Coherent supplies components for datacom transceivers and for optical interconnects. Transceivers and optical interconnects convert electrical signals into optical signals for fiber optic networks within the data center. Traditionally, optical links have been used for compute and storage servers, yet AI/ML servers are driving an increase in demand for the optical interconnects and transceivers that Coherent supplies.
Transceiver speed has been growing with the highest data rates ranging from 100G to 200G to 400G. AI servers are driving a market for 800G data rates, which are shipping in production now, and 1.6T rates, which are shipping in 2025. We’ve been covering these AI-driven upgrades around optical networking in our Marvell write-up and Semtech write-upwrite-up and Semtech write-up.
Coherent’s transceivers work with both Ethernet or InfiniBand, as well as proprietary protocols such as Nvidia’s NVLink and Nvidia’s interconnect chips NVSwitch. The company has stated that their 100ZR pluggable transceivers can upgrade old 10GBps Ethernet links with 100 GBps at the optical network edge, representing a 10X upgrade.
Coherent designs and manufactures the components, such as lasers, detectors and passive optics. Manufacturing the components (as opposed to buying them) is a strength as the company has supply chain resiliency by controlling the end-to-end process, and the manufacturing takes place primarily in the United States with some in Europe. Notably, the discussion as to where AI chips and their components are manufactured is setting up to become a hot topic come 2025.
Overview of Coherent’s Products:
Last quarter, Coherent’s networking revenue increased 12% sequentially and 61% year-over-year to $763 million. The strong growth was led by the AI-related datacom segment, which grew 16% sequentially and 89% year-over-year. The company has seen strong rebounds in this segment since September of last year, partly due to being a supplier for Nvidia’s NVLink and InfiniBand. Coherent also supplies components for Ethernet networking.
Coherent provides laser technologies for 100-gig to 400-gig, plus the AI-centric 800-gig optics and the AI-centric 1.6T optics. Of these, the 800-gig is the primary interconnect for AI deployments with 1.6T shipping in 2025. Artificial intelligence and machine learning drive demand for the 800-gig PAM to increase the speed of input-output and to process the data flows. This doubles the throughput (bandwidth) due to an 8x100Gpbs optical transceiver for inside and between AI clusters.
Marvell was first to launch the 1.6T solution 5nm and 3nm with 200-gig per lane for both. The highly newly launched 1.6T 3nm optical DSPs will reduce power consumption, something that Blackwell is breaking the upper limits of with an estimated 140kW per rack. Right now, any component that lowers power consumption will be in high demand come Q1-Q2. Coherent is expected to launch their solution next year, per the earnings call: “Having delivered initial samples in the preceding quarter, we continue to expect to begin ramping sales of 1.6T datacom transceivers in calendar 2025.”
Although Marvell and Broadcom are the two that first come to mind when discussing PAM4-based networking and DSPs, it’s important to remember that Nvidia is a leader in AI networking in their own right. Last quarter, Nvidia reported networking revenue of $3.13 billion while Broadcom most recently reported AI revenue of $3.7 billion for the quarter. If we remove the $300 million for custom silicon, Broadcom reported $3.4 billion in AI networking. Nvidia is a newcomer to the field following the acquisition of Mellanox in 2019.
Coherent supplied content is likely to increase with Nvidia as Blackwell is a catalyst for this partnership. With the upcoming release of Blackwell, NVLink speeds will double from the fourth-generation to the fifth-generation. We pointed out a few months back the significance of NVLink’s fifth-generation as it’s the most important generational leap to-date for Nvidia’s networking products:
“NVLink Switch is a major component to the Blackwell upgrade. Fifth-generation NVLink enables multi-GPU communication at high speed, reaching 1.8 TB/s bidirectional throughput or 14X the bandwidth of PCIe for a single GPU.
For the NVL72 systems, NVLink Switch can reach 130 TB/second, which is “more than the aggregate bandwidth of the internet.” Therefore, it’s the compute and the communication capabilities of the upcoming GB200 release that are important to consider. The 72 GPUs in the NVL72 can be used as a single accelerator for 1.4 exaflops of AI compute power.”130 TB/second, which is “more than the aggregate bandwidth of the internet.” Therefore, it’s the compute and the communication capabilities of the upcoming GB200 release that are important to consider. The 72 GPUs in the NVL72 can be used as a single accelerator for 1.4 exaflops of AI compute power.”
