Palantir just posted the quarter that breaks whatever mental model you still had about how B2B companies behave at scale. It already had re-accelerated at a rate we’ve never seen before. And then it just … did it again.
- Revenue grew 93% year-over-year to $1.935 billion.
- That is the twelfth consecutive quarter of accelerating growth.
- The company raised full-year guidance from $7.65 billion to $8.15 billion, which is a half-billion-dollar raise in the middle of the year. Net revenue retention hit 157%. Rule of 40 hit 155%.
And yet … going into the quarter, the stock was down 29% for the year. The market was already expecting perfection. This quarter, it beat perfection, and the stock rocketed +20% on the quarter.
Alex Karp called the quarter “otherworldly.” Chief Revenue Officer Ryan Taylor’s framing was sharper: the results were “unprecedented, but entirely unsurprising.”
Here are 5 interesting learnings for B2B and B2B + AI founders.
#1. Growth Accelerated Again at a $7.7 Billion Run Rate
Revenue of $1.935 billion, up 93% year-over-year and 19% quarter-over-quarter. A year ago this was roughly a $1 billion quarter.
The U.S. business is doing the work. U.S. revenue grew 115% year-over-year and 23% sequentially to $1.573 billion, now more than 81% of total revenue. U.S. commercial grew 149% year-over-year and 28% sequentially to $764 million. U.S. government grew 90% to $809 million.
Guidance moved with it:
- Q3 revenue guide of $2.160 billion to $2.164 billion
- FY 2026 revenue raised to $8.150 billion to $8.158 billion, or 82% growth
- U.S. commercial revenue guided to more than $3.424 billion, at least 134% growth
- Adjusted free cash flow guided to $4.5 billion to $4.7 billion
The learning for founders: deceleration at scale is a default, not a law. Every planning model in B2B assumes growth rates compress as the base grows. That assumption holds when the product is a mature category replacement. It breaks when the product sits on top of a platform shift that just got budget. If you are modeling 2027 by discounting your 2026 growth rate by 30%, you may be planning for a decline that the market is not asking you to have.
#2. NRR Hit 157%, and There Are Only 1,049 Customers Total
Net dollar retention was 157%, up 700 basis points sequentially from 150% in Q1.
Now the part that matters more. Palantir has 1,049 total customers. Not 10,000. Not 100,000. About a thousand. Commercial customer count is 870, and U.S. commercial is 653.
At the current run rate, that is roughly $7.4 million per customer annually across the base. In U.S. commercial specifically, $764 million a quarter across 653 customers annualizes to roughly $4.7 million per customer.
Customer count grew 24%. Revenue grew 93%. Almost none of this growth is coming from logo acquisition. It is coming from existing customers buying dramatically more.
And 157% understates it, because net dollar retention excludes revenue from customers acquired in the trailing twelve months. The fastest-expanding cohort is not yet in the number.
The learning for founders: in B2B + AI, the expansion motion is the business. A customer who deploys one AI workflow that works does not add 15% next year. They add 50% or 100%, because the next workflow is already identified and the trust is already paid for. If your NRR is sitting at 105% while you pour money into new logo acquisition, you are solving the wrong constraint.
3. Rule of 40 of 155%, and Most of It Is GAAP-Real
The margin profile:
- Adjusted operating income of $1.194 billion, a 62% margin, up from 46% a year ago
- GAAP operating income of $912 million, a 47% margin, versus $269 million a year ago
- GAAP net income of $1.062 billion, a 55% net margin, up 225% year-over-year
- Adjusted free cash flow of $1.220 billion, a 63% margin, up 115%
- $9.2 billion in cash and short-term Treasuries, no debt
Rule of 40 came in at 155%, up 10 points from Q1’s 145%. Karp noted last quarter that scores in that range had previously belonged to AI infrastructure and chip companies, not software.
Two honest caveats. First, unrealized gains on Palantir’s SpaceX position contributed $0.03 to GAAP EPS and $0.02 to adjusted EPS. Strip that out and adjusted EPS of $0.41 becomes $0.39, still ahead of the $0.35 consensus, but it is not all operations. Second, the 15-point gap between the 62% adjusted operating margin and the 47% GAAP operating margin is mostly stock-based compensation. Real cost, real dilution. A 47% GAAP operating margin at 93% growth is still without precedent at this scale, but the adjusted number is the friendlier one.
The learning for founders: the market stopped paying for growth alone in 2025 and has not gone back. Growth plus GAAP profitability is the bar now. When you can fund your own growth out of operating cash flow, you also get to raise guidance aggressively without worrying about what the financing window looks like next quarter.
4. Backlog Is Growing Even Faster Than Revenue
This is the number most people skip, and it is the most predictive one in the release.
