Meg Whitman controversially directed Hewlett Packard Enterprise (HPE) to spin out of Hewlett-Packard in 2015. At the time, this was one of the largest spin outs in American corporate history. HPE was organized into two divisions playing to HP's historic strengths, enterprise software and server hosting. In the years since, the company has undergone another series of spin-merge transactions, retreating from the increasingly commoditized hosting and hosting services space. During a tense recent Q2 earnings call, HPE's new CEO claimed the second half of the year would be 'more challenging'. Following the call, HPE's stock dropped 11%.
Andrew Macklis published a bullish long on SumZero following this price correction. We sat down to discuss his thoughts on the company. Andrew is currently an MBA candidate at Harvard Business School, and will be working at Capital Group's San Francisco office this summer. Prior to HBS, Andrew worked at Golden Gate Capital, McKinsey, and Blackstone.
Luke Schiefelbein, SumZero: What about HPE caught your eye as a value investor? What catalyzed your entrance into the position?
Andrew Macklis, Capital Group: I am a big proponent of the Joel Greenblatt school of thought that investigating complex situations is one of the best ways to uncover mispriced assets. In the case of HPE, two large spin-merge transactions (the Enterprise Services business in April 2017 followed by the Software business in September 2017) combined with a number of tuck-in acquisitions, significant operational restructuring, and changes to the reported segments created a messy situation that caught my interest. This complexity requires considerable data scrubbing to get a normalized view of what is going on with the fundamentals of the business, making it harder for quantitative screens to pick up on the mispricing. I started to dig in last summer amidst this backdrop, and as I spent time learning about the business it became clear there were some pockets of real strength in the portfolio that were being obscured by all the moving pieces. Based on my analysis of the company, I identified several instances over the past year when I believed the stock was highly undervalued, causing me to build a position. Most recently, I re-established a position following the sharp decline in stock price after the announcement of 2Q ’18 earnings.
Schiefelbein: HPE recently dropped 11% after reporting its Q2 earnings. What happened? Why does HPE’s CEO see a ‘more challenging second half’ for 2018?
Macklis: I believe this is a case where, to use Ben Graham’s language, the market is acting as a voting machine in the short run, but will eventually return to being a weighing machine. Much of the investor fear coming out of the Q2 earnings release was driven by a few snippets of cautious commentary made by HPE’s new CEO Antonio Neri along with a decline in legacy server unit volumes that I believe are being misinterpreted. Headlines on various online forums highlight Neri’s remarks that he “expects the growth rate to moderate” and sees “a more challenging second half” for 2018. While these comments in isolation certainly paint a daunting picture of the next several quarters, the key to understanding the opportunity in this case is to get beneath those soundbites and dig into the full transcript. As Neri makes clear if you listen to the whole discussion, the reason he indicated growth will slow in 2H ’18 is not really due to deteriorating business fundamentals, but instead is largely an artifact of currency impacts and lapping acquisitions. As for the declining hardware unit sales in Q2, these declines were primarily the result of an intentional choice by management to pivot from low/no margin commoditized server sales into higher margin next generation products. The fact that management raised fiscal 2018 full-year earnings guidance again (after already raising guidance on the Q1 ’18 call) offers a much more direct read on their confidence level in the business going into the second half than any soundbite.
Schiefelbein: What key metrics should investors be paying attention to as your thesis matures? What developments in the next 12 months could be most damaging to HPE?
Macklis: Despite all the noise around the company, there are really four key “needle movers” in this business that drive the financial model. The first is normalized growth rates (adjusted for currency and acquisitions) on HPE’s next generation products like hyperconverged, all-flash-arrays, high performance compute, and WLAN. This metric is critical since the business is mix-shifting toward those products, making them the best indication of HPE’s longer-term trajectory. Second, the cost savings achieved in the “HPE Next” program, a clean sheet cost-out initiative launched last year in the aftermath of the spinoffs that has the potential to reduce costs by hundreds of millions of dollars (and my analysis indicates HPE is likely to overdeliver on this front). Third is DRAM prices - since DRAM is such a material portion of the server business’ cost of goods, each 10% change in average DRAM prices equates to a roughly $250M delta in annual cost to HPE. In the near term, this is the most likely metric to hurt HPE since DRAM prices have been rising since January 2017 when prices spiked roughly 50%. Fourth is the trend in customer orders for the technology services business (which is largely break-fix maintenance work) since that is a hidden “crown jewel” that delivers high margin, recurring revenue and generates ~65%+ of HPE’s operating profit.
