease of 0% to 5% YoY in Q3. Management expects client SSD controllers to ramp throughout the year, though the first quarter is seasonally weaker. They expect the new DRAM-less 4-channel PCIe 5 controller that was introduced last quarter to ramp significantly throughout 2026 and the MonTitan controller ramp is expected in the second half of 2026.  

For the full year 2025, SSD Controllers accounted for 45% to 50% of revenue and revenue declined by 0% to 5% YoY.  

eMMC + UFS Controllers  

The company’s eMMC (Embedded Multimedia Card) + Universal Flash Storage (UFS) controller sales increased by 0% to 5% QoQ and 50% to 55% YoY in Q4 2025, compared to an increase of 20% to 25% QoQ and 35% to 40% YoY in Q3 2025.  

For 2025, eMMC + UFS Controllers accounted for 40% to 45% of revenue and revenue increased by 20% to 25% YoY. Given the current backlog and customer outlook for 2026, management expects to significantly outpace the market and deliver another strong year of growth of the eMMC and UFS business despite the difficult market environment, with the two accounting for 35% to 40% of revenue in 2026. 

SSD Solutions  

SSD Solutions revenue increased by 125% to 130% QoQ and 110% to 115% YoY in Q4 2025 compared to an increase of 15% to 20% QoQ and decrease of 40% to 45% YoY in Q3 2025. For 2025, SSD Solutions accounted for only 5 to 10% of revenue and declined by 10% to 15% YoY.  

The company’s boot drives are part of  its  SSD solutions, and this segment is currently growing rapidly due to the strong AI demand.  

Margins  

The company’s margins are improving; however, management expects slight pressure in Q1 primarily due to product mix, with improvement anticipated as the year progresses.  They expect gross margins to return to the target range of 48% to 50% as the year progresses, as the mix of newer products increases, including the PCIe 5 controllers and enterprise SSD solutions.  

• Q4 gross profits grew by 56.2% YoY to $136.8 million with a gross profit margin of 49.1%. The adjusted gross margin improved by 320 basis points YoY and 50 basis points QoQ to 49.2%. The improvement in gross margin was primarily driven by the successful ramp of new products and a favorable product mix shift toward client PC products.   
• The adjusted operating margin came at 19.3%, up 380 basis points YoY and 350 basis points QoQ. Management expects full year 2026 adjusted operating margin to improve despite the higher expenses this year, primarily driven by operating leverage. 
• Q4 net income came at $47.7 million or 17.1% of revenue compared to $21.6 million or 11.3% of revenue in the same period last year. The company benefited from the realized/unrealized gain on investments of $24.2 million in Q4 2025 compared to $0.1 million in Q4 2024.  

Q4 Adjusted EPS grew by 45%  

The company’s  Q4 adjusted EPS grew by 44.8% YoY to $1.26. Analysts expect strong growth to continue and expect Q1 adjusted EPS to grow by 117.2% YoY and 75.9% in Q2 2026. Looking forward, adjusted EPS is expected to grow by 59.4% YoY in 2026 and 21.9% in 2027.  

Cash Flow and Balance Sheet  

The company’s cash flows have been lumpy, and the increase in inventory to support strong future growth has put pressure on cash flows.    

• Q4 operating cash flow was $1.6 million or 0.6% of revenue compared to an operating cash outflow of ($6.2 million) or (3.2%) of revenue in the same period last year.   
• Q4 free cash outflow was ($6.3 million) or (2.2%) of revenue compared to ($17 million) or (8.9%) of revenue in the same period last year.   
• The company had cash of $201.8 million and no debt at the end of Q4 2025 compared to $198.6 million in cash and no debt at the end of the previous quarter.  
• Inventories increased by 24.8% QoQ to $421.8 million to support the strong future growth.  

 Conclusion 

 SIMO is guiding for record breaking revenue in fiscal 2026 despite growing headwinds in the PC and smartphone markets from rising memory costs. Q1 is expected to the lowest quarter of the year for revenue with client SSDs seeing typical seasonal weakness, with sequential growth in each quarter thereafter.  The company’s AI opportunities around Nvidia’s BlueField DPUs are expected to arise in the second half of the year and contribute roughly $50 million with a larger impact in 2027, while MonTitan is expected to quickly ramp in 2H to 5-10% of revenue. 

Revenue growth will also hinge on SIMO’s ability to stably procure NAND supply to meet its end market demand, making increasing memory costs a point to navigate for growth.  

An analyst on the call threw out 2027-2028 as potential timing for when the enterprise segment could get interesting. Our analysis points toward something similar, which is that 2026 may be a solid growth year for mobile and also PC yet is unlikely to be the year SIMO breaks out as a major enterprise AI story.  

Given the growth we are seeing in NAND, we felt it was worth our time to look at this stock more closely, yet will put SIMO on the shelf for further review until end of 2026 to see if a few of these green shoots materialize in the AI market.

Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.

Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.

Recommended Reading:

• Fabrinet: EML Constraints Easing, Capacity Expansion Adding >$2.6 Billion Revenue Potential
• Western Digital: Visibility Extends Through 2028, Catching up to Seagate on Density
• Ciena: Benefitting from Data Center Scale Across Demand
• Seagate: Slow QoQ Data Center Growth, 2027 Capacity Under Discussions

Data is the fuel for training and inference as it enables stronger models and better inference results. As more data is generated and the value of data increases, the demand for storing data is also increasing quickly.  

There are a few ways that Western Digital can awaken an age-old industry to meet the demands of hyperscalers. The first is to increase areal density from 32TB to eventually 100TB as we end the decade. By packing more capacity into the same footprint, Western Digital delivers improved economics to alleviate surging capex. The company’s UltraSMR-enabled JBOD platforms offer TB per drive, lower cost per TB and lower power (and space) per TB, delivering not only increased capacity but also lower total cost of ownership.  

The company seeks to work with a larger customer base beyond hyperscalers by not just saying “go figure out SMR” on your own to enterprises. SMR stands for shingled magnetic recording (more on this below). There are barriers to adoption for SMR as it’s complex to integrate and can have unpredictable performance at times. Western Digital is planning for when AI broadens beyond hyperscalers by offering a validated JBOD platform that tests and tunes UltraSMR, offers predictability through controlled reference architecture, and that is also quick to deploy.  

When translating UltraSMR’s benefits in investor terms, management states it offers a 20% capacity lift over CMR (conventional magnetic recording) and a 10% capacity uplift over industry standard SMR. The UltraSMR solution is software-based, thus it’s “very accretive” from a margin standpoint. Last quarter, there was a 50% mix on UltraSMR with expectations this will increase as the company’s top three customers are onboard with UltraSMR drives today and another two to three are moving toward adopting UltraSMR. 

Looking beyond the recording format, HAMR is an important catalyst for Western Digital as it’s a new write technology that uses a tiny laser to briefly heat the disk surface to write onto higher-stability media at much higher densities. This means WDC can pack more data onto the same drive size to deliver higher capacity. The step-function upgrade to increase capacity is a key element to Western Digital’s future product road map. Overall, HAMR is a primary capacity lever that will lead to 100TB hard drives along with combining higher bits-per-platter and higher platter count.  

The analysis below is dense at times, yet a necessary step to discussing a stock that has seen returns of 591% over the past year.  

Brief Product Overview  

Multimodal datasets are among the largest drivers of incremental storage demand. As inference and physical AI scale, the low-cost and high-density of HDD economics are hard to beat. Video is a data hog that and even modest growth here from multimodal AI can create what’s called “data exhaust.”  

As you'll see below, HDD is the lowest cost per TB for bulk storage. As inference requires more exabytes to be stored, HDDs offer an advantage in that storage tier. In fact, Western Digital’s management team foresees a CAGR of 25%+ over the next 5 years with HDD representing “80% of the storage media that deployed within a hyperscale environment.” 

UltraSMR and ePMR   

To answer this incoming demand, Western Digital’s ePMR tech (energy-assisted Perpendicular MR) tech adds electrical currents to the write head to improve density and write smaller bits. This increase in areal density allows more data to be stored in smaller spaces, helping drive down costs per TB and improve TCO for customers. ePMR is a tried-and-true tech, with WDC noting that it has been the ‘workhorse of the industry’ for the last decade.  

WDC is currently shipping its ePMR-based drives in 26TB CMR (conventional magnetic recording) and 32TB UltraSMR (ultra shingled magnetic recording) configurations. It is also scaling to the world’s first 40TB UltraSMR later this year, a new product addition to its portfolio and a capacity that was previously thought to be impossible to reach with ePMR.  

UltraSMR also features technological and read channel optimizations that boost HDD capacity by up to 18% versus traditional CMR drives. A key enabler is Western Digital’s OptiNAND architecture, which integrates embedded NAND devices, enabling multiple active write zones and increasing track density per disk. WDC says its UltraSMR drives offer among the highest capacities per drive for data-intensive workloads, and consume as little as 5.5 watts of power when ideal, driving TCO lower. The UltraSMR drivers are also available in WDC’s JBOD hybrid storage platforms, integrating 60 to 102 drivers to offer up to 3.26PB of storage capacity. 

HAMR 

WDC is also moving into heat-assisted magnetic recording (HAMR) to scale to higher capacity drives, as ePMR tech is reaching its physical limits in areal density. This refers to how many bits can be packed onto each square inch of disk platter (where data is stored).  

Thus, future capacity gains must come from shrinking bit sizes further, which can be achieved via HAMR. HAMR uses a laser diode to heat a microscopic spot on the disk, enabling polarity of a single bit to be flipped to allow data to be written. This allows for substantial areal density gains, moving from <3 TB per platter under the current UltraSMR drives to up to  ~10TB per platter by the end of the decade.  

WDC acknowledges that its ePMR roadmap can extend up to 60TB, but after that, the shift to HAMR is all but inevitable in order to progress towards 100TB capacities and beyond.

Source: WDC 

WDC is harnessing its patented laser technology to not only reach 100TB drives over the next few years, but scale even further by adding more platters to drives:  

“So for the last 6 years, we've been working on our own patented laser technology. It solves for those 3 problems. By emitting more light, harnessing more of that light into the recording technology, we will increase the aerial density of the HAMR platters from 4 terabytes all the way to 10 terabytes by 2028 per platter. We have 11 platters. It's one of the reasons I'm confident about 100 terabyte HAMR drives by 2029. This technology is not theoretical. It's actually already in the labs. 

We've watched it do in the recording. The other part of it, as you can see from the micrograph, they're shorter. So it allows us to add yet more capacity per drive by packing up to 14 platters into the same 3.5-inch form factor. 10 terabytes, 14 platters, that sounds like 140 terabytes.” 

Accelerating ePMR and HAMR Roadmap 

WDC is accelerating its ePMR and HAMR roadmaps, noting that ePMR shipments are growing double-digits sequentially next quarter. HAMR qualification timelines are also accelerating by six months. 

Management explained in fiscal Q2’s call that ePMR shipments reached more than 3.5 million units, while starting qualification of both its upcoming HAMR and next-gen ePMR drives at two different hyperscaler customers. ePMR shipments were guided to reach closer to 4 million in fiscal Q3, or growth of more than 14% QoQ.  

On HAMR, WDC stated that they pulled forward qualification and started in February with one hyperscaler, with another expected to be initiated relatively soon. Management emphasized at its Investor Day in February that they are “so confident in our road map for HAMR. From last year, we said back in the '27 for the ramp, we pulled it in by 6 months. So it's ramping in the first half of 2027.”  

Inference as an HDD Driver 

We recently covered discussions over the role of SSDs in ‘warm’ storage and Nvidia’s upcoming Inference Context Memory Storage platform in our SanDisk analysis in detail. Simply put, Nvidia is essentially proposing SSDs to take a more central role in the ‘warm’ tier where HDDs sits – what Western Digital and competitor Seagate consider ‘nearline’, where data does not have to be accessed instantaneously but must remain readily available.  

We covered Seagate’s response to this proposal in our analysis, with executives believing there ultimately will not be much change to storage architectures with HDDs remaining critical from a TCO perspective to handle the massive volume of data generated by AI applications. Western Digital mirrored this view, explaining that Nvidia’s new initiative will likely at its core accelerate the growth of volume generated, which means HDDs will remain critical for AI storage requirements – management expects HDDs to still account for 80% of storage solutions deployed by hyperscalers over the next five years. 

However, management also pushed back on the economics of utilizing quad-level cell (QLC) SSDs for AI storage workloads, emphasizing that QLCs offer 10% of performance of HDDs for 10X the cost: 

“This performance has led some customers to think about using QLC flash to provide that performance in addition to hard drives. It's very attractive because — but QLC has a problem when the data is constantly moving as it does in the AI workloads and in the cloud workloads. QLC wears out. And that means you have to make a lot of changes in your software, so it doesn't wear out.  

Otherwise, you end up with silent data corruption. And that is just as scary as it sounds. Hard drives, on the other hand, just don't. They don't wear out. They'll operate for years without wear out and customers like that and it simplifies their code. The other reason to consider QLC flash is that the headline performance of a QLC drive is 6 gigabytes per second. That's way more than the 200 to 250 megabytes per second of a hard drive. But that's a headline number. It's only true when that drive, that QLC drive is attached directly to the GPU with a great big bus. In the real world, in object stores deployed in massive scale, that's not how it's done. Hard drives and QLC drives are connected to the network via a thing called the SAE interface.  

It's a thin pipe that takes data from the drive to the network. It can only support 530 megabytes per second. So customers would get less than 10% of the performance of QLC for 10x the cost of a hard drive. Do you think that's a good deal? I don't now or as a customer.” 

WDC also commented on more specific inference-based HDD demand drivers, such as multi-modal models requiring significantly large data sets to store queries and prompts, video generation, as well as autonomous vehicles and robotics needing extensive data sets to function in real-world situations. All summed up, WDC believes that inference and these upcoming AI applications will drive storage exabyte demand at a >25% CAGR over the next five years, providing a solid foundation for long-term revenue growth as HAMR unlocks a path to higher capacity drives.  

There is one risk with this, in that there will still likely be use cases and applications where customers use a higher mix of SSDs. HDDs offer the lowest-cost per terabyte, ideal for “big data” storage, backups and large AI datasets. However, solid state drives (SSDs), which store data on flash memory chips, are far faster and lower-latency but at a higher cost per terabyte. This means leveraging a mix of HDDs and SSDs is a popular choice, and some AI workloads prioritizing latency may take a higher mix of SSDs.  

High Bandwidth HDD Design and Power-Optimized HDDs 

At its recent Innovation Day, Western Digital shared more information on its product roadmap and technological innovations driving its path to 100TB capacity. WDC is not seeking just capacity gains but also performance gains with high-bandwidth HDD design. This is especially important as throughput (data transfer speeds) must continue to scale alongside capacity to prevent performance bottlenecks. 

This has not been the case over the last several years. WDC explains that since 2017, CMR HDD capacities have increased by 116%, from its 12TB Ultrastar to 26TB, yet maximum sequential throughput only increased 18%, from 255MB/s to 302MB/s. This is critical as a lack of throughput gains will make HDDs less suitable for bandwidth-intensive AI workloads despite strong capacity gains.  

WDC is aiming to boost performance and throughput via two innovations, High Bandwidth Drive (HBDT) and Dual Pivot design technologies. Put simply, WDC is working towards delivering NAND flash-similar performance and throughput but with HDD cost and TCO economics.  

WDC says that its HBDT will enable simultaneous read and write from multiple heads on multiple tracks, which up until now has been done on a single track at a time. Currently, HBDT is able to access two tracks simultaneously, though WDC believes it can ultimately scale to four and eight tracks simultaneously. Under two tracks, WDC says sequential throughput will double and will scale further as HBDT innovation progresses towards eight tracks at once. Importantly, HBDT is already in validation with customers. 

