An early stage Apple
I love the idea of speculating on early stage trends. The earlier you invest, the more likely it is that the valuation isn’t out of whack yet. And the more likely it is that others will buy into the story once you’ve taken the position.
You just need to make sure the company’s market position is so strong that it will be able to capture the tailwinds you are forecasting. I believe Pure Storage may be an example of such a company: a market leader in a fast-growing industry.
Pure Storage (“Pure”) produces all flash array (“AFA”) memory storage systems for data centers. The company was formed by John Colgrove and John Hayes in 2009 and later hired Scott Dietzen as a CEO. It pioneered the selling memory storage systems built on flash technology that could rival HDDs through innovate software technology. Pure’s storage systems cost anything from $25k for a tier 3 system to above $100k for a tier 1 system. A Pure product combined commodity SSDs sourced from a third party as well as their own software that helps improve effective capacity and latency.
The fact that hard disk drives (“HDDs”) are mechanical and flash memory are built on semiconductor surfaces has profound implications. The higher speed of flash is an obvious advantage. Another advantage is Moore’s Law: NAND prices continue to drop faster than HDD prices and at a certain inflexion point, more or less certain 100% of storage systems will be based on flash. Higher speed also enables advanced data reduction techniques such as compression and deduplication, making software a key differentiator between product offerings and competitors.
Another key trend that affects product competitiveness is rise of SaaS and move of data to the cloud. The move to the cloud increases demands on performance for storage devices.
The comparative advantages of flash illustrate why all flash arrays will continue to take market share:
- Price per data request: HDD systems are 5x more expensive than AFA’s on input-output per second (“IOPS”) basis, roughly $1.5/IOPS vs $7.5/IOPS. Flash wins in terms of speed.
- Price per GB: AFA systems are roughly 5x more expensive on a $/GB basis. Thanks to data reduction techniques, flash raw storage capacities can be multiplied by 2-30x, with the actual number depending on the type of data stored. For example: Virtual desktops have a higher proportion of duplicates in the data, making potential data reduction ratios very high. On the other hand, databases offer lower potential for data reduction. The cost measured in $/effective GB after data reduction techniques may already be lower than an equivalent HDD offering, it all depends on the application. Over time, as Moore’s law continues to push NAND prices lower, more customers will find it worthwhile to go over to flash regardless of actual use.
- Failure rates: High end HDD and SSD have about the same mean time between failure (“MTBF”), at about 1.6mn hours for HDD and 1.2mn hours for SSD.
- Space: AFA’s are 3-175x denser than a high-end HDD system on a capacity basis.
- Power efficiency: HDDs use 3-20x as much power as an SSD. Newer products are pushing up this ratio to above 25x.
- HVAC outlays: Flash SSD storage systems have cooling requirements of about 30% less than equivalent HDD.
- Endurance: The number of read/writes for NAND flash is limited, whereas HDDs have practically no limit. However most HDDs die within 5 years, according to research by Google.
- Optimization: given higher IOPS for flash, software is able to tailor storage products for particular applications.
Flash has a technological advantage. As NAND prices continue to fall, there will come a point when even the applications most suitable for HDD will be cheaper to run on AFAs.
The market for performance-intensive storage is $24 billion, stable with essentially no growth in dollar-terms. Pure believes that this market will move to flash storage over the next refresh cycle or two (7-8 years, i.e. by 2024). The below chart from researcher David Floyer echoes Pure’s forecast, predicting that the storage market will be almost all flash by 2025:
The all-flash array market grew 120% in 2014 to $1.3bn in revenue. Gartner expects the market to reach $7bn in 2019, an annual growth rate of 40%.
Pure’s market position
The storage industry is highly competitive with incumbents competing against start-ups such as Pure Storage and Nimble Storage. Even though the products appear to be similar – flash storage systems built on SSDs – they are not. Pure claims to have a 2-year technological lead on competitors, which may be true. Not every customer needs the advanced features that Pure is able to offer however.
Why I believe Pure Storage will win:
- Pure achieves a 5.5:1 data reduction on average, almost double that of its closest competitor EMC. As NAND prices continue to fall, this will become an increasingly important differentiator against competitors. The technological lead has made it simpler to hire top engineers.
- Pure Storage’s latency is now “sub-1ms”, compared to EMC’s claim of 0.5ms. Users report that the latency of Pure’s products rarely spike above 1ms. When it comes to worst-case latency, Pure Storage claims to be market leading.
- Simplicity: Pure Storage’s Chief Architect John Hayes emphasizes ease of use as a key differentiator. Not easily quantifiable, it may matter a lot for customer satisfaction. A random search on Pure Storage reviews on Reddit give you the following feedback:
“Most Pure customers I’ve come into contact with like how stupidly simple it is”
“Pure is awesome. Love the long term TCO per iops on Pure. Totally simplified my life.”
