Hewlett Packard Enterprise (HPE) has quietly undergone a business transformation over the last few years to position itself as a beneficiary in AI servers, networking, hybrid cloud and AI software. Last quarter, HPE’s AI systems revenue doubled sequentially to $900 million with a backlog of $3.1 billion. Compare this to Dell with AI server order revenue of $2.6 billion and AI server shipments of $1.7 billion. Super Micro has total revenue of $3.85 billion with “more than 50%” of this from AI or about $2 billion.

As a percentage of revenue, SMCI is certainly in the lead as a pure play. However, HPE is ahead of Dell in terms of percentage of revenue at 12.5% for HPE and 7% for Dell.

We’ve identified liquid cooling as a leading trend for the back half of 2024. Of those at the cross-section of servers and liquid cooling, HPE has the lowest valuation to the tune of being priced up to 67% lower.

HPE Overview:

HPE has over 13,000 patents, with 300 of these in particular in liquid cooling systems for data centers, and is the owner of four of the world’s top 10 fastest supercomputers. It has also made a number of partnerships, such as with Microsoft Azure, Google Cloud, SAP, and Nvidia to offer seamless cloud experiences or more powerful compute capabilities. HPE operates globally with over 700 channel partners giving it a unique edge in go-to-market capabilities.

HPE has five business segments:

1. Server (53% of revenue in Q2’24): HPE sells servers for general-purpose use through their ProLiant line and more compute-intensive applications such as their Cray line, which is used in supercomputers.

2. Hybrid Cloud (17% of revenue): HPE offers cloud-native and hybrid solutions for customers that would prefer not to host their own servers on-premise. This includes data storage and management, hybrid and cloud-native services through HPE GreenLake, and AI infrastructure as a service.

3. Intelligent Edge (15% of revenue): HPE provides solutions that enable faster data transfer and improved data analytics through edge computing. AI workloads will eventually move to the edge. Intelligent Edge also offers security features like Zero Trust. This segment includes HPE Aruba which is a subsidiary focused on networking capabilities supported by HPE’s planned $14B merger with Juniper Networks, announced in January 2024.

4. Financial Services (12% of revenue): HPE provides investment solutions to allow businesses to deploy technology models and acquire IT solutions.

5. Corporate Investments and Other (3% of revenue): HPE provides consulting and implementation services.

HPE Cray’s supercomputers are high performance computing servers for AI workloads. They are available with both Nvidia and AMD GPUs, and currently rank as the world’s fastest and largest supercomputers. Cray was a supercomputer manufacturer founded in 1972 before it was acquired by HPE in 2019. Cray is roughly 80% government and agency contracts and 20% commercial, with the bulk of the line being used such for supercomputers for research labs within the Department of Energy, among others.

The ProLiant servers accelerate workloads from the data center to the edge, and are used by corporations for hybrid AI. The AI rack servers offer memory intensive AI inferencing and scalable GPU acceleration for enterprise AI. There are SKUs for virtualized workloads for edge applications that offer balanced bandwidth and memory, data intensive workloads for apps that require large storage capacity and high bandwidth, and compute and data storage demanding workloads that require a maximum core count, among others.

GreenLake cloud is a hybrid architecture that allows enterprises to maintain control of sensitive data, while leveraging the benefits of public and private clouds. HPE calls this an edge-to-cloud platform, which offers software-as-a-service for storage, a data control plane to have a complete view of data assets, data analytics, and the ability to scale AI pilots and large language models. The software platform offers the ability to create a private cloud alongside on-premise servers, which may become quite popular as it allows enterprises to balance the security and data sovereignty that on-premise offers with the agility and scalability of the cloud. Among other things, HPE GreenLake lets customer privately train and tune large language models.

HPE Edgeline offers edge computing and processing power in a compact server that is located close to where data is generated. Edgeline servers are important for edge computing where data is quickly turned into intelligence with AI, and to manage AI/ML applications while protecting and governing data.

Aruba ensures reliable and secure connectivity from the edge back to the servers at the datacenter. The most recent Juniper Networks acquisition expands HPE into ethernet networking and switches. HPE’s $14B merger with Juniper Networks is expected to close in late 2024 or early 2025 and will strengthen HPE’s portfolio in AI networking.

Financials:

Hewlett Packard Enterprise (HPE) surprised in its latest earnings report, posting revenue and non-GAAP EPS that both exceeded outlook in Q2 as AI systems revenue more than doubled sequentially. Management also raised full year revenue and non-GAAP EPS guidance, signaling confidence in their ability to convert the current backlog orders to revenue.

Revenue and EPS:

HPE is a low growth company that has promising, initial signs of AI demand.

