Palantir's revenue grew 93% and its US commercial business 149% — an acceleration at the scale where software companies are supposed to slow down

Every figure in this article is drawn from Palantir's second-quarter 2026 earnings release filed with the SEC on 3 August 2026, or from CNBC and The Register as attributed. Sections marked as analysis are identified as such.

The ordinary life-cycle of a software company is that growth decays. A business doubling at $100 million of revenue grows 60 per cent at $500 million, 35 per cent at $2 billion, and 20 per cent after that; the base gets larger, the addressable market gets smaller, and the arithmetic is unforgiving. It happens to almost everyone.

Palantir grew revenue 93 per cent in the June quarter, to $1.935 billion — a rate it never reached in its early years as a public company, when it was a fraction of this size, and comfortably above the 82 per cent it now guides to for the full year. The curve is going the wrong way round.

The quarter

United States revenue rose 115 per cent to $1.573 billion, and within it the two halves both accelerated: US commercial up 149 per cent to $764 million, US government up 90 per cent to $809 million. The company closed 220 deals worth at least $1 million, 98 worth at least $5 million and 73 worth at least $10 million.

Profitability at that growth rate is the genuinely rare part. GAAP income from operations was $912 million, a 47 per cent margin; adjusted operating income $1.194 billion at 62 per cent; GAAP net income $1.062 billion, a 55 per cent net margin; adjusted free cash flow $1.220 billion at 63 per cent. Palantir's own preferred summary — revenue growth plus adjusted operating margin, the industry's "Rule of 40" — came to 155.

The forward book supports it. Total contract value closed in the quarter was $3.373 billion, up 49 per cent, of which a record $2.132 billion was US commercial, up 153 per cent. US commercial remaining deal value stands at $6.238 billion, up 124 per cent — meaning the contracted future revenue in that segment alone is now larger than the company's entire revenue last year.

Guidance was raised to $8.150–8.158 billion for the full year, about 82 per cent growth, with US commercial revenue above $3.424 billion — at least 134 per cent — and adjusted free cash flow of $4.5 to $4.7 billion. The company holds $9.2 billion in cash and short-term Treasuries.

One number in that guidance is easy to skim past and should not be. Palantir expects adjusted free cash flow of up to $4.7 billion on revenue of about $8.15 billion — a conversion rate of well over half. Software businesses of this quality exist, but almost all of them took twenty years to get there and none of them were simultaneously growing at 82 per cent.

What Karp says he is selling

Alex Karp's language in the release is worth quoting at length because it is a positioning statement, not a boast: "Demand for AI sovereignty has now been unleashed... Our customers trust us to provide them with maximal control over their operations, data, and decisions. Their competitive advantage should never become the training data for future models."

Strip the rhetoric and the claim is precise. Palantir is not selling a model; it sells the layer that connects an organisation's private data to whichever model it chooses, inside boundaries the organisation controls. The pitch to a bank, a hospital system or a defence ministry is that they can use frontier AI without handing their operational data to the company that trained it.

That positioning has become considerably more valuable in the past year. Every enterprise that spent 2025 running AI pilots discovered the same two obstacles: the data was not in a usable state, and the legal department would not approve sending it anywhere. A vendor that solves both at once — and that already holds security clearances and defence accreditations — has a narrow but extremely well-defended slice of the market.

This quarter was otherworldly. — Alex Karp, co-founder and chief executive

Analysis: why the acceleration is real, and where it comes from

There is a straightforward explanation for the shape of this curve, and it is not that Palantir has repealed the laws of enterprise software. It is that the company spent fifteen years building deployment infrastructure for a demand event that had not happened yet, and the event arrived. Every large organisation acquired an AI mandate in the space of about eighteen months, and almost none of them had clean, governed, connected data to point it at.

That is the bottleneck Palantir sells into, and it explains the 149 per cent. It also explains why the growth is concentrated in US commercial rather than government: the government work is procured slowly and was largely already won, while the corporate demand is new, urgent and — on the evidence of $2.1 billion of commercial contract value closed in a single quarter — being signed at unusual speed.

The deal-count disclosure supports that reading better than the revenue does. Seventy-three contracts of at least $10 million in a single quarter is not a handful of mega-deals flattering a growth rate; it is a broad base of large customers arriving at once. Palantir spent years being told its business could not scale because every deployment needed its engineers on site. The current numbers suggest it finally productised that work.

The honest question is duration rather than reality. A backlog-driven acceleration of this kind draws forward demand: companies buying a data platform in 2026 because they need one now are not buying another in 2028. Karp has said he intends to hold these growth rates for eighteen months, which is both an ambitious commitment and a tacit acknowledgement that somebody is counting.

Analysis: the part that is not in the numbers

Palantir's government business is the source of both its credibility and its controversy. It holds a consolidated US Army software framework reported at up to $10 billion over ten years, a substantial UK defence engagement, and contracts across the Space Force, the IRS and the Treasury; The Register reported this month on an internal Defense Department memo contemplating up to $244 million more through 2028 on a no-bid basis.

The company's work with immigration enforcement has drawn sustained criticism from lawmakers and civil-liberties groups, and from a group of former employees who went public earlier this year. Palantir's answer has always been that it chooses its customers deliberately and that democratic states are entitled to capable software. Both things can be true; what matters commercially is that this is a political risk of a kind most software companies do not carry, and it does not diminish as the contracts grow.

There is also a concentration question that the acceleration currently obscures. Government revenue is still 42 per cent of the total and overwhelmingly American, which ties a meaningful share of the business to one country's appropriations cycle and one administration's priorities. Fast-growing commercial revenue dilutes that exposure over time — but only if the commercial growth is durable, which is the same open question as before.

Then there is the valuation, which is where even admirers hesitate. On any conventional measure — earnings, revenue, cash flow — Palantir trades at one of the highest multiples in large-cap software, and the price embeds not just this year's 82 per cent growth but several more years of something like it. The business is now unambiguously excellent. The share price is a separate question, and the answer to it depends entirely on whether Karp's eighteen months turn into five years.

What is no longer arguable is the thing sceptics spent five years arguing about. Palantir was long dismissed as a consultancy in software clothing — bespoke, unscalable, forever explaining its margins. It just posted a 55 per cent net margin on 93 per cent growth. Whatever else it is, that is not a consultancy.

The reasonable position on Palantir in August 2026 is therefore uncomfortable for both camps. The bears have lost the argument about the business model outright; the numbers are among the best in the industry and they are audited. The bulls have won that argument and are now betting on something else entirely — that a demand surge created by a single technology shift will last long enough to justify a price that assumes it never ends.