“I couldn’t be happier. Performance boost was obvious but the customer support and ease of use are the biggest advantages”
“Pure’s the easiest thing ever to work with, their interface is ridiculously simple”
- The company’s strategy is similar to that of Amazon: grabbing market share while postponing the eventual need for profits. Given that Pure is a first-mover in some respects and software R&D is a fixed cost offering substantial economies of scale, such a strategy appears to be rational. Operating leverage will reduce losses over time.
- Recurring revenue: its Evergreen Storage solution enables customers to upgrade components non-disruptively without having to take the array out of production. They get free controllers every 3 years together with free software updates. To upgrade their systems, all that customers need to do is adding flash shelves. Around 80% of Pure Storage’s customers are on 3-year maintenance contracts. Pure Storage claims that every $1 spent on its storage products, leads to another $8 spent within 18 months.
- Pure Storage is the only company in the industry that offers a cloud-based support system, called Pure1. Pure collects data about its customers’ devices in real time, which enables Pure’s support team to reach out to customers ahead of failures or other errors.
- Pure has no debt and $600m of cash in the bank. The company became cash flow positive in the last quarter. This helps alleviate customer concerns that the company will not be around, a key reason why Violin Memory is failing.
- Clayton Christensen’s innovators dilemma: it’s hard to disrupt your own 20-year old technology, while looking after your existing business. Technological innovation tends to lead to shake-ups in industry structures.
It is not just marketing slogans. Third-party reviewers and customers all give positive feedback on Pure Storage’s products. In their 2015 “Magic Quadrant report”, Ganrter concluded that EMC and Pure Storage are leading the AFA industry. Pure’s product is said to have a technological lead over EMC, but EMC’s sales channel is more effective. Simply push XtremIO products to your existing customers without the need for Proof of Concept (“POC”) tests.
Management is excellent, both CEO Scott Dietzen, CTO John “Coz” Colgrove and Chief Architect John Hayes. They are nerds, not promoters.
Pure Storage company culture resembles a cult. Employees call themselves “puritans”, indicating their strong affiliations with the company. Marketing employees wear orange when meeting with customers. Glassdoor reviews rate the company at 4.5/5.0, with 100% of reviewers approving of CEO Scott Dietzen. As benchmarks, Nimble Storage has a score of 3.7/5.0, EMC 3.8/5.0 and Violin Memory 2.4/5.0.
Industry ranking provider Satmetrix gave Pure Storage a Net Promoter Score of 79, among the highest in the world for technology brands. Apple iPhone has a score of 63.
In 2015 Pure Storage was awarded the best place to work by The San Francisco Business Times, in the large company category.
Other benchmarks: Google search queries for “Pure storage” keeps on increasing, up 42% YoY in March 2016. Alexa website rank is improving slowly as well.
The stock price
Pure Storage has a long-term operating margin guidance of 15-20%. EMC’s historical operating margins has been in the high teens. Assuming a 15-20% operating margin and a 10x EV/EBIT steady-state multiple, Pure Storage should trade at an EV/S of 1.5-2.0x. If Pure Storage’s prediction that 50% of the high-performance storage market will be flash-based by 2024, and Pure Storage maintains an AFA market share of close to 20%, Pure Storage will have revenues of $4.8 billion and an EV of $7-2-9.6 billion, compared to $2.26 billion today – a CAGR of 21%. This may be conservative: EMC was acquired by Dell for $67 billion.
We have learnt to ride waves of technological change with best-in-class companies, and a run-rate EV/Sales multiple of 3.8x is by no means an excessive multiple to pay for a true disruptor. The idea may be better suited for family offices with longer-term investment horizons.
Why the stock may be mispriced
Investors seem to be in disbelief that an independent company can survive in the cutthroat flash storage industry. Recency bias or “guilt by association” may cause investors to put Pure Storage in the same category as Violin Memory and Nimble Storage, despite Pure’s competitive advantages. Comparing individual metrics such as latency or IOPS typically miss the point that Pure has a much more differentiated offering.
Major hedge fund Tiger Global has suffered from redemptions, leading to a sell down of its stake in Pure Storage.
Shorting pressure has been high, apparently in a bet that Nimble Storage’s new AFA offering will take market share from Pure Storage. Such a view is misguided, for several reasons: Nimble’s software is not written from the ground up and as a newcomer to the AFA market, it is hard to catch up. Pure Storage has also released new products recently, including FlashBlade, a major product release that offers 130 PB in a rack at only $1/GB.
Some insider buying by a director and major venture capital firm Greylock at around $17/share. No insider selling since the IPO.
The major risk with the stock is that the base rate probability of success for storage start-ups is low. No-one has grown to substantial scale since NetApp in the early 1990s. Of the seven major storage IPOs in the last decade, four have been acquired at a decent premium to the IPO price. The others: Nimble, Violin and Box were not acquired and their share prices slumped post-IPO. I am betting that Pure Storage will be an exception to the rule.