The company reported $7.2B in revenue, representing 3.3% YoY growth (5.45% beat), which was the biggest revenue surprise in the last 6 quarters. The company guided for $7.6B in revenue at the midpoint for this upcoming quarter, representing 8.5% YoY growth (2.1% beat).

The company reported ARR of $1.5 billion, up 37% from the prior-year period and 39% in constant currency. Last quarter, ARR grew 41% year-over-year to more than $1.4 billion in Q1, with the company stating to expect ARR growth of 35% to 45%.

GAAP EPS of $0.24 was down 25% YoY and down 17% QoQ.

Adjusted EPS of $0.42 vs $0.52 last year (8.2% beat) is down 19% YoY and was down 13% QoQ. The company guided for adjusted EPS in the range of $0.43 to $0.48 vs $0.49 last year.

HPE’s FY24 revenue guidance represents 1.5% YoY growth at the midpoint, a slight rise from Q1’24 where they guided for 1% YoY growth at the midpoint in constant currency. Adjusted EPS guidance for FY’24 was in the range of $1.85 and $1.95, higher than the guidance given during Q1 results of $1.82 and $1.92.

Key Segments:

The revenue beat was driven by strong performance in AI-related revenues. AI system sales more than doubled sequentially to over $900M for 12.5% of revenue, and the number of enterprise AI customers tripled YoY. AI systems accounted for all of the QoQ growth given the QoQ declines in Hybrid Cloud and Intelligent Edge.

Server Revenue of $3.9 billion dollars in Q2 represents a 16% increase sequentially and up 18% year-over-year, driven by AI servers and HPE Greenlake revenue.

HPE continues to expect sequential growth in both their traditional and AI server business. Operating margins were 11% in this segment, down 340 bps YoY due to pricing headwinds on AI systems, but it is in-line with their long-term operating margin guidance range of 11% to 13%.

HPE has reduced lead times for delivering Nvidia H100 solutions to six to 12 weeks, from over 20 weeks in Q1’24, which it expects will further boost revenues in H2’24, along with more large enterprise orders.

The backlog remained stable at $3.1B, down from $3.4B last quarter but up from $1.4B in the prior year quarter. This acceleration in AI systems revenue comes alongside a recovery in traditional and cloud infrastructure markets, creating a strong set-up for further acceleration into H2. This was also discussed in the Q&A with excerpts below.

Management also pointed to a growing enterprise customer base as evidence of its products’ value proposition.

“Our differentiation – with liquid cooling, software, HPE GreenLake, and increasingly services – is resonating in the market. We have seen a threefold increase in our enterprise AI customer base in the past year.” We have seen a threefold increase in our enterprise AI customer base in the past year.”

Hybrid Cloud revenue was $1.3 billion, down 8% from the prior-year period in actual dollars and 9% in constant currency, with 0.8% operating profit margin, compared to 1.9% from the prior year period.

This decline is being driven by two ongoing transitions. First is from hardware storage to HPE’s cloud-native Alletra storage solution, which reduces current revenues but leads to more predictable recurring revenues with storage ARR up 50% YoY.

Second is the transition from block storage to file storage driven by AI and is seeing strong progress with the pipeline of file storage deals tripling sequentially. Operating margins were down 110 bps YoY to 0.8% due to the decline in revenue as well as a larger mix of lower margin third-party products and traditional storage.

Management provided helpful insights around operational transitions the Company is working through:

“The business is managing two long-term transitions at once. We’ve talked about our migration to the more software intensive Alletra platform. This is reducing current period revenue growth though locking in future recurring revenue. Storage ARR growth of over 50 percent year-over-year offers early confidence in the migration. The second transition is from block storage to file storage driven by AI. While early, this is also on the right trajectory. Our new file offerings plus the sales force investment Antonio mentioned tripled our pipeline of file storage deals sequentially in Q2.”

The GreenLake as a service offering is expected to grow ARR at a 35-45% CAGR through FY’26.

HPE GreenLake, a leader in hybrid cloud infrastructure, is also attracting new customers to HPE’s portfolio with the number of customer organizations using GreenLake increasing 9% QoQ to 34,000 and ARR growing 39% YoY to over $1.5B. This increase is being driven by growth in AI systems and they expect high growth to continue to persist with a target 35-45% ARR CAGR from FY’22 to FY’26.