Dual Pivot tech (DPT) unlocks further performance gains by adding a second independent actuator on a separate pivot. Similar to how read and write were previously done on a single track at a time, HDDs have been limited by having one single actuator containing the read-write heads. With DPT, a second actuator is added on the opposite end of the drive, allowing for independent seeking between the two heads.  

This is the main difference between DPT and previous dual actuator drives. Previous drives had to remove on disk to make space for the second actuator to boost performance, sacrificing capacity, whereas with DPT and the placement of the second actuator, WDC can reduce spacing between disks, and boost capacity and performance at the same time. DPT remains in the lab for testing and is expected to be available in 2028. 

Combining both technologies is expected to drive significant performance and throughput gains while on the path to >100TB capacity. WDC states that “two-track HBDT plus dual pivot is projected to increase throughput from today’s 300MB/s to approximately 1.2GB/s, a 4x increase.” This would mean that a 100TB drive would have similar throughput and performance as the currently-available 26TB drives, preserving HDD performance and cost economics. WDC adds that eight-track HBDT and DPT could theoretically increase throughput to 4.8GB/s, another 4x increase from two-track and 16x versus today.   

Here’s how management explained this increase in performance, and this quote provides some perspective as to why demand may be high for drives integrating these technologies, as it can seamlessly integrate into existing infrastructure: 

The thing [customers] really love about this is that we can go to 4, 6 and 8x the performance. It is scalable. By the time we get to 100 terabytes, we could be 8x the performance of today's drives. We already have the technology to do it, and we're developing it, so we're ready for when customers are ready for it. We'll introduce this capability at the 50-terabyte mark to meet the customers' demand so that they are ready for us to consume and take advantage of all this performance. … 

This design will fit in an existing customer chassis with that change. It can be made on the same manufacturing lines. And they just see more performance. So customers are really excited by this. double the transactions smoothly for customers. … Dual-pivot technology helps customers focus their software effort on improving more performance for AI versus having to deal with how the hard drives are working. And we'll introduce this at the 60-terabyte mark. … 

My biggest problem is finding them enough material so they can start testing. Dual pivot technology will be in their hands in late '27 and '28. So all the performance that the customer's hardware can support their existing boxes, their existing software, their existing networks without having — can be delivered from these drives with the capacity that we are building without having to use QLC. We deliver performance 10x cheaper than QLC can.” 

Management also shed more light on software optimizations and how this will also accelerate deployment timelines for upcoming tech. WDC explained that the “shortcut is a simple, open API that allows customers to integrate that API to their existing file system, their existing object store,” that will be available on flash and extending to all HDDs scaling to 100TB. This will provide faster qualification and faster time to production, with this becoming available in 2027. 

WDC is also working on power-optimized HDDs, aiming to lower power consumption by up to 20% and boost capacity with minimal trade-offs on performance. The lower power consumption not only will offer a lower TCO, but also make these HDDs more attractive from a hyperscaler perspective by saving more power that can be allocated towards more GPUs.  

WDC says that by spinning the drive slower, “we can reduce the power by 20%, but we only trade 5% to 10% of the sequential I/O [input output operations], not something that the customers have seen before or even thought was possible.” This also will add ~10% more capacity, or 10TB for a 100TB drive, or increasing capacity inside the same 3.5 inch form factor without changing the drive’s power profile. WDC adds that this lower power consumption is optimal for cold data, or data that needs to be accessed within seconds for AI inference, helping enable AI data storage at scale. 

Multiple Margin Levers to Pull 

WDC outlined multiple different margin levers at its disposal, from strong yields with its ePMR products to increasing mix of UltraSMR drivers and soon the HAMR ramp. WDC has seen strong margin expansion, with Q2 gross margins up nearly 8 points YoY and operating margin following suit with a nearly 7 point YoY expansion. Perhaps most importantly, WDC expects to drive further gross margin expansion over the next couple of quarters and beyond, due to these levers.  

Management explained that yields for ePMR products are in the low-90% range, noting that “as we get yields up, cost continues to decline. As the UltraSMR mix goes up within those new products as well, that's also going to be a driver of cost down as well.” For context, UltraSMR drives now account for 50% mix in its nearline portfolio, with WDC’s top three customers onboarded with two to three other major customers likely moving to adopt the drives soon. This is expected to drive an increase in UltraSMR mix over the coming quarters and provide solid gross margin tailwinds. HAMR will be another margin lever in play once these drives begin to ramp, as WDC expects its HAMR products to be neutral or accretive to gross margins. 

Touching on the cost front, WDC explained that costs per terabyte were down around (10%) YoY in the quarter, which, combined with a 2-3% increase in ASP, provides room for solid margin expansion. This also pertains to higher capacity drives, which management explained have a cost benefit as well, which likely see lower unit costs, similar to Seagate.  

Analysts questioned about incremental gross margins considering the combined levers WDC has, with management explaining that they are currently running at a 75% incremental margin: 

“On the gross margin line, the guidance that you're giving for 47% to 48%, I guess the back of the envelope math would suggest that you're maintaining what looks to be like a 70%, maybe 75% incremental margin flow-through. So, I guess, my question is, how do you think about the durability of that incremental margin? 

EVP and CFO, Kris SennesaelEVP and CFO, Kris Sennesael 

We delivered 46.1% gross margin, up 220 basis points quarter-over-quarter, up 770 basis points year-over-year. And we are guiding to 47%, 48%, so 47.5% at the midpoint, which is up 740 basis points on a year-over-year basis. And, Aaron, I think your math is working. The incremental gross margin is on or about 75%, depending on how you look at it on a year-over-year basis or a quarter-over-quarter basis. So I've stated before, I'm very comfortable with an incremental gross margin higher than 50% and definitely 75% is higher than 50%. 

I mean in gross margins, there's two sides to the equation. On one hand, you have pricing environment. On the other hand, you have the cost environment. In pricing, I've talked about that before. We see a stable pricing environment with prices on a price per terabyte, kind of, flattish to slightly up. Actually, last quarter, it was up 2%, 3% on an ASP per terabyte basis. So that clearly demonstrate the value that we continue to deliver to our customers. 

And on the cost front, the teams continue to execute really well. We continue to upshift our customers to higher capacity drives, which gives us a cost benefit. And then there is great execution as well on driving down the cost in our manufacturing assets as well as throughout the supply chain. And when you look at it last quarter, the cost per terabyte was coming down on or about 10% on a year-over-year basis. And so when you put this all together, we continue to drive further gross margin expansion. And we believe in the next couple of quarters and beyond, we will continue to be able to do that.” 

What the incremental margin means is that for every dollar of revenue that WDC adds compared to the prior quarter, 75% of that flows through to gross margin – WDC delivered a $199 million QoQ increase in revenue this quarter, with $153 million of that flowing to gross profit. Looking ahead, if WDC can maintain this and deliver >$200 million QoQ growth throughout CY26 (as current estimates suggest), it could exit the year with margins above 50%. 

Visibility through 2027-2028 at Top Customers, Pricing is Stable 

WDC has already signed firm orders through calendar 2026 with its top seven customers, though it has visibility into demand extending out into 2028. Management explained that they have ‘robust’ agreements with three of its top five customers, with two of these covering calendar 2028 and one extending through calendar 2028, with these LTAs including both volume and price conditions.  

Analysts asked questions about WDC’s long-term agreements and what the economics of these new orders would look like. WDC revealed that they do contain conditions for volume and pricing, but added quite an important caveat at its Investor Day – hyperscalers want ‘predictable’ pricing, in sharp contrast to rapidly rising (and fluctuating) SSD prices. This means that upcoming contracts will likely have set price escalators so hyperscalers do not get exposed to rapidly changing memory cost structures: 

Irving, just given the tightness of the HDD market and kind of the significant inflation that NAND is going through right now, can you maybe just talk about maybe your patience in being able to sign purchase orders further into calendar '27 to extract better economics just relative to maybe how you were approaching signing POs last year? Is that making any difference in the economics you're able to extract?  

CEO Irving Tan 

As we highlighted, we're pretty much sold out for calendar year '26. We have firm [purchase orders] with our top 7 customers. And we've also established LTAs with two of them for calendar year '27 and one of them for calendar year '28. Obviously, these LTAs have a combination of volume of exabytes and price. And in relation to pricing, I think first, it's important to recognize [that] there's actually a structural shift in the value that we deliver to them, especially in the impact that we have to their total cost of ownership as the business moves more and more towards inference where monetization is happening.  

So, in this case, the pricing that we've provided there reflects the value that we're delivering to them. And so as Kris mentioned, we continue to see going forward a stable pricing environment that gives us an opportunity to continue to extract more value as we deliver both better TCO value to our customers and to better support their supply-demand needs as well through higher capacity drives. 

This ties in to comments from Investor Day, where management confirmed that they want to ensure there is a fair value exchange between capacity/performance and pricing, noting that the goal is to deliver “predictable pricing” as hyperscalers are concerned about “high volatility of some tiers of the storage space,” referring to the strong QoQ growth in enterprise SSD prices. 

Management also offered some more insights onto pricing trends through 2026, noting that FQ2 had seen stable prices with ASP per terabyte up 2-3%, while projecting mid-to-high single digit YoY ASP growth through the year and remaining stable into 2027: 

“If I look at calendar year 2026, 4 quarters of calendar year '26, I do expect ASP per terabyte to go up mid- to high single digits year-over-year for all 4 quarters, mid- to high single digits year-over-year for '26. And then beyond '26, I expect us to continue to operate in what I call a stable pricing environment, of course, from the higher level that we established in '26.” 

Battling Head to Head with Seagate on HAMR, but Expects to Lead 

It’s safe to say that WDC faces fierce competition with Seagate in the HDD space on both exabyte shipments and on the tech front. WDC exhibits a fair lead in exabytes, shipping 192 nearline exabytes in the quarter versus Seagate’s 165 exabytes. However, Seagate current is ahead when it comes to ramping HAMR drives. 

For an apple-to-apple comparison to Seagate, WDC’s first HAMR products are expected to launch with capacities from 40-44TB, which aligns with Seagate’s Mozaic4+ platform also offering up to 44TB, as both feature 4TB capacity per platter. The key difference is that Seagate’s Mozaic4+ is now shipping to two hyperscalers and ramping through the rest of the year, representing a three quarter head start over WDC in the first half of 2027. 

However, Western Digital expects to quickly take the lead in HAMR-drive capacity, with management confident in achieving 100TB+ products as early as 2029. This goes back to our discussion on the HAMr under the product overview, with WDC aiming to reach 10 TB per platter by 2028 to enable shipping 100TB HAMR drives by 2029.  

One reason that WDC is able to move so fast at scaling capacity is that it begins qualification processes at hyperscalers when introducing new hardware into the labs, cutting ramp timelines by months: “Together with our customers, we started to introduce the hardware into our labs and the qualification of the hyperscaler software so that we can start the qualification process while we're still developing the drive.  

That cuts out months from a qualification process. So by the time we're ready for volume manufacturing, the drive is ready to ramp. As Irving said earlier, we've ramped up our latest generation of drives very quickly. That's one of the reasons. The other part with rapid generation of more capacity points, customers will have a lot of qualifications. So instead of qualifying every single capacity point, they qualify one set of capacity point, let's say, 36 to 41 terabytes, and we will just ship them more capacity as we make it available. One qualification, many capacity points.  

And the next one is going to be at 42 to 56 and so on. So that innovation, not just in the drive design, but also in the processes we do gets us faster time to capacity in customers' hands in their fleets where they need it the most. So putting it all together, our HAMR capacity goes from 40 terabytes to 100 terabytes by 2029.” 

On the flip side, Seagate has that capacity penciled in for the early part of next decade, noting in Q2’s call that its Mozaic3+ and Mozaic4+ “developments align with our long-term areal density road map that extends to 10 terabytes per disk, which we expect to deliver early in the next decade.” This shift will be key to watch, as it suggest that Seagate could soon lag WDC’s capacity roadmap by several years.  

In plain terms, WDC could have a significant capacity and TCO advantage by 2029 if it can realize these density targets, making it a more attractive solution for AI data storage needs 

Financials 

Revenue Growth Accelerating to 40%, Potentially Peak Growth 

WDC delivered fiscal Q2 revenue of $3.02 billion, up 25% YoY and 7% QoQ, although this marked a deceleration from 27% YoY and 8% QoQ in Q1. Management said this was driven by strong demand for its higher capacity nearline drives.  

Exabyte shipments rose 22% YoY and 5% QoQ to 215 EB, including 103 EB from its ePMR product line on 3.5 million shipments. Revenues are closely correlated with EB shipments with pricing contributing only a 2-3 points of upside.  

Looking ahead to FQ3, WDC guided for revenue of $3.2 billion at midpoint, representing an acceleration to 40% YoY growth, up 15 points, while QoQ growth would tick slightly higher to 6.1%. Current estimates suggest WDC will close out the fiscal year with $3.44 billion in revenue in Q4, up 32% YoY and another 7.5% QoQ. At present, this would suggest that fiscal Q3’s YoY acceleration to 40% would be the company’s peak growth quarter moving through calendar 2026. 

For FY26 ending in June, revenue is expected to increase 31% to $12.47 billion, while an initial look at FY27 points to 25.6% growth to $15.66 billion in revenue. This is above WDC’s long-term revenue growth model for >20% growth CAGR over the next three to five years.  

Key Segments 

Similar to Seagate, WDC sees the large majority of its revenue go to Cloud customers, with minimal exposure to Client and Consumer end markets. 

Cloud revenue was $2.67 billion in Q2, up 27.5% YoY and 6.5% QoQ, decelerating from 31.5% YoY and 7.8% QoQ in Q1. Management noted that they are seeing strong demand from hyperscaler customers. Cloud accounted for 89% of revenue in the quarter.  

Assuming similar Cloud mix at 89% in fiscal Q3, though there is potential for slight mix gains on seasonality in its other two end markets, Cloud revenue would be implied at roughly ~$2.85 billion, up 6.5% QoQ and 41.9% YoY. 

WDC’s Client segment represented 6% of revenue at $176 million, up 26% YoY, while Consumer accounted for 5% of revenue at $168 million, declining (3%) YoY. 

Strong Margin Expansion 

As discussed above, WDC is seeing strong gross margin expansion with multiple levers available, and this margin expansion is flowing down the line.  

GAAP gross margin was 45.7% in Q2, up 8 points YoY and 2.2 points QOQ, while adjusted gross margin was 46.1%, up 7.7 points YoY and 2.2 points QoQ. For Q3, adjusted gross margin is expected to be 47.5% at midpoint, up 7.4 points YoY and 1.4 points QoQ.  

GAAP operating margin was 30.1%, up 6.9 points YoY and 2 points QoQ, while adjusted operating margin was 33.8%, up 9.3 points YoY and 3.4 points QoQ. Adjusted operating margin is implied to be 35.5% in Q3, up 9.5 points YoY and 1.7 points QoQ, showing a slight degree of operating leverage.  

GAAP net margin was 59.7% as WDC benefited from a more than $1 billion gain on its 7.5 million share stake in SanDisk, which it has now sold for nearly $3.2 billion to help reduce debt. Adjusted net margin was 26.7%, up 9.3 points YoY and 3.5 points QoQ. 

For a quick view on how current margins stack up to WDC’s long-term model, it expects adjusted gross and operating margins to surpass 50%/40% over the next three to five years, or roughly three to five points of expansion.  

EPS Growth Robust but Decelerating 

Driven by the margin expansion, WDC is seeing strong EPS growth, though growth is expected to largely decelerate moving through the year. 