The result is that the GB200 will deliver a 30X speedup for 1 trillion+ parameter models by leveraging FP4 precision and fifth-generation NVLink. The NVL72 rack-scale systems are exascale computers that will contain up to 5,000 NVLink cables for up to 2 miles of networking.
To put it simply, Nvidia is set to report roughly $200 billion in AI systems revenue next year (we think it’ll be higher than this), with the GB200 systems driving a surge in demand for Nvidia’s in-house networking solutions. There is no other player in the ballpark or zip code of this AI accelerator revenue, and thus, this year will offer a unique opportunity for Nvidia to greatly increase its networking segment driven the scale-out and scale-up of GPUs. That is to say, the 4.5X to 9X increase in systems from 8 GPUs to 36 GPUs or 72 GPUs, plus the scale-out toward clusters with 100,000-GPUs and eventually million-plus GPUs.
Below are a few products that Coherent offers that are set to benefit.
Coherent’s AI-Related Products:
Coherent has a highly technical suite of hardware products. To streamline the discussion, I’ll focus on the three product lines that management calls out as major contributors to the rebound in its AI-related datacom segment (part of the larger networking segment).
- Next-generation transceiver technologies, such as VCSELs, EMLs and CW lasers for silicon photonics – reported in the networking segment
- Datacom optical switch platform – reported in the datacom segment
- Data center interconnects (DCIs) – not part of the AI segment, rather this is reported in the telecom segment.
Next-Generation Transceivers:
Last quarter, Coherent’s networking revenue increased 12% sequentially and 61% year-over-year to $763 million.
VCSELs: >100 meters Distance:
Vertical Cavity Surfacing Emitting Lasers (VCSELs) are for link distances of less than 100 meters. As Coherent explains, these are the lowest-cost, lowest power consumption solution. In March, the company announced its 200G VCSELs that replace the oxide-aperture with a lithographic-aperture. When combined with a DSP, the 200G VCSELs allow for a 1.6T multimode optical transceiver for AI, high-performance computing (HPC) and networking applications.
Silicon Photonics >100m: Nvidia to Launch SiPho Product H2 2025
For distances greater than 100m, silicon photonics-based transceivers are used. We wrote about silicon photonics previously before Nvidia Hopper had shipped in volume. Back in early 2023, an article from Next Platform discussed the previous generation of GPUs, where the NVSwitch fabric on the DGX-A100 systems has “a range of about 300 centimeters and moves data at 8 picojoules per bit. The goal is silicon photonics to do it at half the energy and boost the range to as far as 100 meters between devices.”
Coherent recently announced a continuous wave (CW) InP laser for silicon photonics that “are on the forward edge of data transmission technology, crucially addressing the explosive bandwidth demands brought on by the rapid adoption of machine learning networks within AI-driven data centers.” According to the press release, Coherent’s CW lasers will result in 15% better power efficiency and are more reliable due to eliminating aluminum.
The first CW laser to launch was in September of 2024 with a higher-power laser for cooled operations launching in 2025. The press release cited that production capacity is expanding 500% over the next two years: “And with our upcoming expansion to a state-of-the-art 6-inch InP fab in Sherman, Texas, by 2026, we expect production capacity to be 5x our current production rate, a significant ramp given the complexity of this material.”
At the end of September, Coherent demonstrated a new 1.6T-DR8 transceiver module running a 5nm Nvidia chip. The SiPho product, or silicon photonics product, will read and write data at the speed of 1.6Tbps. The initial reaction from analysts is that this product will become more impactful at the end of 2025, taking up to 10% to 20% of the market in 2026, pointing toward Nvidia customers Microsoft, Coreweave and Lambda as being names that would drive the sales come 2026.
EML Lasers:
Electro-absorption modulated lasers (EMLs) help to enable the 200G per lane transmission, which is enabling the 1.6TBps data rate. In the future, it is likely AI networking will expand to a 3.2TBps data rate. The EML device that Coherent supplies offers signal integrity for 200G optical transmission due to the wide bandwidth of the modulator and on-chip integrated radio frequency (RF) termination.
