Palantir's US commercial revenue overtakes government for the first time as AI demand surges

The commercial number is the headline: $984 million, up 148% year on year, against $953 million from US government work, up 96%. CNBC reported the results after the market closed on Monday, 2 November. It put total revenue past $2 billion for the first time. Net income more than doubled to $1.13 billion, or 44 cents a share, from about $476 million, or 18 cents, a year earlier.

The company also lifted guidance. Full year US commercial revenue should now come "in excess of" $3.56 billion, up from the $3.42 billion forecast a quarter earlier, according to CNBC's account. Fourth quarter revenue should reach about $2.65 billion and full year 2026 about $8.53 billion.

CEO Alex Karp told CNBC's Seema Mody he did not believe anyone had posted such consistent results in the history of software AI. He added that the business was succeeding without salespeople. He also repeated his call to nationalise AI labs, describing it as the only realistic way to manage liability, per CNBC. That is a striking line for a company whose revenue base has long leaned on government work.

Public sector buyers are not waiting for the dust to settle

The same week produced a cluster of signals that procurement itself is being rebuilt around AI, and not only inside the United States. China used a three-day forum in Xiong'an New Area to float a UN-led model for AI governance. It published a consensus document called Global Governance: Xiong'an Vision on the Sunday that closed the event. The South China Morning Post reported on 27 October that the meeting drew delegates from more than 90 countries and international bodies, according to Chinese officials. Beijing paired its governance pitch with advocacy for open-source AI cooperation and a bigger role for the Global South.

That is a rules argument, not a contract award. But the two are linked: whoever writes the procurement standard shapes which models get bought. Germany used the start of this week to publish a national AI plan aimed at closing the gap with the US and China, according to Dataconomy on Tuesday, 3 November. Neither report gives the plan's contents, so the scale of any public spending attached to it is not established here.

The buying side in Washington is moving faster and with more paperwork. Anthropic cut the price of Claude for US government buyers by 40% under a new GSA contract, BigGo Finance reported on 28 October, and extended the arrangement to states and cities. The Manila Times reported on 30 October that Anthropic and reseller Carahsoft had expanded access to Claude across US government through a GSA OneGov agreement. A separate piece from the same week, via news.google.com, put the discount at 40% and tied it to the same GSA deal.

"The attackers behind the activity used tailored social engineering techniques to masquerade as Dubai Airports recruiters, using a trojanized Visual Studio coding challenge to target a high-value job seeker based in Iraq," Unit 42 said.

That quote belongs to a different story, and it matters here for a narrower reason: the security of the systems being procured. Palo Alto Networks Unit 42 described a campaign it tracks as CL-STA-1178. In it, attackers posed as Dubai Airports IT recruiters to deliver malware inside a fake coding assessment. The Hacker News reported on 2 November that the activity was first detected in late March 2026, with infrastructure staging dating to November 2025. Government buyers signing AI contracts at speed also sign up for the attack surface that comes with them.

The paperwork problem

Federal procurement rules have not caught up. Industry and legal experts urged the FAR Council to set standards for AI in federal procurement, Federal News Network reported on 28 October. That signals the rulebook is still being drafted while agencies are already buying. HHS sought a temporary sole-source contract for Anthropic's AI, a request reported on 28 October. That is the kind of shortcut that tends to appear when demand outruns the standard vehicle.

The UK picture is no happier. PublicTechnology reported on 2 November that just one in nine officials believes government has the right approach to procuring tech. That is a poll of sentiment rather than a procurement outcome, but it lands awkwardly next to the push for sovereign AI capacity. The same week, THINK Digital Partners asked on 3 November whether procurement holds the key to Britain's digital sovereignty.

Some of the money is moving through distribution rather than direct award. Carahsoft agreed to distribute Sevii's autonomous AI cyber defence platform to government, ExecutiveBiz reported on 29 October. Polygraf AI and Carahsoft partnered to help public sector organisations adopt AI while keeping control of sensitive data, compliance and operational risk, per a release dated 2 November. Tonic.ai and Carahsoft announced a similar arrangement for AI-native synthetic data on 27 October, according to The Manila Times. The pattern is consistent: resellers, not labs, sit between the vendor and the purchase order.

There are smaller data points in the same direction. Docusign said on 3 November that it had acquired Suplari to speed up AI-powered contract management for procurement, per PR Newswire. Harvey announced FedRAMP Class D certification for federal legal work, Unite.AI reported on 28 October. Blue-chip defence work continues to flow: Palantir awarded L3Harris a $53 million TITAN production contract, Securities.io reported on 2 November. CACI won a $108 million Air Force financial systems contract, the same outlet reported on 3 November.

Look beyond the US and the numbers get bigger and messier. Cohere's C$1.6 billion stake, described by tech-insider.org on 2 November as sovereign AI contract terms, ties a national capability to a single supplier for years. The DCSA's FY27 forecast includes a recompete for entity vetting, GovCon Wire reported on 2 November. DHS plans a contract consolidating biometric identity decisions and accuracy oversight, Biometric Update reported on 29 October.

What the Palantir result actually shows

Palantir's split is worth reading carefully. US government revenue still grew 96% to $953 million. It simply grew more slowly than the commercial line, which has more than quintupled over two years, according to CNBC. The company is not losing public sector work. Private buyers who can sign without a FAR clause or a bid protest are outpacing it.

For procurement officials, that is the uncomfortable part. If commercial demand sets the price and the pace of deployment, agencies end up negotiating from behind. They buy capacity that someone else's requirements have already shaped. The GSA discount on Claude looks like an attempt to answer exactly that problem. It standardises terms and cuts the unit price for everyone from a federal agency to a city hall.

None of this settles the governance question. Xiong'an's consensus document, Germany's plan, the FAR Council's unfinished standards and a single quarter of Palantir earnings are four different kinds of signal. But they point the same way: the contracts are being signed now, and the rules are still being written.