GAAP EPS was $4.73, up 54% QoQ and 272% YoY, again driven by the gain on its SanDisk stake. Adjusted EPS was $2.13, up 78% YoY and 20% QoQ, beating estimates for $1.98. 

For Q3, adjusted EPS was guided to be $2.30, +/- $0.15, up 69% YoY and 8% QoQ at midpoint. Adjusted EPS growth is expected to decelerate further to 61% YoY in FQ4, and slow to 50% by the end of FY27. 

For FY26, WDC is expected to report adjusted EPS growth of nearly 80% YoY to $8.85, with another 52% to $13.46; despite the deceleration on the quarterly view, it should be noted that this EPS growth rate is still quite strong compared to the base it is growing from. Under the long-term model, management expects EPS to surpass $20. 

Cash and Balance Sheet 

Cash flow margins also showed strong expansion, and while debt does look inflated, WDC’s sale of its SanDisk stake is meaningfully improving its debt profile. Management explained at Innovation Day that “if you look at the revenue growth that we've delivered over the last 12 months, the margin appreciation, ultimately, what is done is to translate into very strong free cash flow. And in the last 2 quarters, we returned 100% of the free cash flow that we've generated. And we also were able to bring our net leverage well below the 1 to 1.5x that we laid out last year.” 

Operating cash flow was $745 million in Q2 for a 24.7% margin, up from 16.9% a year ago and 23.8% in Q1. Free cash flow was $653 million for a 21.6% margin, up from 13.9% a year ago and 21.3% in Q3. 

As of Q2, WDC reported cash of $1.98 billion and debt of $4.65 billion, though it already has redeemed $1 billion in its 2029 and 2032 senior notes following the stake sale. 

Inventories were $1.35 billion, down marginally from $1.39 billion in Q1. 

Conclusion 

Western Digital has emerged as one of the top AI winners over the past year with a 591% return, with the company getting added to the NASDAQ-100 index in December of 2025.  

Looking ahead to 2026, Western Digital is already sold out of capacity with set price and volume commitments, with visibility from key hyperscaler customers extending into 2027 and 2028. WDC also has been excelling at driving strong margin expansion with a handful of key levers at its disposal.  

Over the next few years, WDC is leveraging its technological expertise and several new innovations such as HBDT and DPT to drive significant performance gains on the path to 100TB capacity. WDC is also expecting to quickly take the lead in higher-capacity HAMR drives versus key competitor Seagate, potentially giving it a key advantage by the turn of the decade in meeting AI-driven data storage demand.

Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.

Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.

Recommended Reading:

• Ciena: Benefitting from Data Center Scale Across Demand
• Seagate: Slow QoQ Data Center Growth, 2027 Capacity Under Discussions
• Monolithic Power: Strong AI Tailwinds to Drive 50% Enterprise Data Segment Growth in FY26
• Teradyne Q4: Revenue Accelerates to 41% QoQ, Possibly Peak Growth Quarter

Ciena’s role is different than the many “inside-the-data-center" networking stocks the I/O Fund covers as instead of offering networking fabrics at the system level, Ciena offers optical transport across data centers and metro regions. While AI training is already pushing hyperscalers to connect clusters across data center sites, inference offers another leg up for metro/long-haul networking as workloads will be distributed across many data centers to scale significant application usage, especially as agents necessitate responsiveness and reducing latency. 

This quarter, Ciena proved it's a leader in data center interconnects with $1.43 billion in revenue and $1.35 in adjusted EPS – more than doubling year-over-year. Optical revenue grew 41% YoY and 10% QoQ with Waveserver and RLS up 80% YoY and backlog that grew by $2 billion to reach about $7 billion total. 

With 800G pluggables ramping, new platforms like RLS Hyper-Rail and Nubius products (Vesta Optical Engine and Nitro redriver) will help to expand Ciena’s footprint. Notably, Ciena benefits from two major customer pools: hyperscalers building and linking more AI data centers which drives demand for DCI/metro/long-haul optical networking, but also telecom carriers and service providers who are upgrading their networks in anticipation of increased AI traffic across the network.  

Below, we cover Ciena’s product positioning, its most recent earnings report and a few key catalysts to keep an eye on as we approach an inflection driven by the inference market. 

Product Overview: 

Ciena’s foundational business is to address connectivity needs in the wide area network (WAN), spanning long-haul, metro and data center interconnect (DCI). Ciena offers the optical backbone that connects AI data centers and regions and is poised to benefit as AI-driven traffic will lead to higher-capacity metro and long-haul upgrades. 

Below, we break down the main products that Ciena offers currently, such as RLS and Waveserver, and products that Ciena plans to expand into, such as Vesta higher-density optical engine and Nitro linear redriver to extend reach in active copper cabling. 

Optical Networking & DCI 

Ciena offers a range of optical networking solutions targeting scale across, campus and metro DCI (10-20km and 20-100km) to backbone and submarine (10,000+ km reach) optical systems.  

Ciena’s WaveLogic sends information over fiber using fewer lanes, which lowers power use and cost. Ciena uses WaveLogic inside its systems and offers compact pluggable modules that make it easier to add capacity without rebuilding a networking. Ciena’s Waveserver is the platform that connects data centers at scale by offering a packaged solution that can be deployed to expand bandwidth between sites quickly. The newer versions of Wavelogic detailed below allow Ciena to push higher speeds over the same fiber, which can reduce the number of components needed for a quicker and more simplified upgrade.  

• Ciena’s WaveLogic 5 Extreme supports 200G to 800G line rates to enable 400GbE transport across any distance, including ultra-long haul and subsea, while its WaveLogic 5 Nano 100G to 400G ZR/ZR+ pluggable coherent optics are optimized for DCI applications with low power consumption of 15W with reach up to 140km. WaveLogic pluggables also seamlessly integrate into its Waveserver stackable interconnect platforms, scaling up to 12.8Tb/s per 2RU unit for cost-effective DCI solutions. 
• Ciena’s WaveLogic 6 Extreme boosts speeds to an industry-leading 1.6Tb/s per wavelength, extending 800GbE transmission across the longest distances. Ciena says that WL6e delivers twice the capacity per wavelength in its existing 6500 and Waveserver platforms, helping reducing transceiver counts and lowering cost and complexity of systems. It also delivers double the capacity under the same power and space envelope versus WL5e. 
• WaveLogic 6 Nano offers fit-for-purpose 400-800G ZR/LR/ZR+ pluggables and 1.6T coherent-lite pluggable transceivers to meet high-bandwidth AI data center applications. Ciena says that WL6n combines it vertically integrated DSP, electro-optics, and high-speed packaging to support high-performance data center fabric, or 2km, and data center campus, up to 20km, applications. 800ZR pluggables were implied to be ramping later in the year, while Ciena is also developing future 1600ZR/1600ZR+ solutions built on 2nm silicon. 

Reconfigurable Line Systems (RLS) 

RLS offers the infrastructure that routes wavelengths over fiber for metro and long-haul networks. COLORZ pluggables are part of the optical stack and are the endpoints that generate and receive the optical signal for DCI/metro links. 

RLS plays an important role in ‘scale-across’ applications, the third pillar of scaling AI infrastructure. We have detailed the first two pillars, scale up and scale out, in detail, yet the challenge is that both cannot extend infinitely, which is where scale across fits in.   

While scale across is much less discussed, it refers to linking together multiple geographically distributed AI data centers to act as one cohesive compute facility, theoretically facilitating a path to >1GW-scale clusters as it bypasses site/land and power constraints.  This can either be in close proximity such as OpenAI’s Stargate cluster in Abilene spanning multiple buildings, Amazon’s Project Rainier or Microsoft’s Fairwater data centers linking together via ‘AI WAN’ (wide area network). Coherent ZR/ZR+ modules excel in scale across as they maintain the signal integrity and reliability across long distances at the highest data rates.  

According to the management team, Ciena’s upcoming RLS hyper-rail solution is expected to be in high demand: “We are addressing this demand for scale across solutions with our RLS platform, the de facto industry line system standard for cloud providers as well as our 800ZR pluggable optics. To underscore this, we realized a second consecutive record quarter for RLS shipments and revenue. We expect to expand our role in scale across applications with the introduction of our new RLS hyper-rail solution. Hyper-rail delivers an order of magnitude increase in fiber density within existing rack footprints, helping customers scale traffic while reducing and, in some cases, avoiding costs and complexity associated with adding substantial numbers of amplifier huts.” 

• Ciena offers reconfigurable line systems (RLS) in its 6500 RLS family, which addresses the highest-capacity bandwidth requirements in metro, long-haul and DCI applications. RLS can support 60Tb/s capacity on a single fiber pair, offering flexible configuration options while doubling fiber capacity without impacting existing C-band traffic. Ciena also says that its 6500 RLS can reduce physical footprint by as much as 70% versus traditional chassis-based systems from its modular form factor. 
• Ciena is expanding its RLS portfolio with hyper-rail photonics, which feature new amplifier configurations to deliver 32x density of current solutions, scaling to 128 fiber pairs in a single rack. This is achieved by scaling from one rail per module to four, reducing power consumption by up to 75% and space requirements by up to 85% while integrating into existing rack footprints. Ciena explained that the RLS Hyper-Rail was co-developed with its hyperscaler customers specifically for multi-region, large scale AI connectivity, and will begin standardization at the end of 2026 and ramp in 2027. Ciena expects to be first to market in hyper-rail, allowing them to extend current RLS momentum into 2027 driven by share gains. 

RLS is an important cornerstone for Ciena’s Optical Networking portfolio, with management explaining that FQ1 marked a “second consecutive record quarter for RLS shipments and revenue”, with growth of over 80% YoY. Ciena is seeing robust momentum for scale across stemming from its RLS solutions and ZR/ZR+ modules this year, with management explaining that they “are already experiencing extraordinary demand with 3 hyperscalers choosing to use our optical solutions for their training applications across distance.” All three hyperscalers were said to be “significantly ramping including additional orders for multiple additional clusters from the first hyperscaler we announced in Q3 2025.” 

Scale Up, Scale Out and CPO 

Ciena’s $270 million acquisition of Nubis in September 2025 expanded its presence inside the data center, as the aforementioned products primarily offer networking across the data center. With this new acquisition, Ciena is moving into scale-up and scale-out networking and co-packaged optics (CPO).  

As a brief recap, CPO places optical transceivers directly on the chip package, rather than using separate optical modules, resulting in faster data transmission, reduced latency and higher bandwidth. This may be the best of both worlds: the performance of optical yet with reduced power consumption for increasingly power-hungry AI racks. 

Ciena recently unveiled its Vesta 200 6.4T CPX, which it says is the industry's highest-density and lowest-power pluggable CPO solution. Ciena says that the retimer-free linear-drive operation saves up to 70% power versus other retimed options. According to management, samples of the Vesta 200 6.4T optical engine will be available in Q2 2026. 

For scale-up inside the rack, Ciena is advancing Nubis’ Nitro Linear Redriver tech, which extends the distance that signals can be transmitted over copper cables and also reduce power by up to 80% versus AECs. Ciena will also sample the redriver in calendar Q2. As stated on the call, this is especially important for XPUs (such as from Broadcom): “For scale-up opportunities inside the rack, where XPUs are getting faster and driving heat and power concerns, we are advancing the Nitro Linear Redriver technology also from our Nubis acquisition.” 

Data Center Out-of-Band Management  

Out-of-band management is critical yet underdiscussed, as it provides a separate, independent network to control and recover data center infrastructure – for example, it allows hyperscalers to have visibility over servers and other components, and provide a pathway remotely recover systems without physical intervention. However, traditional out-of-band management faces increasing challenges as AI clusters scale up and out. This is because the maximization of GPU density in racks leaves minimal space for OOB connectivity, such as for switches, routers and console servers. It also is copper-heavy (and power hungry) by relying on extensive Ethernet cabling.  

Ciena designed its DCOM solution with Meta to meet hyperscaler configuration requirements, noting that its DCOM solution, leveraging its routers, switches and its passive optical network (PON) ecosystem to reduce rack space by up to 99%, reduce power, and streamline remote access. Ciena says it is continuing to work with Meta on DCOM and had expanded last quarter, with two other hyperscalers in discussion. Management believes that DCOM will be a significant opportunity in the data center and a defensible space for them due to its speed, vertical integration and software.  

Managed-Over-Fiber Networks (MOFNs) 

Advanced networks called managed-over-fiber-network (MOFNs) are projects that connect multiple data centers in a metro area and upgrade capacity and resiliency on existing metro networks. This drives demand for Ciena’s optical transport and DCI solutions, especially across Waveserver, coherent optics and RLS amplifier and routing gear.  

MOFNs are starting to show up as a growth driver because AI is creating significant traffic that has to be distributed between data centers. The customer base includes hyperscalers, neoclouds and service providers with 10% to 15% coming from telecom. 

Ciena is seeing strong momentum in MOFN, noting that its orders in India were up 40% YoY, specifically for these solutions. According to management, there were three greater than 10% customers including two global cloud providers and one Tier 1 North America service provider with strong MOFN activity and “order intake has been incredibly strong over the past 90 days, leading to a new record by a significant margin.” 

There was an interesting quote from management about MOFN and the neoclouds, as Ciena hinted that this is a preferred method of connectivity from its speed to market and less capex requirements from a lack of maintenance on the neoclouds’ part:  

“We're seeing obviously an emerging ramp here around a bunch of the loosely called sort of neo-scalers, which encompasses a fair range of different players. It's, I would say, largely right now MOFN orientated, given the capital expenditures, time to market for them, et cetera. But what is clear from it all is that the network is now a real priority for them. And I think that plays through to the hyperscalers, too. There's been such a maniacal focus and continues to be, obviously, on things like power, GPU accessibility, et cetera, et cetera. 

Now it's really about the network. The traffic is beginning to come out of the network, both for inference and for training. And the neo-scalers are obviously seeing that, too. So they're leaning in on the network. Now we're also beginning to see some of them wish to have control of some of that network as well and do their own builds. We're cautious about that approachment given the financial structure of some of those neo-scalers, not all of them.” 

Backlog Rises 40% QoQ to $7 Billion, Orders Fulfilled in FY27 

Ciena revealed strong backlog growth in fiscal Q1 and robust order activity tied to strong demand from hyperscaler customers. The backlog growth sparked multiple important discussions around conversion and pricing power, suggesting Ciena has strong visibility into next year with pricing emerging later this year as a growth lever.  

Backlog rose 40% QoQ to $7 billion, with management noting that Q1 demand was ‘unprecedented’ with a very strong order intake. Roughly 80% of the backlog is for products and software, or ~$5.6 billion, up from $3.8 billion last quarter. Management would not put a concrete number on RPO, but said that it would be roughly 60% of the orders Ciena took in during Q1.  

Ciena expects its backlog to continue to grow throughout the year yet noted that “nearly all” of the new orders it is currently receiving will be fulfilled in fiscal 2027, suggesting that supply remains extremely tight relative to demand with this providing strong visibility for next year’s growth.  

CEO Gary Smith provided some color on the demand drivers and strength of demand, explaining that Ciena is “seeing growth in cloud, general cloud, you're seeing inference, you're seeing this new market of training now emerge. As I said in my comments, we've now got 3 hyperscalers deploying us for training, and we're at the very early stages of that. 

So you put all of that together, and that yields the incredible demand that we saw in Q1. And as Marc said, despite the fact that we're ramping our capacity for delivery as seen in our results, demand is going to continue, we believe, to outstrip our ability to supply. And that's going to continue for — we believe, this year. And so we're going to end up with a larger backlog than we have right now as we turn the year despite the fact that we're ramping our capacity strongly throughout the year and obviously through '27 and '28.”  