“We have strong momentum in HPE GreenLake. The number of customers that have adopted HPE GreenLake rose 9 percent sequentially. ARR grew 39 percent year-over-year to above $1.5 billion dollars in Q2. Storage and networking are typically the fastest growth elements of ARR and both retain robust growth rates. This quarter, AI was the fastest growth component of ARR. Our software and services mix rose approximately 200 basis points year-over-year to 67 percent. ARR is the best indicator of our model transformation to our as-a-service offerings. This growth validates what our customers are telling us – HPE GreenLake is a key differentiator. We expect HPE GreenLake’s value proposition to key customer, including enterprises and sovereigns, to sharpen with the advent of AI.”This quarter, AI was the fastest growth component of ARR. Our software and services mix rose approximately 200 basis points year-over-year to 67 percent. ARR is the best indicator of our model transformation to our as-a-service offerings. This growth validates what our customers are telling us – HPE GreenLake is a key differentiator. We expect HPE GreenLake’s value proposition to key customer, including enterprises and sovereigns, to sharpen with the advent of AI.”

Intelligent Edge revenue was $1.1 billion, down 19% from the prior-year period with 21.8% operating profit margin, compared to 24.7% in the prior-year period.

This was driven by difficult comps in both periods as HPE went through its backlog and also by soft macro conditions. However, HPE believes that the segment will return to modest sequential growth moving forward. The segment reported a 21.8% operating margin which was down 290 bps YoY due to lower revenues and the high margin switching business forming a lower percentage of revenue. Moving forward, HPE has already lowered its opex for the segment and they expect operating margins to return to the mid-20s by Q4’24.

Financial Services revenue was $867 million, up 1% from the prior-year period with 9.3% operating profit margin, compared to 8.9% from the prior-year period. Net portfolio assets of $13.2 billion, down 1.1% from the prior-year period.

Margins:

Margins contracted across the board with gross margin in Q2’24 the lowest since 2022. Gross margins were 33%, compared to 36% last quarter and in the prior year quarter. A shift from networking revenue to AI systems revenue is a headwind as AI servers are lower in margin (compared to the higher margin offerings i.e., Intelligent Edge). Management did revise forward guidance on adjusted gross margins downward based on the mix shift and expect to be below full year expectations of 35%.

GAAP gross margin of 33.0%, down 300 basis points year-over-year.
Q2 non-GAAP gross margin was 33.1%, down 310 basis points year-over-year.
GAAP operating margin of 5.9%, down 160 basis year-over-year and down 190 basis points sequentially. This is driven by gross margin compression as operating expenses decreased YoY.
Net Margin of 4.4%, down 160 basis year over year and down 130 basis points sequentially. The decreased net income margin is driven by gross margin compression and slightly offset by improvements in total operating expense year over year.

Cash:

HPE reported operating cash flow of $1.1 billion for an OCF Margin of 15.2%, up $204 million YoY.

Free cash flow of $610 million represents an 8.5% margin and an increase from a FCF margin of 4.1% last year due to prepayments for AI systems and the timing of working capital payments. HPE maintained guidance for at least $1.9B in FCF for FY’24, representing 6.4% FCF margins and noting that FCF is seasonally higher in the second half.

Management continues to target returning 65% to 75% of FCF to shareholders through repurchases and dividends. They repurchased $45M worth of shares in Q2’24 and noted that they expect a similar pace of repurchases going forward with an outstanding buyback authorization of ~$0.9B. The dividend yield as of July 10th, 2024 is 2.44%. The dividend yield has come down from >3% in 2019 as the amount of dividends paid has remained consistent at ~$619M since then despite an increase in stock price.

The cash conversion cycle was negative 4 days, which is a reduction of 28 days from Q2 2023. Inventory increased to $7.3 billion vs $4.6 billion in prior quarter. Management noted that “our days of inventory and days payable were both higher to support our expected growth in AI system revenue in the second half”.

Balance Sheet Discussion

HPE has a strong balance sheet despite having a seemingly high net debt balance of $8.6B as of Q2’24 relative to its current market cap of $27.8B as of July 10, 2024. This is because the Financial Services division is managed separately with Net Portfolio Assets of $13.2B which is the total amount of Financing Receivables and Operating Lease Assets, net of reserves against those assets. This is balanced against Gross Debt of $11.5B and Cash of $0.3B for the Financing Division. The actual Operating Company (i.e., HPE excluding Financial Services) has $2.5B of cash with no debt.

Overall, the debt balance has remained manageable with $11.3B of total debt as of Q2’24, down from $13.4B in Q2’23 and a peak of $19.5B in Q3’20. $7.5B of total debt is long-term, with the remainder being comprised of $3B of debt being current, $646M of commercial paper, and $121M of notes payable and lines of credit.