EMLs were traditionally used by telecom customers, yet became attractive for AI servers due to meeting the 200G per second speeds necessary to support AI models and capabilities for 1.6T optical modules. EMLs are more expensive than CW lasers for silicon photonics, with 1.6T SiPho being the more popular choice. Where four EML lasers are required to run 1.6T, two less-expensive CW lasers are needed to run four channels with a silicon photonics module.
Coherent’s press releases point toward EML lasers becoming more impactful with the push toward reading and writing data at 3.2TB per second speeds.
When asked on the call which product will be most influential at 1.6TB ramps, Coherent’s management was declined to state it would be primarily SiPho, rather stating it would be a mix of all three:
Q: “[…] And then if you think that the world is kind of ramping on 1.6T, do you guys see kind of more of a silicon photonics world or an EML world as that 1.6T ramp?”
A: “[…] We look at it as we'll deploy whatever the best technology is for the benefit of the customer and the application that we're trying to drive. So, whether that's an EML, a VCSEl, or silicon photonics, we're developing all of those different options and we'll deploy whatever technology is strongest to create the biggest differentiation for our products and the biggest benefit for our customers. So, that's kind of the approach that we take to the technology. And I think we've got the broadest set of technology options of certainly any of our peers and competitors. So, I think that's a real competitive strength for us.”
Datacom Optical Switch Platform (OCS Switch):
Optical current switches are a new kind of switch for AI clusters that handles the switching optically instead of using transceivers to convert photons to electrons, and back again. According to the earnings call, “Our differentiated switch is based on our highly reliable solid-state liquid crystal technology and was recognized at ECOC '24 with the Best Product Award for data center innovation. We've shipped sample units to key strategic customers and we expect to begin ramping revenue in calendar 2025.”
According to a Dell’Oro analyst, optical switches are a new kind of switch that offers high bandwidth, low latency and are less expensive to operate. Per Coherent: “this kind of optical switching tends to be more reliable – something that will pay dividends in very larger clusters in which mean time to failure tends to be quite low.”
At Hot Chips 2023, Google’s TPU group technical lead, Andy Swing, explained that Google has been able to “switch together very large quantities of AI accelerators” in pods with 64 TPUs. According to Google, the optical circuit switch results in a sizable boost in network bandwidth and can be reconfigured if a node fails. It doesn’t take much imagination to consider other hyperscalers are likely to follow in Google’s footsteps by using an OCS switch as cluster sizes for AI accelerators are set to increase.
Datacom Interconnects (DCIs):
For long distance transmission, Coherent offers data center interconnects (DCIs). This can range up to hundreds of kilometers. There was a recent press announcement that Marvell, Lumentum and Coherent worked on optical modules with transmission links of up to 500km.
DCIs are recognized under the telecom segment, which saw growth of 9% sequentially and 17% year-over-year. Within this, management stated: “We are definitely seeing strong demand signals in DCI, right? And obviously, that's only a portion of the telecom market, but we are seeing very strong demand signals there.”
Financials:
Coherent’s Q1 FY2025 revenue grew by 28% YoY to $1.35 billion, beating estimates by 2.4%. Revenue accelerated by 18.9 percentage points from the 9.1% growth reported in the June quarter. The strong growth was primarily led by AI-related datacom revenue, which grew 16% sequentially and 89% YoY.
Management has guided FQ2 revenue of $1.37 billion, representing 21.1% YoY growth at the midpoint. Analysts expect growth to sustain 15.6% YoY growth to $1.40 billion in FQ3.
Looking further out, analysts expect FY2025 ending in June revenue to grow 17.9% YoY to $5.55 billion and 11.2% YoY to $6.17 billion in FY2026.
Revenue by End Market:
Communications
Communications end market FQ1 revenue grew by 68% YoY and 14% sequential growth to $774 million. It is the largest end market and constitutes 57.4% of total revenue. Revenue accelerated by 49 percentage points from 19% growth in FQ4. The strong growth was led by datacom and telecom revenue.
FQ1 datacom revenue grew by 89% YoY and 16% sequentially due to strong AI data center demand. The company witnessed a continued ramp for 800-G transceivers, benefitting from the increased number of customers ramping 800-G transceivers. Management expects strong growth in 800-G transceivers to continue in the coming quarter with revenue from 1.6T transceivers starting to ramp in CY2025. The company delivered initial samples of 1.6T transceivers in FQ4.