Smith also later clarified that it was purely demand that was driving this order growth and backlog increase, explaining that hyperscalers are ramping significantly in optical technologies both across clusters and within racks.  

Pluggables to Triple to >$500M 

Also intertwined with the scale across momentum is Ciena’s projections for hypergrowth to continue within its ZR/ZR+ pluggables business, where it is expecting its pluggable revenue to triple after doubling last year.  

For context, Ciena stated in Q4 that pluggables had reached more than $168 million in FY25, so management’s guidance implies a ramp to >$500 million this year. This would represent a ramp from ~5% of Optical Networking revenue to likely low-double digit share in FY26 for pluggables, making them a key contributor to the segment and overall topline growth.  

Management offered some commentary on pluggables and the drivers behind this growth, pointing to 800G ramping through FY27 and their expectation to lead the market with the new data rate:  

Amit Daryanani, Evercore 

“How do you see the pluggables market, especially with 800G ramping up through fiscal '26 and '27? And if you just maybe compare and contrast a bit about your position in 400 versus 800, that will be helpful as you go into the next cycle. 

Scott McFeely, Executive Advisor 

So we've seen pluggable revenue increase sort of period-over-period, and we've talked in the past about our interconnect business and when we went from 2024 to '25, that doubling. That's sort of in the rearview mirror. And then we talked about it as a major portion of our inside and around the data center with our aspirations to triple that this year, and we're well on track for that. 

So we do see significant growth from a competitive perspective, as we've talked about in the past, through choices that we made to focus early introduction of the technology in the last generation more on our systems business and our pluggable business because that was the bigger opportunity. We weren't necessarily first movers in that market. So that probably cost us some share and it probably cost us actually, frankly, some margin dollars. That's not the case in the 800G. We're first to market there and 800G is moving quite along.” 

800G is expected to ramp through the first part of 2026 and persist through 2027, and while the tripling of revenue is certainly welcomed, the upcoming cycle is expected to see strengthening margin tailwinds through the course of the year. Ciena explained last quarter that they are expecting to lower 800G unit costs over time, and as the ramp progresses through Q2 to Q4, they expect significantly lower costs versus the start of 2026. This suggests that 800G margin headwinds in the initial early ramp will quickly pass through and potentially shift to margin tailwinds by year-end on a much larger revenue base. 

The primary challenge with pluggables was touched upon by management, in that the market is competitive, any first-mover advantages may be fleeting and pricing power is not a given. For example, Marvell recently unveiled its 1.6T ZR/ZR+ module, the first to come to market, with the new product expected to be initially available in the second half of the year, meaning Ciena’s lead in 800G may be short lived.  

Not Aggressive on Pricing, but Tailwinds Emerging in 2H 

Given the demand environment and supply tightness, analysts were curious about pricing power, noting that other component suppliers are being much more aggressive in raising prices and repricing their backlogs higher, and if Ciena would do the same. 

Ciena has already been upfront about demand outstripping supply for at least the next several quarters and how supply constraints have impacted revenue growth, meaning a more aggressive stance on pricing could drive revenue growth at a much faster pace.  

However, management downplayed the need or desire to reprice the backlog and be aggressive with pricing, explaining that they are already seeing market share gains and tangible impacts to growth and margins. This is likely because Ciena has already raised prices and expecting these to begin landing in 2H: “As we disclosed in Q4, right, the pricing increases that we talked about were really on the new orders. And because we had such a big backlog at the time, most of that was going to be seen in the second half. So you should expect those price increases to show up in Q3 and Q4.” 

Management was also upfront about why they are opting to not be as aggressive as other suppliers when it comes to pricing, rather using it as a lever to get better contractual terms, better cash conversion and longer-term purchasing commitments to reduce the risk associated with converting its larger backlog.  

George Notter, Wolfe ResearchGeorge Notter, Wolfe Research 

“Obviously, you're raising pricing. I know it's going to come through later in the year as you eat down the backlog. But just stepping back and thinking about the space, you've got higher memory costs, you've got component suppliers that are being really aggressive on price. They're repricing their own backlogs. It just seems like it's an environment where you guys could be more aggressive on price and even perhaps reprice your own backlog. So I'm just curious like why not be more aggressive here given the supply-demand dynamics and what's going on in the supply chain? 

Gary Smith, Ciena CEOGary Smith, Ciena CEO 
We've talked a lot about the good things that we're doing to manage our margins and the rest of it, including the value rebalancing. But it is a balance to it all. And that's what we're trying to strike as we go through this. I mean you're seeing it translate into improved financial performance in all dimensions, market share gains, revenue, gross margin improvement and operating leverage. We're seeing that. And it's a confluence of things….  

Marc Graff, Ciena CFOMarc Graff, Ciena CFO 

No, I think Gary said it well. Pricing is a lever, George, but we're also looking at can we improve cash conversion? Can we get better terms and conditions? Can we get longer-term purchasing commits with maybe some more noncancelable, less risky terms as we satisfy this quite large backlog.” 

Ciena is already showing signs of healthier payment activity within its days sales outstanding (DSO), which declined to 72 days in Q1, down from 77 days in Q4 and 90 days in the prior year quarter. Additionally, inventory turns reached 3.2X in Q1, up from 3.1X in Q4 and 2.3X in the prior year quarter, suggesting that demand is strengthening and payment terms are improving. Aiming for longer-term contracts and less risky terms would provide a higher degree of visibility into revenue through 2027 and will prevent Ciena from falling back into an inventory overhang like it had faced in early 2024.  

Financials 

Revenue Accelerates to 33.1% YoY in Q1  

Ciena reported fiscal Q1 revenue of $1.43 billion, representing 33.1% YoY growth, while sequential growth came in at 5.6% QoQ. Revenue also beat consensus estimates by 2.1%, reflecting stronger-than-expected demand across optical networking and cloud connectivity deployments.  

The strong YoY growth marks a notable improvement compared to the muted growth environment seen through much of FY24 and early FY25 as telecom customers worked through inventory digestion and moderated capital spending. The current acceleration suggests spending patterns across service providers and cloud operators are normalizing, particularly as network upgrades tied to AI infrastructure and high-capacity data center interconnect (DCI) deployments expand. 

Looking ahead, Ciena guided for Q2 revenue of approximately $1.50 billion, which would represent 33.5% YoY growth and a modest 4.9% QoQ growth.  

For the full year, Ciena guided for fiscal 2026 revenue of $5.9 to $6.3 billion, raising its growth forecast from 24% YoY to 28% YoY at midpoint. While the raise is certainly welcomed and indicative of the strengthening demand Ciena is seeing in RLS and pluggables, it indicates a rather soft 2H, after 40% and 34% growth in 1H. This suggests that the order momentum and backlog building are not translating over to 2H growth as of yet, but more so for 2027.  

Margins Expanded, Ahead of Guidance 

Profitability improved across all levels during the quarter, with margins expanding sequentially and outperforming management’s guidance in several areas. 

GAAP gross margin came in at 43.8%, generating $625.5 million in gross profit. Gross margins benefited from improved product mix and stronger volumes across higher-capacity optical platforms.  

On an adjusted basis, margins were slightly stronger, with adjusted gross margin of 44.7% on adjusted gross profit of $638.2 million. This result exceeded management’s midpoint guidance of 43.5%, suggesting better pricing discipline and favorable mix within Ciena’s optical portfolio. The company’s WaveLogic coherent optical platforms typically command higher margins, particularly as customers transition toward higher-capacity wavelengths such as 800G and beyond. Looking ahead, Ciena expects that its upcoming products such as RLS hyper-rail and more focused cost optimizations will help deliver improvement in gross margins. On the call, management explained that ” moving forward, you'll see even more aggressive cost reductions. And then the price increases that we talked about at the end of last year, those really haven't started to fully kick in until the second half of the year. So I think that creates additional tailwinds for us. So all in all, again, I think we're making really good progress towards that 45% way point, and you should see that throughout the year.” 

Operating leverage became more visible in Q2 as operating margins improved meaningfully in the quarter, reflecting the benefits of stronger revenue growth flowing through the cost structure. GAAP operating margin was 13.3%, improving 5.8 points YoY and 12.5 points QoQ, and marking Ciena’s first return to above a double-digit margin since the end of 2021. 

Adjusted operating margin was 17.9%, coming in well ahead of management’s 16% guidance. Because Ciena operates with a large fixed-cost R&D base tied to its optical platform development, revenue expansions typically translate into disproportionately stronger operating profit growth once demand improves. Operating leverage will remain a key driver of earnings growth if revenue continues expanding at a high double-digit pace over the coming quarters. 

This improving operating profitability flowed through to net income as well, with GAAP net margin of 10.5%, up 6.3 points YoY and 9.1 points QoQ. Adjusted net margin was 13.8%, with the difference reflecting stock-based compensation and other non-cash adjustments.  

EPS 

Ciena reported a strong 15.6% beat to adjusted EPS estimates in Q1, earning $1.35 in the quarter, up 110.9% YoY. GAAP EPS was thinner at $1.03, up 232.3% YoY on a softer comp. 

Looking ahead, Ciena is expected to see EPS growth accelerate in Q2 on a softer YoY comp, with growth decelerating through the back half of the year but remaining strong. Adjusted EPS is projected to rise 247% to $1.46 in Q2 before decelerating to 134.6% YoY to $1.57 in Q3. Growth is expected to reach 94.1% by Q4 to $1.77. 

For the full year, Ciena is currently projected to report 132.4% growth to $6.14, before decelerating to 34% growth in FY27 to $8.22. 

Cash Flow Generation Remains Healthy 

Cash flow generation remained solid in the quarter and improved alongside stronger profitability. Ciena’s balance sheet remains solid with a manageable leverage profile and ample liquidity. 

Operating cash flow was $227.7 million for a 16.0% margin, improving from a 9.7% margin a year ago. Free cash flow was also positive at $153.8 million, representing a 10.8% margin, up from 7.2% a year ago.  

These cash flow margins remain healthy for a hardware-driven networking company, particularly one with significant R&D investments in optical technologies and silicon photonics. Over time, improving operating leverage and scale could push free cash flow margins higher, particularly if higher-capacity optical products continue gaining share within the product mix. 

Cash and short-term investments totaled $1.30 billion, while total debt was $1.54 billion. 

Notably, capex was $74 million in Q1, which is 2X to 3X more than the company’s average capex over the past three years. The trend toward higher capex is likely to continue given a mix of strong visibility from customers and the need to secure critical components ahead of demand. Here is what was stated on the earnings call: “Second, we are deeply engaged with component vendors, which is where more of the industry challenges exist to secure and expand supply, including through responsible long-term purchase commitments. As shown by our Q1 results, we are navigating the supply environment well and are investing to expand capacity. However, we expect demand will continue to outstrip supply at least for the next several quarters.” 

Conclusion 

Ciena’s quarter reinforced the company is a direct beneficiary of AI-driven connectivity spend – not just inside the data center, but across metro and long-haul networks that connect AI clusters and regions. Ciena was also recently included in the S&P 500 – and for good reason as the $50B+ market cap company is set to report over $6 billion this year in revenue with a bottom-line that is doubling. 

Record revenue and expanding profitability, plus a backlog that pushes into fiscal 2027 all point to demand that is both strong and increasingly visible. The company’s 800G pluggables are ramping, and new platforms like RLS hyper-rail and Nubis products are set to expand Ciena’s footprint ahead of more long-haul/metro demand from distributed inference and multi-site training. 

Although our portfolio is loaded with market-leading AI networking winners, with some up as much as 100% YTD, Ciena holds its own by offering a differentiated way of participating in the next phase of AI infrastructure, where the focus shifts to scaling the bandwidth that connects data centers and regions.

Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.

Every Thursday at 4:30 pm Eastern, the I/O Fund team holds a webinar for premium members to discuss how to navigate the broad market, as well as various stock and crypto entries and exits. Beth Kindig offers weekly deep dives including lesser-known cryptocurrencies and AI stocks, plus the team offers trade alerts. The I/O Fund team is one of the only audited portfolios available to individual investors. If you’d like to subscribe to the Advanced Market Signals plan, email us at premium@io-fund.compremium@io-fund.com.

Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.

Recommended Reading:

• Seagate: Slow QoQ Data Center Growth, 2027 Capacity Under Discussions
• Monolithic Power: Strong AI Tailwinds to Drive 50% Enterprise Data Segment Growth in FY26
• Teradyne Q4: Revenue Accelerates to 41% QoQ, Possibly Peak Growth Quarter
• Talen Q4: Reveals 2028 Timeline for Data Center Power Delivery

Seagate has been a rather under-the-radar AI beneficiary with shares up more than 262% over the past year, keeping pace with heavyweight Micron and other strong memory sector performers. The company is working with cloud service providers (CSPs) to supply hard disk drives, stating they are qualified with 6 out of 8 CSPs on their new HAMR-based Mozaic products. As it pertains to data centers and AI, it’s expected that HDDs will handle the high-volume data storage requirements for agentic AI, which relies on persistent access to large volumes of historic data for planning, reasoning and decision-making. 

With that said, Seagate’s data center revenue growth is quite slow at just 5% QoQ in Q4, as the company is not exposed to the same swift price surges rippling across DRAM and SSD markets due to locked-in pricing for 2026.  

Looking ahead to 2027, Seagate could potentially raise prices with management stating that pricing is not locked-in yet (but notably, management was a bit evasive on the details when asked many times in the Q&A session): “But I would say we have — the vast, vast majority of the volume is already allocated. Calendar '27, we will start working on that fairly soon. Of course, we have very good indication and agreement on volumes, but we have not — we have not fixed the price yet.” 

Regardless of exact timing on when pricing can be renegotiated, Seagate’s fundamentals show exceptional margin growth as HAMR products offer stronger profitability and improved cash flows, leading to net leverage of 1.1x. It spells good things to come as HAMR products are only beginning to ramp now yet are making an impact on margins. 

Brief Product Overview: HDDs and HAMR Technology 

Hard disk drives (HDDs) offer the lowest-cost per terabyte, which makes HDDs ideal for “big data” storage, backups and large AI datasets. Compare this to solid state drives (SSDs) which store data on flash memory chips and are far faster and lower-latency. SSDs cost more per terabyte, thus leveraging a mix of HDDs and SDDs is a popular choice.  

As it stands today, SDDs have illustrated significant pricing power compared to HDDs. Therefore, because HDDs are considered more commoditized in the current market dynamics compared to SDDs, I’m quoting the management team in full as to why HDDs are likely to increase in importance as the AI economy plays out – as this is the predominant question Seagate must answer (when comparing to other memory sector stocks): 

“So if the concept is that drives aren't working hard, they're in the background, just storing data, that's not the way — a good way to think about it. That's not the way hard drives are being used right now. They're working 24/7. A lot of times, they're optimized for performance themselves, largely streaming performance, not random small block workloads. That's more of a memory thing. And so if you had an application that's random small block, it's probably memory. If you have big data, it's probably a little bit of memory on the front end and a lot of hard drive on the back end. And we think that — there are applications across the entire spectrum, of course, but we think that in the future when we start to talk about the concepts in their enormity, about checkpoints and physical AI and video and things like that, it's large, large data. So the architectural tier that stores the data will probably remain constant for the next decade.” 

Heat-assisted Magnetic Recording (HAMR) 

Seagate has developed heat-assisted magnetic recording (HAMR) tech for substantial areal density gains, which refers to how many bits can be packed onto each square inch of disk platter (where data is stored). HAMR uses a laser diode to heat a small spot on the disk, enabling polarity of a single bit to be flipped to allow data to be written.  