Schiefelbein: Where are HPE’s biggest growth markets in the next 2-5 years?
Macklis: HPE is deliberately de-emphasizing zero margin custom-designed commodity server revenue and realigning itself (via organic investments and M&A) toward the types of next generation products mentioned earlier. These new era categories are higher growth, carry higher margins, pull through more technology services, and will serve as growth engines of HPE going forward. In servers for example, HPE’s high performance compute offering gives them access to an $11B market that is growing 6-8% per year. In storage, HPE’s all-flash array products are part of a $6B+ market growing 10%+ per year. Similarly, HPE’s hyperconverged offering allows it to tap into a $4B+ market growing 35%+ per year (with HPE’s product in this space growing triple digits in recent quarters). In networking, HPE plays in the WLAN space - a $7B market growing 8% per year. Together, these components are changing the profile of HPE and imply that simply extrapolating historical revenue growth trends would miss an upcoming inflection point in the business.
Schiefelbein: What caused HPE to spin out so much of its traditional software business?
Macklis: Prior to 2015, Hewlett Packard was a company that housed a set of loosely-related product categories each with disparate capital allocation strategies, cost structures, and go-to-market messages. The competing priorities that resulted from having all these businesses together meant that no line of business was able to operate with the speed or focus necessary to be a true leader in its respective market. I think Meg Whitman deserves considerable credit for having the vision and courage to “shrink to grow”, and last fall’s spin-merge of the software business with Micro Focus was the final step in that evolution. Though the spinoffs certainly created short-term operational distractions, management’s thesis that more focused, nimble companies are able to better execute in their respective markets appears to be panning out.
Schiefelbein: How durable is HPE’s enterprise server moat? Is the increasing commoditization of servers a threat?
Macklis: The industry standard server products have been commoditized for some time, so HPE’s differentiation in that space hinges on its distribution breadth and aftermarket service expertise (HPE is known to literally helicopter an expert over to customer data centers to fix an issue in real time if necessary). While these advantages have afforded HPE a consistent leadership position in the global server market, they do not constitute a durable moat. Importantly though, the new era growth products discussed earlier that HPE is shifting toward like high performance compute and hyperconverged are meaningfully more differentiated and sticky (as evidenced by their premium pricing and lower customer churn) indicating that the business will develop stronger moat characteristics over time. Additionally, since competing in the servers space requires massive scale that few players can match, the threat of new entrants is limited, leading to a consolidated competitive set. Finally, it is important to calibrate one’s expectations of business quality based on the price paid. After the recent price decline, HPE is trading at roughly a 10x FY ’18 ex-cash P/E (a considerable discount to peers like Cisco and NetApp which trade at ~15-20x) granting a margin of safety even if you believe competitive pressures may escalate.
Schiefelbein: What are the biggest risks to your position? What is the most compelling bear case for the stock?
Macklis: I’ll focus on what I see as the three most salient risk areas and explain how I ultimately became comfortable with each when I initiated my position in the stock. First, an acceleration in the migration of workloads to public cloud setups would magnify declines in HPE’s legacy products, which are heavily indexed to on-premise environments. Since guidance and consensus estimates already bake in legacy product declines resulting from a public cloud transition, the critical element for the bear case here would be for those declines to become pronounced enough to swamp growth from the next generation product areas discussed earlier. This scenario becomes less and less likely as the business mix-shifts away from legacy categories.