Telecom revenue grew by 17% YoY and 9% sequentially, primarily due to end market improvement and the ramp of the company’s new products, namely 100G ZR and 400G ZR+ Coherent transceivers. Although management expects the telecom end market to remain weak in the near term, they expect a ramp in revenue from these new products over FY2025. Overall, they expect the communications market to be a long-term growth driver for the company.
Industrial
The industrial end market is the second largest end market and constitutes 30.2% of revenue. The industrial end market grew by 1% YoY and down (-6%) QoQ to $407 million. Revenue decelerated from 5% growth in FQ4 as the strength in display capital equipment was offset by the weakness in the precision manufacturing. The company faced demand headwinds in the precision manufacturing, mirroring broader industry trends. However, despite the near-term weakness, management “expects the industrial market to be a long-term growth driver for the company as the end markets recover and as our new products continue to ramp.”
Instrumentation
The Instrumentation end market constitutes only a small 6.8% of total revenue. FQ1 revenue was down (-8%) YoY and (-10%) sequentially to $91 million.
Electronics
The Electronics end market also constitutes only a small 5.6% of total revenue. FQ1 revenue was down (-16%) YoY and (-27%) QoQ to $76 million.
Segments
- Networking FQ1 revenue reported strong growth of 61% YoY and 12% sequentially to $763 million due to the AI data center demand. Revenue accelerated from 16% growth in FQ4.
- Lasers segment revenue increased 4% YoY and down (-2%) QoQ to $348 million. Revenue decelerated from 7% growth in FQ4.
- Materials segment FQ1 revenue was down (-3%) YoY and (-15%) QoQ to $237 million.
Margins
Margins are expanding yet they remain low compared to other semiconductor companies, and this will remain a predominant concern with Coherent. Management has set a goal of achieving a consistent adjusted gross margin level of above 40% and has two important initiatives: pricing optimization in the industrial business (includes industrial end market, instrumentation, and electronics) and product cost reductions for the datacom transceiver business.
Per the earnings call:
“[…] And I do see opportunity to do a much better job of optimizing the pricing of the [industrial] products and capturing what I would say — what I would call is the fair value for the technology and the innovation that we're bringing to those industrial markets.
Now, in pricing in the datacom transceiver space, there I think there's, I would say, there's not as much opportunity on the pricing side. But what I would say is, but there's definitely opportunity on the product cost side. And that's one of the examples I gave earlier on the call is on product costs within transceivers. And I highlighted, I believe this on the last earnings call, yields as definitely an opportunity. And Sherri and I, as I said earlier, we're in an operational review this morning, spending time talking to the team about yields, the improvements that they've driven over the past few months, and what we need to see in terms of yield improvements moving forward as well. And that is definitely an area of focus for us. So, I would — back to your datacom transceiver question in particular, I would say maybe not so much on pricing, but definitely there's opportunity for us in cost and we're certainly very focused on that.”
- FQ1 gross margin improved five percentage points YoY to 34.1%. Adjusted gross margin improved 2.9 percentage points YoY to 37.7%, primarily helped by higher revenue, favorable product mix, and yield improvements.
- Management has guided for an adjusted gross margin of 37% at the midpoint for FQ2 compared to 36% in the same period last year. The CFO provided more clarity during the earnings call Q&A on the gross margins and assured to provide more details before the Investor Day in May 2025 on its plan to achieve the long-term goal of above 40% gross margins.
“When we look at Q2, the guide for Q2, that is a range, right, 36% to 38%, it is a range, and there can certainly be fluctuation on a quarterly basis with respect to gross margin. But we did talk about, last quarter; Jim mentioned that we launched our gross margin expansion strategy which includes product pricing optimization, as well as product cost reduction. And so that's an area where we're going to focus on because we want to achieve a long-term gross margin of greater than 40%, and so that's really how to think about what our goal is for our long-term gross margin.”
- FQ1 operating margin improved significantly to 5.6% from (-2%) in the same period last year. Adjusted operating margin improved 4.7 percentage points YoY to 17.3%. Management guide for the next quarter is 16.2%. Even though the guide is down sequentially, it is up from 15.2% in the same period last year.
