Had You Bought This Artificial Intelligence (AI) Stock When Warren Buffett Sold It, You'd be Sitting on a 47% Return Now
Berkshire sold its entire position in Snowflake last year for a minimal gain over its estimated purchase price.
Snowflake stock has soared by 47% since Berkshire sold it, but that doesn't mean Buffett and his team made the wrong move.
10 stocks we like better than Snowflake ›
Warren Buffett has served as the CEO of Berkshire Hathaway (NYSE: BRK.A)(NYSE: BRK.B) since 1965. He'll step down from the role at the end of 2025, capping off an incredible run during which he delivered consistent, market-crushing returns for its shareholders.
Buffett prefers to invest in companies with steady growth, reliable earnings, and strong management teams that favor shareholder-friendly policies like paying dividends and repurchasing their shares. That's why it was surprising to many when Berkshire bought Snowflake (NYSE: SNOW) stock in 2020 ahead of its initial public offering (IPO) -- the cloud data warehouse giant hardly ticked any of Buffett's usual boxes.
That means one of his lieutenants probably was the key decision-maker behind the purchase. However, Berkshire Hathaway sold its entire position in Snowflake in 2024's second quarter, and the exit price suggests it booked hardly any gain on the investment. Had you bought shares of Snowflake at the moment the conglomerate sold, you'd be sitting on a 47% return right now -- but I still think Buffett and his team made the right decision.
Snowflake's flagship product is its data cloud, which helps businesses unify their valuable information from across multiple cloud platforms such as Amazon Web Services and Microsoft Azure. This allows them to perform detailed analyses on a complete data set, which produces more useful insights than analyzing fragmented data sets stored on different cloud platforms.
Developing the best artificial intelligence (AI) software requires easy access to massive volumes of data, so Snowflake is perfectly positioned to help its customers deploy this game-changing technology. It launched a new platform in late 2023 called Cortex AI that provides businesses with access to large language models (LLMs) from leading third-party developers like Meta Platforms and Anthropic.
Businesses can plug their internal data into those LLMs to rapidly create AI software for a range of purposes. But Snowflake continues to empower Cortex with new capabilities. It offers a tool called Document AI which extracts valuable data from unstructured sources like contracts and invoices, which businesses can plug into their AI models. Then there is the Cortex Agents tool, which allows businesses to create custom virtual assistants to complete specific tasks.
For example, a business could train an agent to autonomously create transcripts of conversations between salespeople and customers, and then train another agent to analyze those transcripts for opportunities to generate more revenue. In the past, those tasks would take human workers hours or even days to complete, which is why agentic AI is becoming so popular.
Snowflake had 11,578 total customers at the end of its fiscal 2026 first quarter (which ended April 30), and about 5,200 of them were using its AI products and services on a weekly basis. That suggests rapid uptake considering the Cortex AI platform is barely 18 months old.
Snowflake's product revenue grew 26% year over year in its fiscal 2026 first quarter to $996.8 million. That was its slowest rate of increase on that metric since the company went public in 2020. In fact, Snowflake's revenue growth is consistently trending lower with each passing quarter.
Slowing revenue growth isn't necessarily a bad thing if a company is spending less money on things like marketing to improve its bottom line. But that isn't the case with Snowflake, because it continues to ramp up its spending. The company's operating costs grew by 26% during the first quarter, which resulted in a blowout net loss of $430.1 million. That was a whopping 35% larger than its net loss from the year-ago period.
Snowflake's net loss would have remained stable or even declined if its increased spending resulted in accelerating revenue growth rather than decelerating growth. However, there is no sign of that trend reversing, because the company is forecasting even slower revenue growth of 25% in its fiscal 2026 second quarter (which ends July 31).
Even Snowflake's remaining performance obligations (RPOs) are flashing signs of weakness. Despite growing by 34% year over year to $6.6 billion during the first quarter, they actually shrank from their peak of $6.8 billion from the previous quarter three months earlier. RPOs are essentially an order backlog that can be expected to eventually convert into booked revenue, so any dip in that figure could be a signal of future weakness for the company's top line.
Snowflake went public at a price of $120 per share, and Berkshire Hathaway bought the stock right before it listed, so its entry price was probably around that level. Berkshire sold its position during the second quarter of 2024, when the stock was hovering at around $135, which translated into a gain of just 12.5% over four years. A risk-free asset like U.S. Treasuries might have yielded a better return.
But Snowflake stock has soared by 47% to around $200 since Berkshire sold. Does that mean Buffett and his team made a mistake? I don't think so, because I would argue it's now overvalued. The stock is trading at a price-to-sales (P/S) ratio of 17.3, which is significantly more pricey than cloud and AI leaders like Amazon, Microsoft, and Alphabet:
Snowflake ticks at least two of Buffett's boxes, with a caveat on both counts: The company is growing its revenue quite fast, but that won't last long if its current rate of deceleration continues. The company is also returning money to shareholders through a stock buyback program, but that can't last forever if it keeps losing money.
On the flip side, Snowflake's valuation and its steep losses would normally have excluded it as an investment option for Berkshire entirely, so I'm not surprised the conglomerate eventually exited its position. Besides, that stake was worth just $800 million at the time of sale, so it represented a tiny sliver of Berkshire's $276 billion portfolio.
In summary, it doesn't matter if you're Buffett or an average retail investor -- some stocks will rise after you sell them. That's part of the investing journey, and we have to accept that timing the market reliably is impossible. It's better to focus on the fundamentals of the business in question, and on that basis, I think Berkshire made the right call.
Before you buy stock in Snowflake, consider this:
The Motley Fool Stock Advisor analyst team just identified what they believe are the for investors to buy now… and Snowflake wasn't one of them. The 10 stocks that made the cut could produce monster returns in the coming years.
Consider when Netflix made this list on December 17, 2004... if you invested $1,000 at the time of our recommendation, you'd have $653,389!* Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you'd have $830,492!*
Now, it's worth noting Stock Advisor's total average return is 982% — a market-crushing outperformance compared to 171% for the S&P 500. Don't miss out on the latest top 10 list, available when you join .
See the 10 stocks »
*Stock Advisor returns as of May 19, 2025
Suzanne Frey, an executive at Alphabet, is a member of The Motley Fool's board of directors. Randi Zuckerberg, a former director of market development and spokeswoman for Facebook and sister to Meta Platforms CEO Mark Zuckerberg, is a member of The Motley Fool's board of directors. John Mackey, former CEO of Whole Foods Market, an Amazon subsidiary, is a member of The Motley Fool's board of directors. Anthony Di Pizio has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Alphabet, Amazon, Berkshire Hathaway, Meta Platforms, Microsoft, and Snowflake. The Motley Fool recommends the following options: long January 2026 $395 calls on Microsoft and short January 2026 $405 calls on Microsoft. The Motley Fool has a disclosure policy.
Had You Bought This Artificial Intelligence (AI) Stock When Warren Buffett Sold It, You'd be Sitting on a 47% Return Now was originally published by The Motley Fool
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Yahoo
5 minutes ago
- Yahoo
Vistra Stock Could Chart Path Back to Record Highs
Vistra Corp (NYSE:VST) stock is up 0.6% to trade at $198.53 at last check, not too far off from its Aug. 5 all-time high of $216.85. The equity is looking to snap a three-day losing streak, with the $190 region providing support as it pulled back from that record peak. Better yet, a historically bullish signal now flashing could soon help shares extend their healthy 150.4% lead. Per Schaeffer's Senior Quantitative Analyst Rocky White, Vistra stock's pullback placed it within one standard deviation of its 50-day moving average. The equity was above this trendline in at least eight of the last 10 trading days, and spent 80% of the past two months above it. Within these parameters, eight other signals occurred over the last three years, after which the security was higher one month later 75% of the time, averaging an 8% gain. From its current perch, a similar move would place back above $213. At the International Securities Exchange (ISE), Cboe Options Exchange (CBOE), and NASDAQ OMX PHLX (PHLX), VST's 50-day put/call volume ratio of 1.17 ranks higher than all readings from the past year. In other words, the security could benefit from an unwinding of pessimism. Options are affordable, making this an opportune time to bet on VST's next moves. This is per its Schaeffer's Volatility Index (SVI) of 49%, which stands in the 10th percentile of readings from the past year.