- Total remaining deal value of $13.1 billion, up 83% year-over-year
- U.S. commercial remaining deal value of $6.24 billion, more than double a year ago and up 124%
- U.S. commercial TCV bookings of $2.132 billion in the quarter, up 153% year-over-year and 81% sequentially
- Total TCV of $3.37 billion, up 49%
U.S. commercial bookings grew 153% while U.S. commercial revenue grew 149%. Bookings outrunning recognized revenue means the next several quarters are already substantially sold. That is why management could raise the full year by $500 million without needing to be brave about it.
The learning for founders: track your contracted-backlog-to-revenue ratio as a leading indicator, and report it internally every month. Revenue tells you what you sold two to four quarters ago. RPO and remaining deal value tell you what the next four quarters look like. When backlog growth exceeds revenue growth, you can invest ahead of the curve with actual evidence rather than optimism. When it inverts, you have two quarters of warning before the revenue line shows it.
5. 73 Deals of $10M+ in a Single Quarter
Deal counts for Q2:
- 220 deals of at least $1 million
- 98 deals of at least $5 million
- 73 deals of at least $10 million
One third of the disclosed million-dollar deals were $10 million or larger. In Q1 the $10M+ count was 47. It went to 73 in one quarter.
The individual deals show the mechanic:
- A multinational technology company that started with a single engagement at one operating company expanded into a three-year deal worth nearly $370 million
- A nonprofit health system converted a pilot into a three-year partnership worth $37 million in TCV
- A global asset manager signed a three-year, $35 million deal
- A global software and services company signed an initial $15 million agreement, five months long, after attending an Agent Camp
A hands-on workshop produced a $15 million contract with a five-month term. Short duration, high value, fast to sign. Palantir has built a motion where the sales event and the proof event are the same event.
The learning for founders: land in one workflow with a measurable operational result, then expand across operating companies and functions. The $370 million expansion did not start as a $370 million negotiation. It started as one deployment at one subsidiary. In enterprise AI, credibility earned inside one P&L is the cheapest possible path into the next twelve.
5 More Quick Hits from the Quarter
- $9.2 billion in cash and short-term Treasuries, and no debt. With adjusted free cash flow guided to $4.5 billion to $4.7 billion for the year, the balance sheet is now a strategic weapon rather than a runway calculation. Nothing about the growth plan depends on the financing window staying open.
- The government business is not the slow half. U.S. government revenue grew 90% to $809 million and accelerated sequentially. CTO Shyam Sankar noted that more than 25,000 builders are now developing agents and applications on Maven, and that trailing-twelve-month revenue from the Department of Defense still sits below 25 basis points of the Pentagon’s budget. That is the definition of headroom.
- Agent Camp is a real go-to-market channel. One of the quarter’s named deals was a $15 million, five-month agreement with a global software and services company that signed after attending an Agent Camp. Hands-on build sessions that produce a working artifact are outselling traditional enterprise pipeline motions.
- Model-agnostic is the actual pitch. Sankar said a standard Nemotron Ultra model outperformed frontier models on five production tasks within 24 hours of deployment. The message to buyers is that general benchmarks do not predict performance on your workflows, and that you should not be locked into one provider. Karp’s framing of it: a customer’s competitive advantage should never become training data for someone else’s model.
- U.S. commercial customer count grew only 6% sequentially while U.S. commercial revenue grew 28%. Almost five points of revenue growth for every point of logo growth. If you want one number that explains the whole quarter, that is it.
What This Quarter Signals for B2B + AI Founders
- Concentration is the real risk in the model, not growth. The U.S. is 81% of revenue. International government revenue grew 42% to $181 million, well behind everything else, and the international commercial business (derived from the segment disclosures at roughly $181 million) remains the weak spot. This is a company winning enormously in one geography. Karp said on the call he is trying to hold U.S. commercial growth at or near the current rate for the next 18 months. That is the whole thesis in one sentence.
- Execution and expectations are separate problems. The stock was down 29% for the year going into a quarter where revenue grew 93%, margins expanded 16 points, and guidance moved up half a billion dollars. It jumped about 12% after hours. When a company is priced at extreme multiples, even historic quarters produce volatile outcomes. That dynamic applies to private rounds too. Raising at a price that requires perfection means every good quarter is merely adequate.
- The operating model is the story under the numbers. Palantir is running roughly $7.7 billion in annualized revenue on approximately 4,500 employees, based on the most recently disclosed headcount. Karp put annualized revenue per employee at $1.5 million in Q1. On the same basis, this quarter’s run rate implies something closer to $1.7 million. That is a calculation, not a disclosure, so treat it as directional. But the direction is the point: revenue per employee is compounding faster than headcount, and the companies winning in 2026 are treating headcount as a constraint to engineer around rather than a lever to pull.
93% growth. 157% NRR. 155% Rule of 40. 73 deals over $10 million in ninety days.
The question of whether Palantir is the best-performing enterprise software company ever built at this scale is getting harder to argue against every quarter.