Earnings Q&A:

HPE’s IP Portfolio and Liquid Cooling:

On the call, an analyst asked how HPE separates itself given the large IP portfolio the company has with over 300 patents related to liquid cooling. Given we are Dell and Super Micro investors, this question was of importance to us. If we look at the sales we see today, these three companies are neck-and-neck in terms of total AI server $ revenue. However, HPE took the time to discuss how the company is differentiated, although we have yet to see that differentiation (or competitive edge) show up in revenue.

Question

Aaron Rakers – Wells Fargo:

Yes. Thanks for taking my question. I guess sticking on the AI topic, if I could first ask, when you referenced the AI enterprise customers starting to show up, and I think the comment on the conference call was…it's now north of 15% of your AI orders. Can you give a little bit more context of that? What has that been over the last couple of quarters? I'm just trying to think about the trajectory of that.

And then Antonio, on the liquid cooling side, as we and investors think about Blackwell product cycle from NVIDIA. I'm curious of…can you be a little bit more specific of exactly where, from a technology perspective, you differentiate at liquid cooling? Is there something unique that HPE does within the 300 patents that you would want to highlight for us, as sustainably differentiated. Thank you.

Answer:

Antonio Neri – CEO

As for the differentiation, HPE has three different ways to cool systems. So, one is the traditional way, which is called the liquid-to-air cooler, think about that, basically running water supply in chilled locations where basically cools the air around the systems. Everybody has done that for a long time. The second is what most of the industry is doing today, which is what I call 70% direct liquid cooling or hybrid liquid cooling. Those companies still use fans to cool aspects of the systems. Some of our competitors talk about direct liquid cooling, but that's exactly what they're doing, and they are doing only a hybrid direct liquid cooling. And HPE has…and by the way, in that environment, we have 10 systems already in market today that we are shipping and configuring for customers.

And then we have what I call 100% direct liquid cooling. And this is a unique differentiation HPE has because we have been doing 100% direct liquid cooling for a long time. And today, there are six systems in deployment, and three of them are for generative AI. And as we go to the next generation of the silicon and you talk about Blackwell, when you go to the B200, that will require 100% direct liquid cooling.

–End Quote

Here was another moment the CEO discussed the top 4 ways HPE is differentiated for AI:

“So Amit, on the differentiation, I will summarize this on four key elements. One is our ability to deliver and run systems at scale, so AI system scale, that's a unique expertise and we have decades of experience. Number two is our infrastructure cooling intellectual property. We actually have all the IP necessary to cool systems in three different ways for them other. Our manufacturing footprint, which is very unique. We have one of the largest water-cooled manufacturing footprints in the world with two very important locations in the US and in Europe, which are close to customers.

And then last but not least is services. What I think people are coming to realize that running the system of scale requires unique services capabilities. And that's why with Marie, we started showing you what the services pull-through is, which is also over time, a lever to improve the gross margin in this business. And we cover all aspects from day zero which consulting, to day one, which is advisory and professional services design and then — and build, and then day 2, which is a running part with our — in our operational services side and deep expertise when it comes down to this system of scale, including direct liquid cooling.”

Guidance Looks to Be Conservative:

An analyst called out that the guidance looks conservative. HPE could be setting up for a beat/raise in the second half of the year, which would be key to market enthusiasm and stock performance. Primarily, the key metrics including backlog support higher growth than management’s current guide. FY’24 revenue guide seems to assume no additional growth in the backlog, to traditional servers, or to intelligent edge, despite management noting their expectation for all three to experience sequential growth through year-end.

Management toned down expectations by noting that it takes longer than 6-12 weeks to install and that a decent percentage of AI system deals are in generative AI which are all GreenLake and therefore the services portion is deferred over the life of the contract. While this latter point may mean that the current backlog will take longer than 2 quarters to realize, expectations are still low enough to set for further large beat and raises.

Toni Sacconaghi:

[…] You talked about enthusiasm for the second half, but you beat revenues this quarter relative to your expectations by $400 million and by guiding up an additional percent, you're actually only guiding up the full year by $300 million. So, I'm wondering, are you just being conservative, given the commentary around enthusiasm and forces at work in the second half or how do we reconcile that discrepancy? And then also just on AI servers for the second half, I think you talked about six-week to 12-week lead times. So if you have $3 billion in backlog and lead times for six weeks to 12 weeks, why can't you deliver $3 billion in AI systems like next quarter or certainly in the second half? Thank you.

Marie Myers

[…] What I did point to though, Toni, is I pointed to the higher end of the range, so that's really what's giving us confidence based on the increase that we made on revenue. So you're seeing that higher top-line and then also the confidence I got around just the cost discipline […] So overall, Toni, keeping the guide at $1.85 to $1.95, but really pointing to the higher end of the guide in terms of just the confidence that you articulated. So I'll turn it over to Antonio to cover the second question.