- The operating margin is expected to improve along with the gross margins in the long term. The company is prioritizing its R&D investments in the growth areas and divesting non-profitable businesses. Another key takeaway from the earnings call is the CEO’s comment, “I just want to reiterate what I said in the prepared remarks that even though the overall non-strategic category is a relatively small part of our revenue, again, it is dilutive to our operating margins.”
- Net income was $25.9 million or 1.9% of revenue compared to a net loss of (-$67.5) million or (-6.4%) of revenue in the same period last year. Adjusted net income was $149.7 million or 11.1% of revenue compared to $55 million or 5.2% of revenue in the same period last year.
EPS
Coherent has strong EPS growth. Analysts expect EPS growth to continue going forward, and there is potential for further upgrades to the EPS estimates due to the new CEO’s streamlining/restructuring plan.
- FQ1 adjusted EPS grew by 357% YoY to $0.74, primarily helped by operating leverage, favorable product mix, improvement in yield, product price optimization, and product cost reductions.
- Analysts expect adjusted EPS to grow 86.3% YoY to $0.67 and 43.2% YoY to $0.76 in the subsequent two quarters.
- Looking further out, analysts expect the adjusted EPS for FY2025 ending June to grow 80.5% YoY to $3.01 and 45.1% YoY to $4.37 for FY2026.
- Jefferies analyst believes that the new CEO’s streamlining and cost reduction initiatives could potentially double the adjusted EPS from the current about $3.0 to as much as $6.40 annual EPS in 2026.
Cash Flow and Balance Sheet: 2.4X Debt Leverage
Operating cash flow margins have been hovering around 10% of revenue and are expected to improve with higher revenue and profits in the coming quarters.
- FQ1 operating cash flow was $152.98 million or 11.4% of revenue compared to 18.9% in the same period last year. The operating cash flow margin was in line with the full year operating cash flow margin of 11.6% for FY2024 and 12.3% for FY2023.
- FQ1 free cash flow was $61 million or 4.5% of revenue compared to 13% in the same period last year. The free cash flow margin was in line with the full year free cash flow margin of 4.2% for FY2024 and 3.8% for FY2023.
- Cash was $1.02 billion; debt was $3.99 billion compared to $926 million and $4.1 billion at the end of FQ4.
- During FQ1, the company repaid $118 million in debt from the cash from operations and proceeds from the sale of the Newton Aycliffe fabrication facility. The sale is part of the company’s ongoing efforts to streamline operations.
- With the $118 million debt repayment, the company has reduced the debt leverage ratio to 2.4x, and debt reduction will continue to be the priority for the management.
Divestment
The company is looking to divest or shut down non-strategic product lines and assets. The company recently sold the Newton Aycliffe fabrication facility in the UK. The proceeds from the sale of the facility were used to repay debt and will also help to reduce the overhead expenses and interest expenses.
Similarly, it is looking for strategic alternatives for the battery technology business. Although the business accounts for a small percentage of revenue, it is dilutive to the company’s margins. It will allow the company to focus on investments that have better returns on investments, such as the AI data center transceiver business. The management is expected to provide more updates in the coming months and during the Investor Day in May 2025. The company might also divest the silicon carbide business, in which Denso and Mitsubishi Electric have a 25% non-controlling stake.
New Management from Lattice Semiconductor:
Jim Anderson was appointed CEO of Coherent Corp. and a member of the Board of Directors on June 3, 2024. He previously served as CEO of Lattice Semiconductor Corporation since September 2018. Prior to joining Lattice, Mr. Anderson served as the Senior Vice President and General Manager of the Computing and Graphics Business Group at AMD.
Jim, who is credited for bringing about a turnaround in Lattice Semiconductor, is also expected to repeat his success at Coherent along with the new CFO Sherri Luther. The company appointed Sherri Luther as CFO on Oct 11, 2024. Prior to that, she was the CFO of Lattice Semiconductor since 2019. Prior to joining Lattice in 2019, she worked at Coherent for 16 years, including as Corporate Vice President of Finance.
Commentary from FQ1 2025 Earnings Call:
800-Gig drove the Revenue Beat, 1.6T Shipping in 2025:
There were questions as to where the upside from the quarter came from, given the top line and sizable 20%+ bottom-line beat. As noted, there was sizable growth in the datacom transceiver business of 16% QoQ growth and 89% YoY growth. The CEO points out customer breadth is expanding on the 800G product.