Other HDD tech such as energy-assisted magnetic recording (EAMR), like Western Digital’s ePMR (energy-assisted Perpendicular MR) tech, adds electrical currents to the write head to improve density and write smaller bits. While EAMR is simpler and builds on familiar PMR tech, offering faster time to market, it does not scale to the same capacity as HAMR. WDC acknowledges that EAMR is expanding to 30-40TB drive capacity with roadmaps to 60TB, while HAMR enables capacity to expand to >100TB by 2032. 

Seagate says HAMR allows data bits to become even smaller and more densely packed, while remaining thermally stable, and its HAMR-based drives can offer more exabytes of storage while taking up less space in the rack, offering substantial increases in TCO. To put this in number form, Seagate says a 1 exabyte deployment with HAMR would consume ~2 million kWh per year in 204 sq ft of space, versus 3.2 million kWh per year and 400 sq ft of space for a similar PMR deployment. 

Management believes that HDDs and higher-capacity storage systems will play a key role in meeting the storage needs for agentic AI, which need constant access to substantial volumes of data for reasoning and decision-making. Seagate projects AI workloads will drive “7.2 zettabytes of nearline storage demand over the next four years, exceeding the entire industry's consumption over the last decade.” 

Seagate’s Mozaic3+ HAMR, offering >3TB per disk, is qualified with all major US CSPs and the company says it remains on track to qualify with all global CSPs within the first half of CY26: “We ended the year shipping 3 terabyte per disk Mozaic-based HAMR products to our first CSP customer. And by year's end, quarterly HAMR shipments exceeded 1.5 million units and have continued to ramp. Mozaic 3 HAMR drives are now qualified with all of the major U.S. CSP customers and qualifications for our second-generation Mozaic 4 terabyte per disk products are tracking well to plan.” 

In terms of future catalysts, Mozaic4+ HAMR is currently qualifying, with the ramp expected to begin later this quarter and multiple CSPs expected to be qualified over the coming months. Mozaic is powering Seagate’s Exos family of products, which scale to multi-petabyte block storage systems, offering up to 2.5 PB (2,500 TB) in storage.  

In terms of volume, Seagate says that quarterly HAMR shipments have exceeded 1.5 million units and are continuing to ramp, or roughly 20% of its estimated ~7.2 million nearline unit shipments in Q2. To note, Seagate is ahead of Western Digital when it comes to HAMR, as WDC is not expecting to ramp its HAMR products until the start of CY27. 

HDD Role in ‘Warm-Storage’ 

We recently covered discussions over the role of SSDs in ‘warm’ storage and Nvidia’s upcoming Inference Context Memory Storage platform in our SanDisk analysis. We had stated: 

“Nvidia believes that “AI factories need a complementary, purposebuilt context layer that treats KV cache as its own AInative data class rather than forcing it into either scarce HBM or generalpurpose enterprise storage.” For example, the current inference context hierarchy begins with HBM (G1), providing near-instant access to latency-critical context in active generation, down to SSDs (G3) in the third tier to handle ‘warm’ data, or data that is used regularly but less frequently and still requiring efficient, cost-effective storage. Enterprise or shared storage sits at the bottom of the hierarchy (G4), handling ‘cold’ data, or data stored for long-term retention but much less frequently accessed.” 

The connection here to Seagate is that Nvidia is proposing SSDs to take a more central role in the ‘warm’ tier where HDDs sits – this is what Seagate and WDC consider ‘nearline’, where data does not have to be accessed instantaneously but must remain readily available.  

Analysts questioned about this possible shift to SSDs in warm storage: 

“Dave, while we have you, a little bit of a technical one, are you — what kind of activity are you seeing at sort of the so-called warm tier of storage? It's a question that comes up a bunch in our conversations. We've heard that it's obviously growing, it's growing both hard drive and flash storage is participating nicely, but would love to get your input on it. Because I think there's still — first of all, we love to know if what we're hearing is accurate. 

But secondarily, I think there's a lot of people that are assuming that that's really like it's becoming a NAND tier, largely a NAND tier in the GenAI world. And anyway, just love to get any context there that you have. 

CEO and Chairman William MosleyCEO and Chairman William Mosley 

“When you start talking about big data storage, if you will, in data centers, the tiering architecture is fairly well set and probably won't change based on economics and also architectures that are well known. People know how to play. So if the concept is that drives aren't working hard, they're in the background, just storing data, that's not a good way to think about it. That's not the way hard drives are being used right now. They're working 24/7.  

A lot of times, they're optimized for performance themselves, largely streaming performance, not random small block workloads. That's more of a memory thing. And so if you had an application that's random small block, it's probably memory. If you have big data, it's probably a little bit of memory on the front end and a lot of hard drive on the back end. … So the architectural tier that stores the data will probably remain constant for the next decade.” 

While there has been some chatter about this newfound role for SSDs to command a leading position in meeting inference-driven storage demand, Seagate does not believe that there will ultimately be much change to storage architectures, with HDDs remaining critical to handle the massive volumes of data generated by AI applications such as agents, video generation, world models and more.  

2026 Capacity Sold Out, Beginning to Contract for 2027 

One of the reasons for this shift to SSDs is that HDD capacity is already sold out through 2026, with Seagate already turning to signing long-term agreements (LTA) for calendar 2027. Seagate said its nearline capacity is already fully allocated through CY2026, and management expects to “begin accepting orders for the first half of calendar year 2027 in the coming months,” supported by strong demand visibility from some of its major cloud customers seeking supply assurance through 2027 and into 2028. 

Considering the supply tightness and elevated level of demand in the market for HDDs, analysts prodded about how pricing would look under new LTAs, and if new agreements would be closed at higher prices. Management noted that 2027 prices are still in discussion and have not been locked in, and implied that higher prices are possible due to demand: 

Q, Christopher Muse, Cantor FitzgeraldQ, Christopher Muse, Cantor Fitzgerald 

“And then, I guess, maybe bigger picture, as you think about overall average pricing per exabyte, we've gone from kind of down double digits to high single digits. And I think we just exited the quarter down 4% year-on-year. Do you see a world where pricing could flat or even move positive year-over-year?” 

CEO William MosleyCEO William Mosley 
“Pricing will be dictated by the demand. Right now, the demand is really strong. So I think as we roll through into '27 and '28, we look at how much capacity we're having. We're bringing online by virtue of the fact that we're making all these aggressive product transitions. We'll bring more exabytes to bear and then people go out there and renegotiate for those. I think flat to slightly up is certainly possible. And that's the way we're really managing it as we talk to our customers.” 

This ties into a later response where Seagate revealed that pricing has not been locked in for 2027, while 2026 volumes and prices are well defined – meaning any potential upside in 2026 would only come from excess capacity being sold in the open market.  

CFO Gianluca Romano explained that Seagate has “very good indication and agreement on volumes, but we have not fixed the price yet,” so if prices begin to move higher as higher-capacity next-gen HAMR devices ramp, considering how tight supply remains relative to demand, Seagate has tailwinds to both revenue growth and profitability.  

It should also be noted that in clear contrast to SSDs, HDD prices are not surging – analysts had implied HDD prices exited Q4 down (4%) YoY, while other reports placed prices up ~4% QoQ, a far cry from SSDs rising 40% to 100% QoQ – this is why Seagate’s QoQ Data Center revenue growth pales in comparison to competitors such as SanDisk. 

If Seagate can only realize a small low to mid-single digit price increase with 2027 contracts, this suggests growth will likely follow its projections for mid-20% exabyte growth with a few points' upside, rather than sharply accelerating on a YoY or QoQ basis. 

Financials:  

Revenue Growth

Seagate reported fiscal Q2 revenue of $2.83 billion, up 21.5% YoY and 7.5% QoQ, with sequential revenue growth across nearly all end markets. Growth was relatively steady on both a YoY and QoQ basis compared to fiscal Q1. 

Looking ahead, Seagate guided for FQ3 revenue to be $2.90 billion, +/- $100 million, pointing to YoY growth of 34.3%, accelerating nearly 13 points, though QoQ growth would be just 2.7% at midpoint. This softer QoQ read is due to typical March quarter seasonality from edge IoT markets, though management expects Data Center to more than offset that impact.  

Seagate also did not provide guidance for the full-year, but did state that its current outlook expects “sequential improvement to both the top and bottom line throughout calendar 2026,” meaning QoQ growth in each quarter through fiscal Q2 2027.  

Assuming low double-digit QoQ growth in FQ4 from the midpoint of Q3’s guide, FY26 revenue would roughly project to ~$11.64 billion, up 27.9% YoY, slightly above current estimates for $11.50 billion for 26.4% growth. 

Looking out to the first half of FY27, considering management’s commentary and similar QoQ growth rates as the past two years at ~7-8% QoQ, Seagate could exit calendar 2026 (FQ2 27) at ~$3.75 billion, or about 4% above current estimates for $3.61 billion. 

Key Segments – Data Center Growth Decelerates to 5% QoQ 

Seagate reports under two segments – Data Center and Edge IoT, with Data Center being the company’s primary revenue stream, accounting for 87% of its shipment volume in the quarter but only 79% of revenue. 

Data Center revenue increased 28% YoY and 5% QoQ to $2.22 billion in FQ2, decelerating from 34% YoY and 13% QoQ in Q1. Driving this was a deceleration in shipments – Data Center shipments were 165 exabytes in the quarter, up 4% QoQ and 31% YoY, slowing from 17% QoQ and 39% YoY in Q1.  

Seagate said it is seeing “sustained demand growth for our high capacity nearline drives across global cloud data centers as well as continued improvement from the enterprise edge,” expecting strong demand trends to continue for some time. In the enterprise OEM specifically, Seagate said it is “benefiting from slight improvement in traditional server units, along with increasing demand for storage servers, driven in large part by the adoption of AI applications and need to store data at the enterprise edge.” 

Outside of Data Center, Edge IoT revenue rose 2% YoY and 17% QoQ to $601 million, though it should be this sequential growth followed a soft FQ1 which saw an (11%) QoQ decline. Management said this was driven by anticipated seasonal improvement for consumer products.  

Looking ahead to FQ3, management expects Data Center to more than offset typical March quarter seasonality in edge IoT – assuming a similar (7%) QoQ decline in Edge IoT, this would project Data Center revenue to be up just over 5% QoQ, matching Q2’s pace.  

Gross and Operating Margin Reach Records 

Seagate is witnessing solid margin expansion, driven by pricing and operating leverage, with management pointing to high-capacity drives as a gross margin tailwind while opex continues to decline. Both gross and operating margin reached company records in Q2.  

Q2 GAAP gross margin was 41.6%, up 6.3 points YoY and 2.2 points QoQ, while adjusted gross margin was 42.2%, up 6.7 points YoY and 2.1 point QoQ. Management said this was driven by its pricing strategy and improving mix of high-capacity drives as HAMR shipments ramp. Seagate noted that the upcoming Mozaic4+ adds more content per unit, helping reduce costs and improve profitability. 

Q2 GAAP operating margin was 29.8%, up 8.8 points YoY and 3.4 points QoQ, while adjusted operating margin was 31.9%, up 8.8 points YoY and 2.9 points QoQ. Seagate continues to see a greater degree of operating leverage, as adjusted opex as a percent of revenue declined to 10.3%, from 12.4% in the year ago quarter and 11.1% in Q1. 

Q2 GAAP net margin was 21%, up 6.5 points YoY and just 0.1 points QoQ. Adjusted net margin was 24.8%, up 6.2 points YoY and 2.6 points QoQ. 

Looking ahead to FQ3, Seagate projected a notable uptick in both gross and operating margins. Adjusted operating margin was projected to be in the mid-30% range, up around 3 points QoQ assuming this would correspond to roughly 35%. Based on adjusted opex guidance to be ~10% of revenue, this would place adjusted gross margin at ~45% under this framework.  

Evercore’s Amit Daryanani questioned about the QoQ strength in margins, more specifically the gross margin – management would not offer much aside from pricing and mix, with a HAMR ramp at a recently qualified customer aiding this expansion: 

“Gianluca, I'm hoping you can talk a little bit about the March quarter guide because there seems to be a really sizable uptick in gross margins. I think it's up like 250 basis points or 100% plus incrementals. Could you just — is there anything you would call out in March quarter that's unique that's helping drive that kind of margin expansion? And is this really all coming from the core HDD business? Or is there a potential benefit from the old systems business helping you as well?” 

EVP and CFO Gianluca Romano 

“Amit, well, I would say, we expect to be a very good quarter. I don't think it's different than what we have done before. It's always based on the pricing strategy and the mix, as you know. We qualified another customer on HAMR, so we will ramp a little bit more volume on HAMR. This is helping us to get better margin. But fundamentally, is not really different in how we think we are going to execute the quarter and is good. I think the incremental margin looks very good.” 

Romano had also clarified that Seagate had previously presented a model with a 50% incremental margin above $2.6 billion of revenue, noting that this model covers the next two to three years, with the HAMR ramp helping drive further margin expansion.   

EPS 

Driven by the margin expansion, Seagate delivered strong sequential EPS growth, with adjusted EPS up 19% QoQ to $3.11, beating estimates by 9.6%. YoY growth for adjusted EPS was 53.2%, decelerating from 65.1% in Q1. GAAP EPS in Q2 was $2.60, up 67.7% YoY, decelerating from 72.3% in Q1.  

Looking ahead to Q3, Seagate guided for adjusted EPS to be $3.40, +/- $0.20, representing a more than 25 point acceleration to 78.9% YoY at midpoint. Growth is expected to decelerate back towards ~50% YoY in FQ4 to $3.87.    

For FY26, current adjusted EPS estimates sit at $13.02, up 60.8% YoY, while GAAP EPS is projected to be $11.71, up 73% YoY. 

Cash Flows and Balance Sheet 

Seagate noted that its free cash flow generation reached the highest level in the last eight years, while they also retired $500 million in debt to strengthen its balance sheet.  

Operating cash flow in Q2 was $723 million for a 25.6% margin, up 17.1 points YoY and 5.4 points QoQ.  

Free cash flow was $607 million for a 21.5% margin, up 15 points YoY and 5.3 points QoQ. Management guided for free cash flow to expand further in Q3, supported by strong demand trends and operational efficiency. 

Cash and equivalents totaled $1.05 billion with management noting that total liquidity was $2.3 billion including its undrawn revolving credit facility.  

Inventories remained flat QoQ at $1.5 billion.  

Debt was $4.5 billion exiting Q2, down from $5 billion at the end of Q1. Seagate said its net leverage ratio was 1.1x, improving 16% QoQ and 63% YoY, with the expectation that net leverage will continue to trend lower as profitability and cash flows increase.  

Valuation 

Seagate is trading close to peak multiples, a tougher pill to swallow considering the slow sequential growth the company is seeing in the Data Center. Seagate trades at 7.8x forward PS, well above its 5-year average of 3.0x, and just below its recent peak of 8.5x. Shares had traded as low as 1.5x in April. 

Looking at the bottom line, Seagate trades at 31.5x forward PE, below its 39.9x average, though it should be noted this is skewed higher due to margins falling negative in 2023. Over the past year, shares are trading nearly double its average of 17x. 

Conclusion 

Seagate’s data center growth is much slower than other memory peers at just 5% QoQ as the HDD space is witnessing soft pricing power relative to SSDs or DRAM. Data center growth is also unlikely to meaningfully inflect moving through CY26 as price and volume agreements have already been locked in. Seagate does see potential for some pricing upside in 2027 as contracts are open for discussions, yet for now, pricing tailwinds are muted compared to what is seen in enterprise SSDs and DRAM.

Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.

Every Thursday at 4:30 pm Eastern, the I/O Fund team holds a webinar for premium members to discuss how to navigate the broad market, as well as various stock and crypto entries and exits. Beth Kindig offers weekly deep dives including lesser-known cryptocurrencies and AI stocks, plus the team offers trade alerts. The I/O Fund team is one of the only audited portfolios available to individual investors. If you’d like to subscribe to the Advanced Market Signals plan, email us at premium@io-fund.compremium@io-fund.com.

Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.

Recommended Reading:

• Monolithic Power: Strong AI Tailwinds to Drive 50% Enterprise Data Segment Growth in FY26
• Teradyne Q4: Revenue Accelerates to 41% QoQ, Possibly Peak Growth Quarter
• Talen Q4: Reveals 2028 Timeline for Data Center Power Delivery
• Nova: Memory Revenue Accelerates Sharply, though 1H 2026 Expected to be Soft

SanDisk was the best stock in the S&P 500 in 2025 with a 559% return, and the company is continuing this strong performance in 2026 with shares up 41% in barely two weeks. Much of this performance is rumored to be linked to Nvidia’s CES presentation where the GPU leader discussed context windows as the next bottleneck for AI inference, and hinted at rack-scale and network solid state drives (SSDs) becoming key components to address this. 

On a broader level, data center/enterprise SSDs are often overlooked but equally critical as HBM when it comes to AI training and inference. This is because data center SSDs offer higher read-write speeds critical for accessing and transferring data rapidly, along with higher performance and energy efficiency, vital factors for larger-scale AI training and inference workloads.    

SanDisk operates independently after being spun out of Western Digital in February 2025. The company expects to ride enterprise SSD demand tailwinds with management projecting sequential growth in its data center segment through 2026, with two hyperscaler qualifications underway and an additional hyperscaler expected in 2026.  

However, data center remains a smaller portion of revenue, contributing $269 million last quarter or less than 12% of revenue, with client (PC/smartphones) and consumer products (SD cards/USB) remaining core to its business.  

The Context Bottleneck, and Extending KV Cache to SSDs 

Moving to some longer-term tailwinds for SSDs, Nvidia discussed how context windows could soon be the new bottleneck for AI inference performance at CES this week, as it unveiled its new Inference Context Memory Storage platform (ICMS) to address growing key-value (KV) cache capacity limits. KV cache capacity is a known pain point when working to balance long-context reasoning and memory capacity.  

Put simply, the KV cache is a memory optimization technique that stores calculations during the inference phase, allowing the model to remember those prior calculations instead of repeating them, thus enabling faster response times; without it, latency would be much higher and computation much slower as every new token would require recalculation of all prior tokens. KV cache essentially serves as a model’s long-term memory that is reused and extended throughout many steps or requests.  

However, the KV cache has a substantial memory footprint, especially for long contexts, and during deployment it can consume ~30% of GPU memory, making it a major bottleneck for large-context applications, such as coding, natural language processing, or handling simultaneous requests from many users on large models.  

Let’s first take a look at KV cache from a tensor parallelism perspective, and how distributing memory across tens to thousands of GPUs significantly increases available KV cache memory, translating directly to larger maximum token context windows. 

For example, a single AMD MI300X GPU running a Llama-70B model would have ~17GB of memory capacity available for the KV cache, after accounting for 140GB to store model parameters (2 bytes per parameter on FP16, so 70 billion * 2), and 35GB for the activation buffer (estimated ~25% of model storage), per TensorWave. With an estimated 2.6MB required per token (or 2.6GB per 1K tokens), a single GPU can handle a max request of ~6,500 tokens.  

When you distribute model parameter and activation buffer across an 8-GPU server, or ~17.5GB and 4.4GB per GPU, this frees up 170.6GB per GPU for the KV cache, a ~10X increase; for the server, KV cache memory availability is now 1,360GB. This means that an 8-GPU server could now handle a max request of 523,000 tokens, an ~80X increase, with gains that only compound as server size and memory increase.  This can then be optimized for longer contexts, or 8 requests of 64k context lengths, for high-throughput, or 64 requests of 8k context lengths, or other combinations.  

Here’s where SSDs fit in – by extending or offloading KV cache to local SSDs, model prefill and time to first token can be significantly reduced, thus significantly decreasing latency and increasing throughput.  

AI inference acceleration startup WEKA states that when it tested Llama-3.1 70B with no optimizations, a 100K token prompt took 24 seconds to prefill into the model before any output could be generated, but “extending GPU memory to ultra-fast storage [NVMe SSDs] can dramatically improve token processing efficiency.” When configuring an Nvidia DGX H100 server with an 8-node exabyte-scale NVMe SSD pod, WEKA says its “tests demonstrated a staggering 41x reduction in prefill time on LLaMA3.1-70B, dropping from 23.97 seconds to just 0.58 seconds,” significantly improving model efficiency with zero optimizations – simply from adding SSDs to extend GPU memory.  

Google ran tests using an 8-GPU H100 server on Llama-3.3 70B, extending system memory to larger, lower-cost CPU RAM and enterprise SSD tiers. For a ~4 million token cache, Google found that utilizing CPU RAM and SSDs in conjunction with HBM decreased end-to-end latency and time to first token by 64% to 79%, while increasing throughput by 179-264% on 50k to 100k token prompts.  

Nvidia Working to Tackle the Context Bottleneck with its ICMS Platform  

With its new ICMS platform, Nvidia is working to mitigate context windows becoming a major bottleneck as agentic AI and physical AI scale over the coming years. Scaling of models to trillions of parameters and the shift to multi-step reasoning, multi-agent workflows or advanced multimodal applications will generate substantial volumes of context data and require significant KV cache reuse to maintain accuracy and context in prolonged interactions.  

As such, Nvidia believes that “AI factories need a complementary, purpose‑built context layer that treats KV cache as its own AI‑native data class rather than forcing it into either scarce HBM or general‑purpose enterprise storage.” For example, the current inference context hierarchy begins with HBM (G1), providing near-instant access to latency-critical context in active generation, down to SSDs (G3) in the third tier to handle ‘warm’ data, or data that is used regularly but less frequently and still requiring efficient, cost-effective storage. Enterprise or shared storage sits at the bottom of the hierarchy (G4), handling ‘cold’ data, or data stored for long-term retention but much less frequently accessed.  

Source: Nvidia 

Nvidia is essentially proposing an architectural redesign of this hierarchy, positioning the new ICMS platform between G3 and G4, or as it calls it, G3.5. ICMS is a new Ethernet-attached NVMe SSD storage tier, likely integrated into the fabric and optimized specifically for KV cache usage at the pod level. It is powered by Nvidia’s new BlueField 4 data processing unit (DPU) packing 512GB of on-board SSD capacity, a 4x increase from BlueField3’s 128GB, and is combined with Nvidia’s GPUDirect Storage, which bypasses the CPU and provides direct memory access from GPU to SSDs to reduce latency.  

ICMS will provide petabytes (millions of GB) of shared KV memory capacity per GPU pod, capable of storing context for many models or agents simultaneously, while being located close enough to GPUs to frequently share inference context with lower power consumption and better efficiency versus shared storage. Nvidia claims ICMS can enable up to 5x improvements in power efficiency and 5x increases in tokens per second versus shared storage. 

It is this new platform and Nvidia CEO Jensen Huang’s comments relating to the storage hierarchy redesign that have fueled optimism for SanDisk and SSDs, as Huang believes it is a completely new market, integrating SSDs to the fabric, that could ultimately become the largest storage market: 

“For storage, that is a completely unserved market today. The way that storage works is SQL. SQL is structured data. Structured database is lightweight. AI database KV caches insanely heavy weight. You're not going to hang that off of your north-south network. I mean that's just a horrible waste of network traffic. You want to put it right into the computing fabric, which is the reason why we introduced this new tier. 

This is a market that never existed. And this market will likely be the largest storage market in the world, basically holding the working memory of the world's AIs. And that storage is going to be gigantic, and it needs to be super high performance.” 

Because Nvidia is positioning NVMe SSDs to become the backbone for this new shared memory tier, there is the potential for SSD suppliers to see solid medium/long-term tailwinds from increased SSD capacity requirements in inference-optimized deployments over the next few years. For example, Bernstein estimates that Huang’s CES comments on SSDs and KV cache requirements suggest an additional 16TB per GPU, compared to 3-4TB per GPU today, or 4-5X growth. This will be more weighted towards year-end and into 2027 as ICMS rolls out with Rubin.  

SanDisk’s BiCS8 Tech, Kioxia JV and Data Center ‘Stargate’ SSD Line 

SanDisk is eyeing strong growth in the enterprise SSD market with its ‘Stargate’ NVMe SSD products, based on its BiCS8  architecture jointly-developed with Kioxia, offering industry-leading capacity, energy efficiency and performance.  

NVMe (Non-Volatile Memory Express) is a protocol designed specifically for NAND-flash based SSDs that optimizes performance by reducing latency and increasing data transfer speeds by utilizing the PCIe bus, enabling high throughput and fast data transfer speeds necessary for AI training and inference. 

SanDisk and Kioxia’s joint venture is one of the longest-standing JVs in the industry, signed in 2000 and lasting through 2034. It is primarily a shared manufacturing and capex strategy, with the two both splitting JV capex and wafer output and then selling NAND products independently. For example, the mega-fab in Yokkaichi, Japan produces nearly one-third of all global NAND bits (500K wafers per month), with the JV taking 80% capacity, split 50/50 between SanDisk and Kioxia, and Kioxia taking the remaining 20% (for an overall split of 60-40 for Kioxia and SanDisk).  

The two also recently started operations at their new second fab in Kitakami, Japan, which is geared towards BiCS8 3D NAND and future advanced 3D NAND production, with output expected to ramp meaningfully in the first half of 2026. This will help scale SanDisk’s enterprise SSD line, based on BiCS8, and potentially aid in future development of high-bandwidth flash. BiCS8 accounted for 15% of bits shipped in fiscal Q1 and is expected to reach majority of bit production exiting FY26, providing a clue into the ramp profile for the year. 

BiCS8 is the duo’s eighth-generation BiCS (bit cost scalable) 3D NAND architecture, which stacks NAND cells vertically, creating more layers and reducing costs per bit. BiCS8 scales to 218 layers from 162 layers in BiCS6, with SanDisk saying that BiCS8 increases memory density by more than 50%, program and read bandwidth by 35% and 26%, and data transfer speeds by more than 80% versus BiCS6. The two also have previewed the next generation of BiCS, scaling to 332 layers and further improving interface speeds by ~33%.  

Additionally, SanDisk believes that its BiCS8 QLC (quad-level cell) die underpinning its Stargate data center SSDs delivers substantial performance, latency and efficiency advantages over competitors: 11% to 67% faster input/output speeds in Gb/s, along with 27% to 34% lower latency. Management expects its BiCS8 QLC line to go from ~20% to 40% of its data center business by the end of FY26. 

Source: SanDisk 

SanDisk’s ‘Stargate’ line, built on BiCS8, debuted this year with 64TB and 128TB capacities now shipping. 256TB products are scheduled for launch in mid to late 2026 and 512TB targeted in 2027, with the combination of fast performance, low latency and high capacity making the SSDs suitable for managing massive AI datasets and workloads. SanDisk says Stargate “is growing in demand with 2 hyperscaler qualifications underway and a third hyperscaler along with a major storage OEM planned for calendar year '26,” with current qualification focused on 128TB. 

SanDisk also says that its SN861 NVMe SSDs were the first to be certified to support Nvidia’s GB200 NVL72, though this does not mean that its shipments will be correlated 1:1 with Nvidia’s racks. Management explained that the certification puts them on the approved vendor list, and “when the ODMs are picking their design, they pick vendors from the approved vendor list. So that's how we are getting into qualifications with the system partners that are building on top of Nvidia, not necessarily the Nvidia build.”

‘AI SSDs’ and a Path to ~33X Increase in Performance  

Nvidia’s ICMS also could create a new market for SLC-based ‘AI SSDs’ in the data center, which had predominantly focused on TLC and QLC-based SSDs. SLC, or single-level cell, stores one bit of data per cell, offering the fastest data retrieval and best performance, though it is typically the most expensive; TLC and QLC (triple-level and quad-level) store three and four bits per cell, increasing storage capacity significantly and reducing cost but with slower performance versus SLC. Overall, SanDisk’s role in these first performance-based ‘AI SSDs’ remains somewhat unclear, though the company is playing a much more integral role in high-bandwidth flash (HBF) development. 

The AI SSD push is currently being spearheaded by SK Hynix and Kioxia, and ultimately aims to boost SSD performance by up to 33X by 2027. SK Hynix is creating a three-product family, ‘AI-N’, with its AI-N P focusing on maximizing performance, AI-N B on maximizing bandwidth, and AI-N D on maximizing storage density. Reports suggest that Hynix’s first-gen AI-N P product will target approximately 25 million input/output operations per second (IOPS), or how many read/write operations a device can perform per second, with a higher number meaning faster performance. This is about an 8X leap from current SSDs at ~3 million IOPS today, while Hynix’s second-gen and Kioxia’s product in conjunction with Nvidia are said to be targeting 100 million IOPS by 2027, a 33X increase. 

High-Bandwidth Flash Could Boost GPU Memory by 21X 

HBF is being proposed as a future alternative/replacement for HBM memory on GPUs, stacking up to 16 3D NAND BiCS8 dies using through-silicon vias (TSV) to deliver up to a 21X capacity boost with similar bandwidth and cost as HBM.  

For example, SanDisk’s first-gen HBF could pack capacity of 512GB per stack with HBM-like bandwidth. At partial HBM replacement, such as in six out of the eight dies on the GPU package, HBF could boost total GPU memory to ~3,120 GB per GPU, a nearly 17X increase versus individual Blackwell GPUs featuring 186GB HBM per GPU. In full replacement of HBM, HBF could provide 4,096 GB of total memory per GPU.  

The significance of this is that it could allow frontier models to be stored entirely within a single GPU, rather than needed to be partitioned across a rack. SanDisk explains, “Think about a frontier large language model, let's say, something like GPT-4. GPT-4 has 1.8 trillion parameters with 16-bit rates. And that model alone would take 3.6 terabytes or 3,600 gigabytes of memory space. I can put that entire model on a single GPU now. I don't need to shuffle around data any longer.”  

In terms of potential commercialization of HBF, Hynix’s aforementioned AI-N B line is built on HBF and expected to be developed in collaboration with SanDisk. The first ‘alpha’ version sample could be released as early as January with the first proof-of-concept samples in 2027, followed by full-scale evaluation afterwards.  

SanDisk stated that the “gating item indeed is going to be enabling the ecosystem, aligning with the customers at their system level, integrate it and then bring it to the market,” but the product is highly executable as it is based on its existing NAND architecture. CEO David Goeckeler clarified in Q3 that the company “announced a time line last quarter of having the memory later in '26 and then having the controller for that in '27 we're still working towards that time line.” 

However, timelines for HBF are still unclear given the newness of the technology, with some estimates suggesting 2027 to 2028 as a possibility. It is far too early to tell whether HBF will be commercially viable or successful.  

Training and Inference are Long-Term SSD Demand Drivers 

This section includes a brief excerpt from our 6,000+ word thematic deep dive into the current AI memory boom recently published for our Pro subscribers: 

AI training and inference are two main long-term drivers for SSD demand, which is projected to rise ~6X from 2024 to 2030, from 181 exabytes (EB, or equal to 181,000,000 TB) to 1,078 EB, under McKinsey’s base case scenario. Training demand projected to rise at a 62% CAGR to from 7 EB in 2024 to 127 EB by 2030. On the flipside, demand from AI inference is expected to grow at a 105% CAGR from 6 EB to 447 EB by 2030, giving inference a 41% share of demand versus less than 12% for training. base case scenario. Training demand projected to rise at a 62% CAGR to from 7 EB in 2024 to 127 EB by 2030. On the flipside, demand from AI inference is expected to grow at a 105% CAGR from 6 EB to 447 EB by 2030, giving inference a 41% share of demand versus less than 12% for training.  

This is not only driven by development of more LLMs, but also the increasing size and complexity of frontier models, where training data sets and context windows for inference are getting increasingly large.   

For example, EpochAI estimates that training data set sizes are rising 3.7X per year on average, or nearly doubling every six months, though there are some models that are scaling much quicker. For example, Meta’s Llama2-70B from 2023 was trained on 2 trillion tokens, while Llama3-70B, from 2024, was trained on 15 trillion tokens, a 7.5X increase. Multi-modal models, those integrating audio, video, image or more, are also likely to require significantly more SSD storage, with McKinsey estimating in the hundreds of TBs depending on the mix of data needing to be stored.  estimates that training data set sizes are rising 3.7X per year on average, or nearly doubling every six months, though there are some models that are scaling much quicker. For example, Meta’s Llama2-70B from 2023 was trained on 2 trillion tokens, while Llama3-70B, from 2024, was trained on 15 trillion tokens, a 7.5X increase. Multi-modal models, those integrating audio, video, image or more, are also likely to require significantly more SSD storage, with McKinsey estimating in the hundreds of TBs depending on the mix of data needing to be stored.   

Source: EpochAI  

The increasing size and complexity of models also ties directly to a major pain point when it comes to inference: “As models grow in complexity and require longer contexts, their memory footprint expands beyond what a single GPU can handle. This results in inefficiencies where GPUs are memory-starved, causing significant bottlenecks in AI token generation.” inference: “As models grow in complexity and require longer contexts, their memory footprint expands beyond what a single GPU can handle. This results in inefficiencies where GPUs are memory-starved, causing significant bottlenecks in AI token generation.”  

This is exactly what Nvidia is addressing with Rubin and ICMSP, creating a new storage tier within the cluster fabric that is designed to extend GPU memory and facilitate high-speed KV cache distribution among racks.  

There are also tailwinds to SSD growth from increasing cluster sizes, with compute-focused eSSDs seeing a 1:1 attach rate per GPU. For example, SanDisk says that a real-world 32,256 GPU cluster (or eight pods of 252 16-GPU racks) would require 4,032 compute eSSDs such as its SN861 product. This could create a strong tailwind for SSD growth as clusters scale to 100K+ GPUs towards 1 million, assuming the correlation for compute eSSDs to GPU remains 1:1. 

Financials 

Revenue 

SanDisk reported a strong sequential revenue acceleration in its fiscal Q1, driven by NAND demand outpacing supply and increasing demand in its data center, edge and consumer end markets. Q1 revenue increased 22.6% YoY and 21.4% QoQ to $2.31 billion, accelerating from 8% YoY and 12.2% QoQ growth in fiscal Q4. Higher-than-expected bit growth drove the outperformance in the quarter relative to guidance of $2.1-2.2 billion, per management.    

SanDisk’s Edge segment was the primary growth driver in Q1 with revenue up 30% YoY and 26% QoQ to $1.39 billion, driven by increasing NAND content in PCs and smartphones and a positive PC refresh cycle. Consumer revenue rose 27% YoY and 11% QoQ to $652 million, while data center revenue was down (10%) YoY but up 26% QoQ to $269 million.  

Q2 revenue was guided to be $2.55 to $2.65 billion, up 38.6% YoY and 12.6% QoQ at midpoint. CFO Luis Visoso clarified that “the key message is most of the growth in revenue will be pricing driven in the quarter.”  

Revenue growth is then expected to accelerate further to 55% YoY in fiscal Q3 (even with a seasonal slowdown in consumer products following the holidays) and then decelerate slightly to 51% in Q4. Pricing tailwinds could strengthen significantly in FQ3 on reports that NAND prices for enterprise SSDs could rise ~100% QoQ in the March quarter, according to supply chain checks by Nomura. Citi estimates SSD prices will rise ~32% QoQ in the March quarter following a 21% QoQ increase in the December quarter.  

For fiscal 2026, SanDisk is currently expected to generate revenue of $10.6 billion, up 44.1% YoY. SanDisk sees demand outpacing supply through the entire year, currently estimating supply to support mid-teens demand growth, and potentially lead to strong pricing tailwinds from this tight/tightening environment:  

“We saw supply growth in calendar year '25 of about 8%. We see it at about 17% in '26. We see demand — constrained demand around 14% [mid-teens] because that's all that's out there from a supply point of view. But unconstrained demand is in the — literally, a couple of weeks ago, we thought it was 20%, it's probably mid-20s by now. So we see the supply pretty much being able to service that kind of mid-teens level demand for '26.” 

This tightening environment comes despite fabs running at 100% utilization, with management adding that they do not plan on adding capacity to any end market, but rather remain prepared with the optionality to shift capacity as visibility into product mix strengthens.  

AI Segment Growth 

SanDisk’s data center revenue, as mentioned above, declined (10%) YoY but rose 26% QoQ to $269 million, driven by increasing demand for its ‘Stargate’ enterprise SSD product line. However, revenue contribution remains small, at less than 12% of revenue.  

SanDisk did not provide a numerical guide for Q2 for data center, but management noted that they are expecting sequential growth throughout fiscal 2026 with faster growth in the back half, driven by the current hyperscaler qualifications planned and underway. SanDisk did clarify that they are “working with 5 major hyperscale customers through active sales and strategic engagements” across its data center portfolio.  

Data center growth is supported by solid visibility, with management explaining that they are either “striking deals that are multi-quarters, let's say, through the first half of next calendar year” from customers looking to lock in supply, or working with customers with demand visibility through 2027 to align supply with those demand forecasts. Management also sees undersupply conditions extending potentially into 2027 now, supporting strong pricing in deal negotiations.  

Management also increased their forecast for data center exabyte growth, explaining that last quarter, exabyte growth expectations were in the mid-20% range, but now are in the mid-40% range. As a result, data center is expected to be the largest market in NAND on an exabyte basis in 2026, surpassing mobile.  

Earnings 

SanDisk stands out for its strong expected earnings growth through fiscal 2026 and fiscal 2027, with adjusted EPS expected to reach more than $21 by then, or >7X higher than the $2.99 it earned in fiscal 2025.  

Q1 GAAP EPS was $0.75, a strong improvement from a ($0.16) loss in Q4, though this was down (49%) YoY from $1.46 in the year ago quarter as margins remained lower YoY. Adjusted EPS was $1.22, up 321% QoQ but down (33%) YoY.  

For Q2, SanDisk guided for adjusted EPS of $3.00 to $3.40, up more than 162% QoQ. Adjusted EPS is expected to further increase to $3.78 in fiscal Q3 and $4.82 in fiscal Q4.   

For fiscal 2026, SanDisk is expected to generate $13.29 in adjusted EPS, up 344.6% YoY, while GAAP EPS is projected to be $11.53, up from ($11.32) in FY25 due to the spin off. Fiscal 2027 is expected to see earnings power surpass $21, with GAAP EPS estimated to be up 86% to $21.47 and adjusted EPS up nearly 62% to $21.50. 

Margins  

Margins are lower YoY compared to pre-spinoff margins, but Q1 saw strong sequential margin expansion that is expected to accelerate in Q2.   

• Q1 GAAP gross margin was 29.8%, down 8.8 points YoY but up 3.6 points QoQ. Adjusted gross margin was 29.9%, down 9 points YoY but up 3.5 points QoQ.  
• GAAP operating margin was 8.3%, down 8.3 points YoY but up 5.6 points QoQ. Adjusted operating margin was 10.6%, down 8.2 points YoY but up 5.3 points QoQ.  
• GAAP net margin was 4.9%, down 6.3 points YoY but up 2.7 points QoQ, and adjusted net margin was 7.8%.  

For Q2, SanDisk guided adjusted gross margin to be 41-43%, or up just over 12 points QoQ at midpoint on higher pricing and cost reduction tailwinds, while adjusted operating margin is implied to be 24.2% at the midpoint of opex guidance, or up 13.6 points QoQ. Fab startup costs are expected to transition from headwinds to tailwinds during the quarter, potentially aiding more margin expansion into fiscal Q3 and Q4. 

Cash 

SanDisk noted that in Q1 it reached a net cash position, six months ahead of schedule, though debt is still almost equivalent to its cash on hand. Cash flows were quite strong, and adjusted FCF margin showed strong expansion.  

• Operating cash flow was $488 million in Q1 for a 21.1% margin, up from a (7%) margin in the year ago quarter and a 4.9% margin in Q4.  
• Adjusted free cash flow was $438 million in Q1 for a 19% margin, up from a (10.5%) margin in the year ago quarter and 2.6% in Q4.  
• SanDisk’s total gross capex to support the JV was $387 million in Q1, though its cash capex spend was only $40 million (1.7% of revenue) as the remainder was funded through external sources such as subsidies or tool depreciation recorded in COGS. 

Cash and equivalents totaled $1.44 billion while debt totaled $1.35 billion.  

Valuation 

SanDisk’s valuation is somewhat hard to pin down given the company’s limited history on the public markets after its February spinoff, and its rapid 362% ascent since the end of August.  

SanDisk trades at 5.3x forward PS, surpassing its prior peak at 4x in November and a substantial re-rating higher from 0.6x in the summer. For comparison, this is now on par with former parent Western Digital at 5.1x forward PS, though the two are focused on different memory market segments with WDC primarily in hard disk drives.   

For forward PE, SanDisk currently trades at an 28.7x multiple, nearly double its 15.8x average from the second half of fiscal 2025 prior to its fiscal year readjustment in June. Shares traded as low as 3x in July and August due to the sharp earnings increase expected in fiscal 2026.  

Notable Risks 

The NAND flash market has historically been quite volatile, and is shifting from significant oversupply in 2023 to expectations for substantial supply shortages through 2026 and potentially into 2027. However, if NAND capacity begins to come online quickly through next year, or if demand for PCs and smartphones falters due to rising memory prices, the NAND cycle could reverse and lead to pricing pressures cutting into revenue growth and margins.  

Competition is also quite stiff in enterprise SSDs, and SanDisk is a small player with <4% market share, versus Samsung at >35%, SK Hynix/Solidigm at nearly 27%, and Micron and Kioxia in the 14% range. Jefferies analysts also warned that there is “no idea” what market share China’s YMTC could take as it ramps up output. 

There’s also the risk of having a limited viewpoint on where normalized earnings could land if/when the cycle peaks and reverses, as SanDisk is currently benefiting from strong pricing and a tightening supply-demand environment. Combined with the sharp 1,000%+ rally since its summer lows and peak valuation multiples, there could be a higher degree of risk if the supply-demand imbalance and pricing revert sooner than expected.  

Conclusion 

SanDisk has a multi-faceted growth opportunity ahead over the next few quarters, with supply-demand imbalances widening with strong enterprise SSD demand, a potential doubling of prices in the March quarter supporting more upside for revenue, and multiple hyperscaler qualifications on deck. 

Nvidia’s CES keynote discussion around the context window becoming the next bottleneck could have positive implications for the SSD market from Nvidia’s ICMS platform utilizing NVMe SSDs to significantly boost KV cache memory and increase throughput for inference applications. HBF is also a potential long-term opportunity later in the decade as it could dramatically boost total GPU memory to allow frontier models to run on a single GPU, though it is too early to tell if it will be viable.

Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.

Please note: The I/O Fund conducts research and draws conclusions for the company’s portfolio. We then share that information with our readers and offer real-time trade notifications. This is not a guarantee of a stock’s performance and it is not financial advice. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis. Beth Kindig and the I/O Fund do not own shares in SNDK at the time of writing and may own stocks pictured in the charts.

Recommended Reading:

• The I/O Fund’s Top 10 New Ideas List for Q1 2026
• Celestica Eyes FY26 Acceleration on Strong Networking Switch Demand
• Nebius: Financing its Data Center Ambitions Will be Challenging
• Lumentum: EMLs Driving Results, CW Lasers Ramping with Q2 Guided for 22% QoQ Growth

The trend toward neoclouds is a high risk/high reward opportunity for investors. Nebius is similar to CoreWeave, dubbing itself as AI native cloud infrastructure, which means the infrastructure was created specifically for AI workloads with architectures built on bare metal servers instead of hypervisor layers, for example, creating an important differentiation from the Big 3 which we’ve previously covered here and here. 

Nebius offers an Nvidia-optimized cloud platform for teams that need to adjust compute resources, want high-performance storage and an easy-to-use AI environment yet do not want to manage the infrastructure or AI operations. 

Nebius shares surged in September following the up to $19.4 billion deal with Microsoft. Recently, the company expanded its hyperscaler customer list recently with a $3 billion deal with Meta expected to ramp over the next three months. Backed by these two deals ramping through 2026, Nebius now projects reaching $7 billion to $9 billion in annualized run rate revenue (ARR) by the end of next year, up more than 13X from its current ARR of $551 million from Q3.  

However, Nebius is a high-risk stock given its success depends on how much capital the company can raise, and the current financials do not support an easy path to ramping capacity and reaching these targets. Capex needs have already moved much higher, and meeting management’s more aggressive capacity targets for the end of 2026 will require significant capital to deploy the necessary GPUs – there is a chance for capex needs to remain at 7-8X of revenue in 2026, making financing a major challenge. 

Regardless of how Nebius executes relative to CoreWeave, it remains an AI bubble stock as the company has to hope the stock price goes up to raise cash, which will dilute shareholders (or raise debt). It’s a vicious cycle as during any months/quarters that AI stocks are soft, Nebius stock will carry outsized execution risk. 

For a closer look at Nebius’ deal with Microsoft, its vertical integration and custom servers, read more from our Advanced analysis, Nebius: Mega Microsoft Deal, 5x Growth in Power Fueling AI Cloud Hypergrowth, But High Risk Remains.Nebius: Mega Microsoft Deal, 5x Growth in Power Fueling AI Cloud Hypergrowth, But High Risk Remains. 

Revenue Growth Decelerates in Q3, but Expected to Reaccelerate 

Revenue decelerated to 355% YoY and 39% QoQ in Q3 to $146.1 million in Q3, down from 605% YoY and 90% QoQ in Q2. Core AI Infrastructure was the primary driver as revenue grew 400% YoY and 40% QoQ to ~$131.5 million, or ~90% of total revenue in Q3. 

On that note, Nebius tightened its 2025 revenue outlook, from $450-630 million down to $500-550 million, citing timing of capacity as the primary reason. This is also below current estimates for 372% growth to $555 million. 

On the other hand, revenue growth is expected to reaccelerate to 549% to $246.1 million in Q4 and further to 749% to $469 million by Q1, with more capacity coming online next quarter to support the Microsoft deal ramping through 2026 and its new $3 billion, five-year deal with Meta, with capacity rolling out over the next three months.  

For fiscal 2026, revenue is expected to reaccelerate to 521% to $3.45 billion, with estimates having doubled since Nebius signed the Microsoft deal in September. Nebius said this quarter that it plans to provide full-year guidance for 2026 in the upcoming quarter.  

However, reaching these targets will likely require strong/perfect execution as the company must build and deliver substantial capacity for both the Microsoft and Meta deal, as well as additional capacity to meet external demand. Doing so will likely require capex at >7X of revenue next year, an incredibly challenging position to be in as peers are spending far less yet still struggling to find funding. 

Ambitious $7-9B ARR Target by End of 2026, up >7X YoY 

Nebius provided an ambitious new medium-term annualized run rate revenue (ARR) target in Q3, forecasting reaching $7 billion to $9 billion in ARR by the end of 2026. For comparison, this would represent >7X YoY growth from 2025’s target of $1 billion at midpoint if it materializes.  