Forbes
7 minutes ago
- Forbes
Cold Data At FMS And Magnetic Tape Data Recovery
The FMS conference has moved from a concentration on solid state storage to other types of storage as well. In this article we will explore a session on cold storage. In addition, we will report on an on-going project to recovering data from old tapes for training AI and other applications. The cold storage session was moderated by Rich Godomski from Fujifilm and included talks by John Monroe from Further Market Research, Ilya Kazansky, CEO, SPhotonix, Steffen Hellmold, President, Cerabyte, Inc., Dave Landsman, Distinguished Engineer, Director Industry Standards, Western Digital and Alistair Symons, VP, Storage Systems Development, IBM. John Monroe's message was that about 70% of the data stored in enterprise and data center applications is cold data, but that it may need to be accessed at any time to support modern AI workflows. Thus, it should be stored on an active archive, which would save on power consumption and operating costs. He also felt that in the future AI will be used to manage what could otherwise become an enormous amount of stored data. Although much of this cold storage is currently on HDDs, he projected that magnetic tape and emerging optical archive storage technologies would grow to over 20% of the total shipped storage capacity by 2030, compared to about 16% in 2024 (which is somewhat higher than my projection of about 12% in 2024). Ilya Kazansky CEO of SPhotonix spoke about the company's optical storage technology, its 5D fused quartz-based memory crystals, an example is shown below. He argued that conventional storage technologies need to be replaced every 10-15 years and that migrating content to new storage technology takes a significant amount of effort and money and that it results in significant waste. The commonly used storage technologies are also susceptible from damage from heat, humidity, chemical reactions and electromagnetic pulses and that they require controlled environments to preserve the storage media. The company's memory crystals provide a write once, read many times or WORM media that can persist for a thousand years or more. Recording on this media are also not subject to damage from heat, humidity and the other factors that can cause loss of data in other storage media. Thus, it would not require migration to new media on a regular schedule and is easy to recycle, since it has a composition similar to sand. He said this volumetric storage media can support storage densities up to 10GB/mm3. The femto-second lasers used for recording record nanostructures that can encode data in their width, height, depth, polarization and birefringence, hence the 5D name. Steffen Hellmold, President of Cerabyte, spoke about their optical media technology for archiving applications. He also stressed the costs of archiving and management required with current digital storage media. Similar to the SPhotonix media, Cerabyte media could last a long time since it is not sensitive to heat, humidity and other factors that can damage the media or the data on conventional digital storage media. He stressed the need to move from traditional archiving approaches to active archiving to support data needs of modern AI workflows. The Cerabyte media uses glass substrates, which are coated with a ceramic material. The Cerabyte solution uses digital light processors, DLPs, from Texas Instruments to effectively write and read millions of bits at once and thus providing high data throughput, as shown below. Although not represented in this session, there are other optical storage startup working on archival technology. These include Folio Photonics and Optera, both of which are developing higher capacity circular optical media that could be placed into traditional optical storage libraries. David Landsman, Distinguished Engineer and Director of Industry Standards, at Western Digital spoke about why HDDs will remain relevant for cold storage applications. He said that HDDs are where cool, versus cold data, lives in data centers. He said that HDDs for data centers are 6X less expensive than data center SSDs and would remain so through 2030 and beyond. One of his slides showed 36-44TB HDDs in 2026, 80-100TB HDDs by 2030 and over 100TB in the years after, as shown below. Alistair Symons, VP of Storage Systems Development at IBM, spoke in favor of magnetic tape for archive storage in the session. He argued that the increasing amount of storage that AI and other applications is a significant factor in the increasing energy requirements at storage centers and this will favor storing less frequently accessed data on magnetic tape. Compared to SSDs and HDDs the energy consumption for magnetic tape storage is considerably less. Chuck Sobey of Channel Science and the organizer of the Monday training sessions before the regular FMS sessions, has been pursuing the development of a multi-format, minimal-contact legacy tape reader that can recover data from old magnetic tapes. There are vast libraries of older magnetic tapes in now obsolete formats that contain valuable data for AI training and other applications or that may contain other content that has a long-standing cultural value. Training on this often-irreplaceable data, such as cultural, seismic, climate or astronomy data, could make better domain specific AI-based models. Currently many efforts to recover data from older tapes requires having the original tape equipment to read the tapes. In addition, even if you can find the old tape drives, replacing worn out parts such as heads is hard because these heads are no longer in production. In addition, even if you have a functioning vintage tape drive it only has the capabilities that were designed into the original drives. To meet the need to read these old magnetic tapes Channel Science created its multiformat tape reader. The company has one patent granted on the technology and more pending. The prototype tape transport device is shown below. Developed using SBIR grants the device uses GMR sensors, in mass production for HDD heads, to read the data from the tapes with a gentle tape path and a high-speed tape transport with advanced signal processing, detection and decoding and uses AI to optimize its operation for different types of tape. To find out more, contact Chuck Sobey at csobey@ The 2025 FMS Conference included sessions and exhibitors working on the increasing demand for colder storage for AI and other applications. Channel Science developed a multiformat magnetic tape reader for recovering data from old tapes.
Bloomberg
7 minutes ago
- Bloomberg
Baiocchi on ELFY's Performance and Electricity Demand
SS&C Alps Advisors Head of Fund Sales & Strategy Paul Baiocchi discusses ELFY's performance and the impact AI has on electricity demand with Scarlet Fu and Katie Greifeld on 'ETF IQ.' (Source: Bloomberg)