Q: Simon Leopold, Raymond James: “[…] And I guess what I'm trying to understand is where was the, really, upside surprise this quarter, was it really 800 gig and above or was there more strength from the more traditional products below — 400-gig and below, did that provide any upside or was it all coming from the higher performance? Thank you.”
A: Jim Anderson, CEO: “We did see sequential growth there [400-gig and under speeds]. That was very nice to see. When you look across the customer base, customers are at different stages of adopting the different transceiver speeds. So, we still have customers that are doing significant volume on 400G and below as well. And so, it's really a mix of different transceiver speeds.
And then, back on 800-gig, I would say, look, we're really pleased with the ramp, the overall ramp of our 800-gig transceivers. And then, the other color I would add is that one of the things I'm really pleased to see is the breadth of customers that we have. The number of customers that are ramping 800-gig has significantly increased. If I look like a year ago, it was only maybe a couple customers. Now, we have many customers ramping 800-gig. So, there's a much bigger diversity of revenue streams underneath that 800-gig ramp. And we do expect 800-gig to continue to grow over the coming quarters as well.”
Although the customer breadth has grown with 800-gig, it was stated on the call the 1.6T will launch with a customer breadth that is smaller and matches the start of the 800-gig.
Supply Constraints Could be a Catalyst for Coherent:
There were three questions on the call from three separate analysts about potential supply constraint in lasers, with two analysts calling out the supply constraint being specifically with EMLs: “certainly on the EML side as we're hearing about constraints, et cetera, as you think about '25 and '26?”
The CEO pointed out this is an area incremental strength for Coherent as they make most of their parts internally (which is quite rare) yet will source when the demand requires it. The CEO also pointed out their breadth of technologies helps to meet demand if there were to be a shortage in a specific laser, such as EML.
Here are a few statements from the CEO that addressed Coherent having an advantage by building their own products in a fairly crowded market of networking components:
“On that first area of technology roadmap, I think our customers really recognize the breadth and the depth of the technology portfolio that we can bring to bear, specially in the optical networking space, where we don't just assemble the modules, but we build a lot of the ingredient components that go into the module; the lasers, whether they're VCSELs, EMLs for silicon photonics that we design, or a lot of the other ingredients that go into those modules […] and then, our verticalized structure could be a real advantage, especially in a very fast ramp situation, which we're in right now with our datacenter customers. When demand is increasing very quickly, it's really important to have that verticalized strategy and structure that we have, because I think that's really allowed us to supply them in a really reliable.”
This statement was reiterated a few times, which is that Coherent’s advantage lies within the breadth of the technologies they offer alongside the capability to build internally.
United States Domiciled:
The discussions around supply constraints potentially helping Coherent as the company builds its own products are further supported by the fact Coherent components are primarily manufactured in the United States, with some in Europe. The company has plans to expand its presence in the United States with a 6-inch InP wafer fab in Sherman Texas. According to an announcement last week, Coherent was awarded $33 million from the United States CHIPS act to support the expansion of the Texas facility.
Competitors:
According to the 10-K, the company “had one customer who contributed more than 10% of revenue during fiscal 2024.” This is understood to be Nvidia. Yet, B.Riley downgraded the stock due to Nvidia seeking more suppliers for the 1.6T products, with the analyst note naming Eoptolink, Innolight and Fabrinet as notable competitors.
Conclusion:
If you had “reading about AI networking components” on your Christmas list, then consider yourself in luck. The I/O Fund is leaving no stone unturned in delivering a full hardware stack analysis on what’s to come in 2025 as AI systems increase in complexity. Although we saw some outliers in AI software this year, we continue to foresee 2025 to be the year for AI hardware – far more so previous years. This goes back to how the Blackwell systems are being built, by scaling out 4-9X, and also requiring new components to handle the surge in AI server power consumption.
Of course, with every great opportunity, there comes some element of risk. The clear risk to semiconductors are tariffs with rumors that Chinese tariffs could be as high as 60% on imports from this region. This would technically be a tailwind for Coherent — a key Nvidia supplier as we go into 2025 that would not only circumvent the majority of tariffs but potentially come out on top.
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.
Additional Readings: