Snowflake Dives Into the Generative AI Data Pool

Nearly 90 years ago, the good old U.S. government enacted the Securities Exchange Act (SEA) of 1934 after a little market meltdown circa 1929. Sure, the Great Depression sucked, but look at all of the great hiking trails we got out of it. Anyway, one of the new rules to ensure greater transparency and less fraud by companies was the requirement of regular financial statements, including quarterly reports. Some people love them, others hate them. The latter camp believes that focusing on three-month business cycles encourages a short-term obsession with numbers while undermining long-term strategy. Yup, that sounds like ‘Merica.

We usually check in with stocks in the Nanalyze Disruptive Tech Portfolio once per year because we believe that offers enough resolution into a company’s business without needing to get caught up in the hype cycles that quarterly earnings reports live and die on. On the other hand, hype cycles like generative AI, which has consumed the tech landscape this year, can create enough gravitational pull to change the semi-immutable laws of investing. Or at least make us pay closer attention to the short-term rhythms. That’s certainly the case with big data powerhouse Snowflake (SNOW), a company with a data storage solution that’s being used by marquee names across all industries.

Let’s quickly check in with the financials before covering some of the major moves Snowflake is taking to stay ahead of the wholesale adoption of generative AI by all of Big Tech.

About Snowflake Stock

Snowflake has developed a robust cloud platform where customers can slice and dice their stored data for numerous business applications. Almost all of the company’s revenue comes from consumption-based fees it charges to customers based on the compute, storage, and data transfer resources they consume rather than a traditional fixed-cost subscription that you’ll see with most software-as-a–service (SaaS) models. Regardless, the paradigm works: Snowflake has maintained dynamo product revenue growth since its IPO in 2020 and claims it will hit $10 billion by 2029.

Triple-digit revenue growth usually isn’t sustainable over the long term, and we see there is a significant drop predicted for this year – “just” 36% year-over-year growth in 2024. (Unfortunately, the Security Exchange Act of 1934 does not regulate a standardized fiscal year, so Snowflake tracks its finances differently.) There is also a downward trend in net retention rate, which measures how much money existing customers are spending on the platform. Again, a 151% dollar-based net retention rate would be enviable in most circumstances.

The slowdown is real and management said in its Q1-2024 earnings call that lower platform consumption will probably persist for the short term. One big reason is that some of the company’s biggest customers are scrutinizing costs associated with using the Snowflake platform. For example, some organizations have “reevaluated their data retention policies to delete stale and less valuable data. This lowers their storage bill and reduces compute cost.” Snowflake management puts a positive spin on it naturally: “History has shown that price performance benefits long-term consumption.” Or maybe the company’s marketing and sales staff did a good job of upselling customers on all the bells and whistles, and a tightening of the purse strings is having a ripple effect on Snowflake’s bottom line. In response, management is promising to rein in costs. (More on that later.)

A Blizzard of Generative AI Actions

Remember all of that talk about drowning out the noise of quarterly reports? Well, here’s an example of how hard a company can pivot in just three months. In the Q4-2023 earnings Q&A wrap-up, an analyst asked about the opportunity that Snowflake anticipated from generative AI and large language models (LLMs). CEO Frank Slootman gave the we’re-still-in-the-evaluation-stage answer, sounding almost a bit dismissive or at least nonchalant:

One of the challenges with these new technologies [is] that people come up with a lot of interesting questions, but without a solid business model, that’s not going to take off. So we take a very pragmatic view. We do anticipate that Snowflake data will be a very, very big driver of a large language model in conjunction with many, many other data sources. So we think that the gravity around data will drive a lot of this action activity to our platform.

CEO Frank Slootman

Fast forward to Q1-2024 and Slootman pretty much leads off his remarks on generative AI and spends half of his intro talking up the company’s machine-learning bonafides. He even throws in this nugget near the top of the fold:

Data science, machine learning and AI use cases on Snowflake are growing every day. In Q1, more than 1,500 customers leveraged Snowflake for one of these workloads, up 91% year-over-year.

CEO Frank Slootman

Generative AI craves huge amounts of data, and it sounds like Snowflake’s platform is designed to keep those algorithms well fed (data transfer) and safely caged (storage) so that enterprises can build their own ChatGPT-like models (compute) using their own massive amounts of proprietary or specialized data. 

Toward that end, Snowflake acquired Neeva in May for an undisclosed sum. The Silicon Valley startup had raised $77.5 million from high-rolling venture capital firms like Sequoia Capital (also an investor in Snowflake) and Greylock to develop a generative AI search engine. The emergence of ChatGPT et. al. pretty much squashed plans to compete in the general search engine market, so Neeva and Snowflake will leverage the tech for enterprise customers. Less than a year ago, Snowflake also acquired Applica, a Polish startup that had developed automation technology that could turn documents such as invoices or legal contracts into structured data for AI algorithms to digest. 

In addition, just a few days ago, Snowflake and the trillion-dollar AI chipmaker NVIDIA (NVDA) announced a partnership to leverage the power of the latter’s hardware to help enterprise customers make custom large language models (LLMs) for advanced generative AI services, including chatbots, search, and summarization.

Hype or Substance?

Every company out there (it seems) is latching onto generative AI in hopes of attracting the attention of investors. Companies like C3, Palantir, BigBear.ai, and SoundHound are enjoying considerable hype at the moment, something that we don’t view as a positive. Hype is unnecessary noise for long-term investors, though management teams that recognize it for what it is will often take advantage of the situation to raise more money or take some chips off the table while the going is good.

Snowflake’s sudden conversion to the religion of generative AI should be taken with a grain of salt. The thesis – he who stores all the big data is in the best position to sell generative AI solutions – makes sense, but we don’t invest in stories. There’s enough to like about Snowflake even if their future generative AI potential is dismissed, and their latest investor deck shows a company that’s handled the recent macroeconomic headwinds just fine. But is it a company worth investing in based on current valuations?

Should You Buy More Shares of Snowflake Stock?

Needless to say, Slootman was singing a very different tune about the company’s interest in generative AI in Q1-2024 than he was in Q4-2023. Was he just being coy, waiting to drop the mic on the opening act of the 2024 fiscal year? Or was he just too busy spending all the money he gets paid to be CEO: Snowflake had more than $890 million in stock-based compensation (SBC) in 2023 alone, representing more than 40% of total revenues (average stock-based compensation for the tech industry was around 22.5% in 2021). Ultimately, shareholders need to be concerned about dilution, and the company assures us that the 3% increase in outstanding shares last year will move to 2% next year and going forward.

A side effect of all that SBC is the amount of cash burn being obfuscated from investors. Snowflake’s operating loss of $842 million in Fiscal 2023 all but evaporates when you back out the $889 million of SBC.

Snowflake has also managed to achieve positive operating cash flows – over $545 million in 2023 – which mean they don’t have to worry about runway, though their plan to buy back $2 billion in stock over the next two years seems odd. Isn’t there a better way to show investors an ROI on the $5 billion in cash and investments they have laying around?

Snowflake seems to be still operating on the growth-at-all-costs philosophy that was still in effect in 2020 when the company went public. With a simple valuation ratio (market cap of $57.5 billion/$2.36 in annualized product revenue) of nearly 24 – where anything over 20 is considered overpriced – we won’t be adding to our position in Snowflake at current prices (see last year’s piece on What’s a Fair Valuation for Snowflake Stock?). Tech stocks are volatile, so setting a valuation target (reasonable or aggressive) and waiting will usually turn out just fine. Don’t chase momentum. No FOMO.

Conclusion

Snowflake remains a high-growth company in a high-growth industry, with an estimated total addressable market (TAM) of $248 billion (based on a pile of Gartner forecast reports, so at least they’re not relying on content factories in Mumbai). If Snowflake can reach its 2029 goal of $10 billion in revenue, it will have penetrated just 4% of its TAM, so the long-term outlook seems good. The company appears to be positioning (and rebranding) itself as the cloud-based tool for enterprise applications using generative AI, so it won’t have a Blockbuster moment.