Two ways to earn from enterprise AI
Palantir and Snowflake sell different parts of the enterprise AI stack. Their October 2025 integration partnership lets customers combine Snowflake's data platform with Foundry and AIP. A customer can use both; this comparison is about the investment cases. Partnership announcement.
Snowflake has a fresh development today: database researcher Jeff Naughton has joined as Technical Fellow to help turn research into capabilities for AI workloads. That adds context to its ambitions. The financial test is still whether customers' usage turns into revenue, cash and returns per share. October 12 announcement.
Palantir's operating lead is substantial
Palantir's quarter ended June 30 produced $1.935 billion of revenue, up 93%, and a 47.1% GAAP operating margin. Snowflake's quarter ended July 31 produced $1.547 billion of total revenue, up 35%, with a 17.0% GAAP operating loss margin. Snowflake's often-quoted 37% growth figure applies to product revenue. Palantir results, Snowflake results.
That is a strong lead for Palantir on current execution. The periods finish a month apart, and its government exposure and implementation model differ from Snowflake's consumption business. We should carry those differences into the valuation discussion.
Stock compensation also deserves a place beside the growth numbers. Across the latest trailing periods, the cash-flow-statement stock compensation add-back was $836 million at Palantir and $1.642 billion at Snowflake: approximately 13.6% and 30.2% of sales. Snowflake's figure is net of capitalized amounts. These exclude related payroll taxes and are not cash payments. Palantir annual results, Snowflake annual results, with the latest half-year statements linked above.
Quarterly weighted-average basic shares increased 1.46% year over year at Palantir and 4.19% at Snowflake. Basic shares give us the same measurement on both sides; Snowflake's losses exclude potential dilution from its GAAP diluted count. Stock compensation and share growth are related checks, not interchangeable figures.
Snowflake also priced a $3.75 billion zero-coupon convertible-note offering in late September, after its latest reported balance-sheet date. Capped calls are intended to reduce potential conversion dilution, subject to a cap; proceeds also fund note repurchases and other corporate uses. That funding and its obligations belong in the next balance-sheet review. Financing announcement.
What investors pay for that difference
Friday's closing prices were $209.05 for PLTR and $368.89 for SNOW. Multiply those closes by the common shares reported on the latest quarterly filing covers, and the equity-value references are roughly $502.36 billion and $130.14 billion. Palantir's share count is dated July 27; Snowflake's rounded count is dated August 21. They are valuation references, rather than exact October 9 market capitalizations. Palantir filing, Snowflake filing.
Against reconstructed trailing revenue of $6.156 billion and $5.435 billion, that is about 81.6 times sales for Palantir versus 23.9 times for Snowflake. Both denominators cover twelve reported months, ending June 30 and July 31 respectively. Neither uses a fiscal-year forecast.
For a consistent cash check, subtract cash purchases of property and equipment from operating cash flow. Over those trailing periods, Palantir retained $3.358 billion and Snowflake $1.195 billion. That represents cash margins of 54.6% and 22.0%, but cash yields on the equity-value references of only 0.67% and 0.92%.
This is a non-GAAP cash calculation, without either company's payroll-tax add-backs. It excludes acquisitions and financing. Stock compensation remains a non-cash add-back in operating cash flow; cash spent on equity-award withholding classified as financing remains outside this measure. It is not cash available free of every obligation.
Palantir demonstrates much greater operating profitability and cash conversion. Its valuation asks more of future growth. Snowflake's lower sales multiple comes with weaker GAAP profitability and a heavier stock-compensation burden. A relative discount does not establish an absolute bargain.
The price recovery has been powerful
PLTR ended Friday above its $176.98 50-session and $152.33 200-session averages. In the last 252 completed sessions, its largest closing drawdown ran from $207.18 on November 3, 2025 to $107.27 on June 25, 2026, a 48.2% decline. The subsequent rebound to Friday was 94.9%, putting it 0.9% above that earlier peak.
SNOW ended above its $330.25 and $231.05 averages. Its largest closing drawdown in the same window ran from $277.14 on November 3, 2025 to $121.11 on April 10, 2026, a 56.3% decline. The rebound was 204.6%, leaving it 33.1% above that earlier peak. Closing and intraday extremes differ.
From the April 1 close to October 9, PLTR gained 42.7%, SNOW 140.9% and SPY 18.8%. Those are independently calculated price returns, excluding dividends, with the same start and finish. They describe the path already taken. They do not identify why it happened or forecast a repeat. PLTR price cross-check, SNOW price cross-check, with Yahoo daily data and YieldTerminal calculations.
Our Price & Levels archive tracks dated observations and subsequent outcomes. For this comparison, continued closes above the intermediate averages would support the trend; losing them would weaken the technical case without, by itself, settling the business valuation.
Read the insider sales before judging them
In the July 12–October 12 window, our Insider Analyzer found $100.78 million of reported PLTR common-stock sales and no eligible purchases. All 39 sale rows referenced plans. Approximately $70.70 million of those sales explicitly covered tax withholding after vesting, based on the row footnotes in the Karp and Buckley filings. Karp Form 4, Buckley Form 4.
For SNOW, the same scan found $507.41 million of reported sales and no eligible purchases. Of 104 sale rows, 100 referenced plans. The remaining four had no identified plan reference; that does not establish that they were discretionary. One example is Benoit Dageville's October 7 sale of 50,000 shares at a reported $333.42 under a plan adopted in April. Dageville Form 4.
These are gross reported transaction values. The totals use P-coded purchases and S-coded sales, with grants, exercises and F-coded withholding separate. S-coded sales to cover taxes remain included. Marker sizes reflect reported value; their position on the price line is that day's close, not the execution price. The scan covers filings available at retrieval and eligible transaction dates, so later filings can change the picture.
My interpretation: there is no reported insider-buying confirmation in this window. The sales need their plan and tax context; the larger dollar total is not a bearishness ranking.
What would change my view?
For Palantir, the case strengthens if rapid growth continues to produce GAAP profit and cash per share while dilution stays contained. It weakens if growth slows substantially and the valuation still assumes extraordinary execution.
For Snowflake, the case strengthens if usage growth leads to a narrower GAAP loss, better cash conversion and a smaller compensation burden relative to revenue. It weakens if acquisitions and equity issuance keep absorbing the gains from customer growth.
The next reported results are the checkpoint. Our Earnings Calendar distinguishes company-confirmed dates from estimates. Revenue guidance, realized margins and share counts belong in the same follow-up; an AI product announcement alone cannot replace them.
My read: Palantir has the stronger reported business performance. Snowflake has the lower sales multiple. At these reference valuations, both need substantial future progress to make the current cash yield more attractive. I would keep the operating lead and the price paid as separate judgments.
Educational commentary, not personalized investment advice. Market calculations use October 9 completed prices; news and filing scan were reviewed October 12. Trailing figures are reconstructed as prior fiscal year plus current half-year less prior half-year. No forward P/E, historical-low valuation or price target is claimed.