Nebius’ two hyperscaler deals with Microsoft and Meta would account for more than half of this at ~$4.1 billion in ARR at full scale combined (approx. $3.5B for Microsoft and $0.6B for Meta). However, this would require Nebius to generate ~$3.9 billion in ARR at midpoint from other customers and other capacity, a challenging task given how costly its capacity expansion plans will be.  

Additionally, the QoQ decline in incremental ARR, from $181 million in Q2 to ~$121 million in Q3, highlights a major challenge – this growth boils down to timing, and when Nebius can bring capacity online. This could be power bottlenecks, GPU supply bottlenecks, delays in building out physical infrastructure, funding bottlenecks, etc. 

Thus, if the company cannot raise enough capital to afford its ambitious plan of reaching 2.5GW of contracted power and 0.8-1.0GW of connected power by the end of next year, these revenue targets may be unattainable.  

Capex Forecast Raised by 2.5X to $5 Billion, 9X of Revenue 

Capex requirements will be the number one focal point for Nebius considering it does not have the same depth of cash as hyperscalers, yet is competing with them on securing power, GPUs, and AI workloads.  

On this note, Nebius is projected to spend ~9X its estimated 2025 revenue on capex after boosting its 2025 capex forecast by 2.5X to $5 billion, versus its prior view for $2 billion. This capex-revenue ratio is far above what peers such as CoreWeave and even Oracle are spending, at 2.5X and 0.75X. It’s also important to note that the latter two are struggling to in the financing department – CoreWeave is having to take on debt at >9% interest rates, while Oracle’s 520% debt-to-equity ratio has pushed its credit default swaps up to the highest level since 2009. 

Financing this capex will be the primary challenge, and it is critical considering the bulk of this capex will go to revenue-generating GPUs. To put this in perspective for 2025 and 2026, here’s what CEO Arkady Volozh explained about capex and Nebius’ connected power targets: 

And if we look at it from the CapEx point of view, roughly speaking, it breaks into 3 spending blocks. So first stage, securing land and power. It's pretty cheap. It's around — again, it depends on the scale, but it's around 1% of total CapEx for securing those blocks and electricity. The second stage, building the data centers, building connected power is something around, I don't know, 18%, 20%. And the remaining 80%, the main part is for deploying the actual GPUs. This is the main part of CapEx. So if we want to build as much as our capital will allow us, what should we do?1% of total CapEx for securing those blocks and electricity. The second stage, building the data centers, building connected power is something around, I don't know, 18%, 20%. And the remaining 80%, the main part is for deploying the actual GPUs. This is the main part of CapEx. So if we want to build as much as our capital will allow us, what should we do? 

First, we should secure as much capacity as we can because the cost [is] immaterial at this scale. Second, we should build as much as our capital allows us. And third, we will fill GPUs in line with contracted or clearly visible demand. We will need this massive 80% spend [that] will come only when we see real demand. That's why we say that in 2026, we will be securing 2.5 gigawatts total contracted capacity. And we are planning to physically build 800 to 1 gigawatt of connected data centers. This will be done by the end of next year.” 

For the updated 2025 capex guide, this would assume ~$50 million towards land, and ~$1 billion towards the physical data center shell and necessary equipment to connect power, and the remaining $4 billion for GPUs. This aligns with management’s forecast to have 220MW of connected power (not yet active) and 100MW of active power by year-end. This would be around ~$10 million per MW of active power based on the comments above, slightly below averages around $12 to 14 million.  

Looking ahead to 2026, Nebius is planning to have 2.5GW of contracted power and 0.8-1.0GW of connected power (up 2.5X from a prior view for 1GW contracted). Per management, this includes scaling its existing data centers in the UK, US, and Israel, new data centers under development in the US and Europe, and several large sites with up to hundreds of MW under review, with the chance that some go online by the end of 2026. 

Building out this capacity pipeline to reach the connected power targets with powered shells would likely require approximately $3 to $4 billion in capex, while GPUs would likely require $20 billion to $24 billion, potentially higher, depending on mix and how much of that connected power Nebius aims to have active. This is anchored by GPU costs for next-gen hardware now running at $25 billion to $30 billion per GW.  

Financing Potential >4X Capex Growth, Still 7-8X of Revenue 

There is a likelihood that Nebius’ capex needs for 2026 rise at a multiple of >4X YoY to meet these aggressive capacity expansion targets. This would also be ~7-8X current estimated revenue of $3.45 billion, meaning the company will be unable to break free of this extremely elevated capex cycle next year.  

Compare this to Nebius’ balance sheet, which currently show $4.79 billion in cash on hand and $4.1 billion in debt following the company’s recent combined debt and equity offering raising $4.2 billion. This is not likely not even 20% of what the company could need to reach its capacity targets, and thus its revenue targets.  

CFO Dado Alonso covered the financing aspect, though it is unlikely that these financing avenues could raise the necessary amount needed for these capacity targets without significantly stressing the company’s balance sheet:  

“In order to support our aggressive growth plans in 2026 and to maintain this pace of growth in 2027, we will be utilizing at least 3 sources: corporate debt, asset-backed financing and equity. We are in the process of raising asset-backed debt, which we'll be able to secure with attractive terms supported by creditworthiness of our largest customers. Tomorrow, November 12, we will be putting in place an at-the-market equity program for up to 25 million Class A shares and plan to file a prospectus supplement. We will evaluate the program regularly based on our capital needs. The program enables us to access equity funding on an efficient ongoing basis. However, we will remain dilution sensitive as we prepare to finance future growth opportunities.” 

Alonso’s comments suggest that the current cash on the balance sheet likely will go towards other data center opportunities outside of its hyperscaler deals, with the asset-backed debt more likely to fund the Microsoft and/or Meta buildouts in similar fashion to some Bitcoin miners raising substantial cash via asset-secured debt.  

The at-the-market program could provide around $2.25 to $2.5 billion in capital around current share prices, if exercised in full between $90 to $100/share, while leading to approximately 10% dilution. Considering management’s goal of remaining dilution sensitive, the ATM program may be utilized at higher share prices to raise more capital, or as a second source behind debt. Again, raising only a few billion via 10% dilution to shareholders is still far from enough from meeting estimated capex needs next year.  

While Nebius remains in a better position than CoreWeave at the moment in terms of debt-to-equity, at 0.94x versus 4.85x, there is a very high likelihood that this ratio will move rapidly in CoreWeave’s direction through 2026 given the capital intensity of building this capacity at an accelerated pace. There is also the potentiality for Nebius to be unable to raise $20B+ in capex given that it is approximately equivalent to the company’s current valuation and still 7-8X of revenue. 

Why Nebius Must Spend Aggressively – Capacity is Sold Out 

The reason that this capex growth is necessary, at least from management’s point of view, is because demand continues to far outstrip the capacity that Nebius can offer. Put another way – it is a spend or get left behind market.  

Management emphasized numerous times in Q3’s call that capacity is the main bottleneck to revenue growth, and their current main focus is adding capacity to remove this bottleneck. CEO Arkady Volozh explained that demand was very strong in Q3 with Nebius selling out of all available capacity, and each time capacity was brought online, it was sold. Nebius is currently “selling the remnants of Q4, but [also] now preselling new capacity being delivered in future quarters” in 2026, helping lock in future revenue growth. This would also include the company’s trance of GB300s coming online in Finland in Q4.  

One of the more important comments this quarter related to the demand pipeline. Management stated that pipeline generation, or customers wanting to buy capacity, expanded 70% QoQ to $4 billion in Q3, yet they “were only able to convert a portion of that given to the constraints of our capacity.” For context, core AI infrastructure revenue was approximately $131 million in Q3, so the pipeline would be ~30X its current run rate. It makes sense why Nebius is aiming to aggressively grow capacity given it witnessed well over $1.5 billion in sequential growth in the pipeline, as meeting a larger portion of this pipeline via more capacity would quickly translate to revenue growth.   

To that point, management said that lead times from power connection and start of GPU deployments to revenue generation range “anywhere from 6 to 12 weeks” or potentially faster in existing facilities, again underscoring why they are willing to pursue this rapid capacity growth through 2026 as they can quickly shift to revenue generation from power connection. 

However, this has highlighted a major downside to a capacity-constrained model. Nebius stated that it is “learning to say no to customers as we routinely sell out and have to actually let them down lightly and try to convince them to purchase in the future,” but there is an equal chance that these customers will simply go to CoreWeave or a hyperscaler who has capacity, get locked in to that ecosystem and not return to Nebius.  

It also sheds light on the puts and takes of Nebius’ deployment strategy at this small scale. By prioritizing hyperscaler deals with Microsoft and now Meta (though this was ultimately constrained by capacity), Nebius is locking in strong future revenue streams over the next few years, but turning away these smaller startup customers by also locking up a larger portion of its near term capacity to the hyperscalers.   

Building Core AI Cloud Business  

As we discussed in our prior analysis on Nebius for Advanced members, Nebius: Mega Microsoft Deal, 5x Growth in Power Fueling AI Cloud Hypergrowth, But High Risk Remains, the company believes its vertical integration and proprietary cloud serve as a key advantage and the ‘real future opportunity’: 

“In the longer term, Nebius believes that its vertical integration with a full stack of AI services will help broaden its customer base, increase platform stickiness and capture higher margin revenue and services. Nebius offers a proprietary cloud platform with managed MLops services, low downtime and high cost efficiency, combined with its inferencing platform AI Studio. With AI Studio, Nebius says it can offer up to 3x token savings with low latency, and up to 4.5x faster time to token versus other competitors in Europe.”offer up to 3x token savings with low latency, and up to 4.5x faster time to token versus other competitors in Europe.” 

As such, the company is prioritizing building out its core AI cloud platform and continuously adding new features, with the predominant goal currently being geared towards enterprise adoption. Two main features from Q3 include Nebius’ new enterprise-ready cloud platform, dubbed Aether, as well as Nebius Token Factory.  

With Aether, Nebius brings enterprise-grade security and compliance features along with a comprehensive observability suite, developer tools and more, in an effort to make its platform more attractive and accessible for large enterprise customers. Nebius also is focusing on improving reliability of its network with active health checks to reduce maintenance tasks, and boosting performance and storage speeds.  

Token Factory builds on Nebius’ AI Studio and embeds enterprise-grade security in a production-scale inference platform, letting customers run open-source AI models from OpenAI, Alibaba, Meta, DeepSeek and others with 99.9% uptime, per the company. Management says Token Factory will help customers “transform open source models into optimized production-ready systems with guaranteed performance and transparent cost per token” and the “best total cost of ownership.” While it is too early to see how Token Factory contributes to monetization, Nebius’ dedication to improve its platform and expand its suite of offerings can serve as a strong differentiation and potentially aid in customer acquisition over rival platforms. 

Financials 

Revenue to Reaccelerate 

Following the deceleration in Q3, Nebius tightened its 2025 revenue outlook, from $450-630 million down to $500-550 million, citing timing of capacity as the primary reason. This is also below current estimates for 372% growth to $555 million. 

For fiscal 2026, however, revenue is expected to reaccelerate to 521% to $3.45 billion, with estimates having doubled since Nebius signed the Microsoft deal in September. Nebius said this quarter that it plans to provide full-year guidance for 2026 in the upcoming quarter. 

Margins Improving, but Widely Negative 

Gross margin was 70.6%, down slightly from 71.3% in the prior quarter and up from 69.2% in the year ago quarter. 

Operating margin began to show signs of improvement, coming in at (89.1%), compared to (105.8%) in the prior quarter and (251.1%) in the prior year. At this rate, Nebius could break even in five quarters assuming it can maintain such improvements consistently, though this may be challenging as capacity ramps up rapidly. 

Net margin was (81.9%), not comparable to the prior quarter’s 556% on Toloka’s deconsolidation but up from (293.5%) in the year ago quarter. Adjusted net margin was (68.7%), improving from (123.7%) in the year ago quarter. However, it is important to note that net losses has widened (from ($39.7 million) to ($100.4 million) for adjusted net loss) though margins are showing improvement from the rapid revenue ramp. 

Adjusted EBITDA improved to just ($5.2 million) or a (4%) margin, up from a (20%) margin last quarter. Nebius noted that its Core AI Infrastructure business continued to generate positive adjusted EBITDA at a nearly 19% margin in Q3, with the metric weighed down by Avride and TripleTen investments.  

Nebius lags CoreWeave by a significant degree for adjusted EBITDA, with CoreWeave posting a 61% adjusted EBITDA margin in Q3, down 4 points YoY. This suggests that there is room for substantial expansion over the upcoming quarters as the business scales to a much larger size. 

Earnings Remain Far From Profitability 

Nebius reported a 23% beat on adjusted EPS in Q3, though the company remains far from profitability and is not expected to reach profitability for quite some time.  

Q3 adjusted EPS was ($0.40), beating estimates for ($0.52) but widening slightly from ($0.38) in the prior quarter. Looking ahead to Q4, adjusted EPS is expected to be ($0.58), before widening to ($0.74) in Q1. 

For fiscal 2025, adjusted EPS is projected to be ($1.73) before widening to ($2.29) in 2026, likely driven by increasing expenditures to quickly ramp capacity. 

Cash Needs Increasing 

The challenge for Nebius is very similar to that of CoreWeave, with the neocloud spending significantly on GPUs and raising substantial debt to fund said spending. As a reminder, Nebius spent nearly $1 billion in capex in Q3, up from $510 million in Q2 and representing more than 6X its revenue. Capex is on pace to be >9X of revenue this year and potentially remain at 7-8X of revenue next year. 

Operating cash flow was ($80.6 million) for a (55.2%) margin, while free cash flow was ($1.04 billion) for a (709.1%) margin. This widened from free cash flow of ($678.3 million) in the prior quarter as capex surged more than 87% QoQ to $955.5 million. 

Cash and equivalents totaled $4.8 billion, while debt was $4.1 billion. As mentioned previously, surging capex this year and the potential for tens of billions next year means Nebius will likely turn to debt markets for significant funding. Debt to equity sat at 0.94X in Q3, though this is likely to worsen significantly moving through 2026 based on estimated capex needs north of $20 billion.  

Valuation 

Nebius trades at 37x forward PS ratio, slightly above its average of 32x, though data is limited considering its recent launch on the public markets post-Yandex breakup. Shares have traded as low as 8.6x forward PS and as high as 61.5x. 

Rapid revenue growth in 2026 is expected to bring forward PS down to 5.9x next year, though this remains a premium to CoreWeave at 3.0x next year’s revenue estimate of $12.07 billion. 

Conclusion 

Nebius is on a trajectory of high growth-high debt for the foreseeable future, with the company spending nearly $1 billion this quarter on capex alone, in preparation for capacity ramps for Meta and Microsoft occurring in quick succession.  

On one hand, revenue growth is expected to accelerate sharply to 750% over the next two quarters and maintain a hypergrowth profile with >380% growth for the next five quarters, while YoY growth is expected to accelerate 148 points to 521% in 2026. However, capex guidance for 2025 was raised to ~9X of revenue, and aggressive capacity expansion targets for 2026 mean capex will likely remain 7-8X of revenue, far above peers and making finding financing a significant challenge. 

Ultimately, even if high-beta stocks catch a bid, there are far cleaner and less capital-intensive ways to gain exposure to the AI buildout. Until Nebius can demonstrate that its growth is not being funded by an expanding cash shortfall, we see limited risk-reward and will remain on the sidelines.

Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.

Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.

Recommended Reading:

• The I/O Fund’s Top 10 New Ideas List for Q4 2025
• Lumentum: EMLs Driving Results, CW Lasers Ramping with Q2 Guided for 22% QoQ Growth
• Cloudflare: Revenue Accelerates to >30%, Key Metrics Strengthen
• Talen Energy: Acquisitions May Unlock Data Center Deals, Amazon Ramping Soon