Antonio Neri

Yes. Toni, I think there is an opportunity to potentially exceed that. I think the limiting factor is not the supply, to be honest with you, is the availability of data center space. I made this comment in Q1, if you recall, data center space and power and cooling. And so some — we are working with the customers to time everything correctly, 6 to 12 weeks, think about it, maybe less than a quarter, but then you have to go and install it.

And there is a nice percentage of our deals in generative AI, which are all actually GreenLake. And so while we can recognize the revenue upfront, we are deferring all the services piece of it. So it really is going to come down to the timing of the data center and the power and cooling. And if that all aligns correctly, then we may have an opportunity to do better. But we felt prudent at this point in time to keep it the way it is and raising by 1%.

–End Quote

With an AI backlog of $3.1B and lead times being reduced to 6-12 weeks, HPE can conservatively recognize their backlog over the next 1-2 quarters, resulting in at least $1.55B of AI systems revenue per quarter. This alone would meet consensus estimates for H2’24, and with management expecting sequential growth in traditional servers and intelligent edge as well as a backlog that is growing almost as fast as revenues, HPE is well-positioned to deliver a surprise or two over the next few quarters.

Valuation:

Despite top-line growth being positioned to accelerate from the low-single-digits over the last three years to the mid-single-digits and potentially higher going forward, HPE still trades in-line with its historical range on a NTM EV/EBITDA basis. With a NTM EV/EBITDA of just 6.6x, it trades below peers like DELL at 10.8x despite comparable growth (Dell is projected to grow in the high-single-digits going forward).

However, as previously mentioned, HPE’s enterprise value is overstated since it accounts for all of the liabilities of the Finance division and none of its assets. Backing out the book value of the finance division yields an EV of $23.3B as of July 11th, 2024, compared to EBIT of $297M in the division in the TTM. Backing out Financial Services EBIT yields non-GAAP operating profit of $2.6B for an EV/EBIT of 9x.

Comp Table

Assuming HPE meets consensus revenue expectations of $31.8B of revenue in FY’26 (3.5% CAGR) and maintains adjusted EBIT margins of 11%, it would generate $3.5B of EBIT. Assuming Financial Services forms the same percentage of EBIT as it does today and backing that out from total EBIT yields Operating Company EBIT of $3.1B. Assuming a similar pace of buybacks, HPE can reduce its share count by 0.8% annually, along with a 2.3% dividend yield. Assuming no multiple expansion, this would yield a high-single-digit IRR.

If we assume that HPE re-rates to 13.3x EV/EBIT by 2026, a valuation in-line with CSCO which is projected to see flat growth through 2026 and is the second cheapest in the group, then HPE would generate a low-20s IRR. This scenario could become more likely as AI systems revenue continues to accelerate and the market starts to recognize HPE as a direct liquid cooled server company. This scenario is still conservative as it doesn’t take into account the likely possibility that HPE will beat consensus estimates.

Conclusion:

HPE will test mental flexibility as it’s a sleeper stock; viewed as an outdated tech company from the dot-com era. Yet, 1990s hardware players are auspiciously positioned to capture AI server revenue, and to also offer software platforms for hybrid AI architectures.

There will be many investors too set in their ways to consider the possibility that we are in a new era. Those who see HPE or Dell doing well will cry “dot-com bust.” AI servers will drive the revenue for these companies in the near-term, yet keep an eye on the AI software segments (GreenLake, etc.) for dot-com-defying longevity.

Our portfolio’s motto is the best AI hardware players will make the best AI software players, which gives a strong nod to how we plan to secure future gains as opposed to resting on our laurels with the current hardware-driven AI cycle. HPE is certainly a candidate that fits this motto, yet it’s too early to tell if HPE has what it takes to compete in what is shaping up to be a highly competitive space.

This analysis is a preview of what you can expect in our upcoming Discovery tier, which will provide additional analysis on new idea generation stocks that are not currently in the I/O Fund portfolio. We look forward to launching this tier late August/early September. There will be no changes to our current service tiers, rather I/O Fund Discovery is a service for those who want more new stock ideas beyond what our service currently provides. Stay tuned for more information!upcoming Discovery tier, which will provide additional analysis on new idea generation stocks that are not currently in the I/O Fund portfolio. We look forward to launching this tier late August/early September. There will be no changes to our current service tiers, rather I/O Fund Discovery is a service for those who want more new stock ideas beyond what our service currently provides. Stay tuned for more information!

Richard Chu, Equity Analyst for the I/O Fund, contributed to this analysis.

Recommended Reading:

Category: Enterprise

Hewlett-Packard Enterprise: Sleeper Stock with AI Potential