Second, as free cash flow conversion improves, there is a risk that HPE squanders the cash either on a blowout deal like the 2011 Autonomy acquisition, or on low-yield internal investments. While recently appointed CEO Antonio Neri’s capital allocation strategy remains to be tested, his commentary and actions thus far have been consistent with those of Whitman: a returns-based approach that prioritizes capital return (in the form of both dividends and aggressive share buybacks) and only considers M&A in select cases that uniquely leverage HPE’s distribution breadth and provide truly complementary technology.
Third, competitive pressure should be closely monitored going forward, particularly in the more commoditized server and storage products. Though it is possible that pricing pressure from players like Dell, Huawei, or unbranded “white box” servers emerges, this does not seem likely in the near-term given that recent strength in enterprise IT budgets is currently supporting price increases.
Schiefelbein: How vulnerable is HPE to the continued growth of CSCO and ORCL in the enterprise business services space?
Macklis: Prior to the spinoffs, I would have said that HPE was relatively vulnerable to attempts by competitors to steal share. The old HPE was clunky, faced channel conflicts between its business units, and tried to be a one-stop shop. The new slimmed down HPE has instead chosen a few core markets and put together an ecosystem of “best of breed” partnerships with companies like Arista, Mesosphere, Hedvig, and Cohesity that position the business for success. While players like Cisco are certainly formidable and they will no doubt continue to compete and innovate, customer discussions I have had indicate that the HPE value proposition remains highly compelling on both an absolute and relative basis.
Schiefelbein: How does the surge in GPU and DRAM demand from crypto miners impact your thesis?
Macklis: In the DRAM discussion, it is important to delineate between the short run and long run. In the short run, the supply of DRAM is largely fixed (since building new capacity from scratch can take years, and existing facilities are so capital intensive that manufacturers have a strong incentive to constantly run their production lines at full capacity to leverage fixed costs). This reality means that changes in demand dictate DRAM price fluctuations in the short term. The rise of cryptocurrency mining (in addition to continued growth in smartphone sales) has thus absolutely fueled the recent spike in DRAM prices by creating a surge in demand. It is worth noting, however, that recent weakness in cryptocurrency prices may dampen mining activity, since an analysis published by Morgan Stanley estimates that even large Bitcoin mining pools breakeven at $8,600 per coin.
Even if these robust demand dynamics continue, my research has led me to believe that over the medium to long term, it is likely that DRAM prices will normalize due to supply-side changes. The long-term trend for DRAM prices is clear: a pattern of consistent (despite short-term cycles) declines since the early 1990s as technology improves and memory costs continue to fall. Moreover, there’s a potential nearer-term catalyst: the adoption of 3D NAND. 3D NAND is a new type of memory that is both lower cost and higher capacity than older “planar NAND”. This superior form of memory is thus expected to displace Planar NAND capacity, leaving memory vendors with the choice of either decommissioning Planar NAND production lines (since they will be obsolete and no longer cost competitive) or converting them into something else. I suspect DRAM would be a prime candidate to convert into given the supply tightness in that market. This conversion process takes 6-9 months, so it is a much faster way for manufacturers to capitalize on high DRAM prices than if they built capacity from scratch. The corresponding increase in DRAM capacity and decrease in DRAM costs would create meaningful value for HPE, since a return to Oct. ’16 DRAM prices would save HPE ~$550M+ even assuming the company has to give back 50% of the cost savings in the form of pass-through commodity price concessions.
Schiefelbein: Where else do you see value in the market today? Where else do you focus your research?
Macklis: While valuations are elevated on average, the recent uptick in volatility combined with skittishness in investor sentiment related to fears about being near the top of the cycle has created select opportunities to bargain hunt. I am focused on two main methods for idea sourcing right now – looking at special situations (such as the spinoffs in the case of HPE) to find value investments, and building a library of “best-in-class” long-term compounders that I can have on my wish list to pick up if they ever temporarily fall out of favor. One example of the latter is Advanced Disposal Services Inc., which I established a position in and published my thesis on a few months ago in SumZero.