In July, Palantir Technologies Inc. (PLTR) chief executive Alex Karp went on CNBC and called the AI industry’s token pricing model “effing insane”, accusing frontier labs of overcharging enterprises while quietly absorbing their data.

The rant went viral, but it also doubled as a pitch. Karp’s argument was that renting AI by the token means handing a vendor the workflows and proprietary data that make a business valuable in the first place.

That case only works if Palantir has somewhere else to send customers. Its software, AIP, Foundry and Ontology, runs on top of compute infrastructure the company itself does not own.

Sovereign AI is industry shorthand for systems a customer fully controls, its data, its model weights, and the hardware underneath, rather than a subscription to someone else’s cloud.

For a bank or a defense contractor, that distinction can decide whether a workload is allowed to touch a public cloud at all.

Palantir had already patched part of that gap. In June, it paired with Nvidia Corporation (NVDA) to run open Nemotron models for government agencies under the same sovereign AI label.

On September 8, Palantir named Nebius Group N.V. (NBIS) its preferred sovereign AI infrastructure partner for commercial customers, extending the same model beyond government work.

Nebius fills the compute gap Karp complained about

Once integration is complete, Nebius’s compute and inference endpoints will sit inside Palantir’s enterprise perimeter.

Eligible customers can then deploy open-weight models on Nebius hardware, fine tune them with proprietary data, and keep both under their own control.

The companies also plan modular data centers at sites where power is already secured, a detail that matters because electricity, not chips, has become the real bottleneck in AI capacity. Neither company disclosed financial terms of the arrangement.

Palantir has named Nebius Group its preferred sovereign AI infrastructure partner for commercial customers.

Bloomberg / Getty Images

Wall Street rewards the compute side, not the software

The market’s verdict was immediate and lopsided. Nebius shares climbed as much as 6% to $239.23 Tuesday morning before extending gains intraday, touching a high near $250, while Palantir slipped about 1%.

That split says more about where analysts see leverage than about either company’s fundamentals.

Related: Nebius stock quietly defies its own history after earnings

Nebius carries a consensus Strong Buy rating and an average 12 month price target near $286.69, according to S&P Global Market Intelligence data, implying upside even after this week’s rally. Targets range as wide as $144 to $415, a spread that shows how unsettled the compute business model still is.

Palantir’s consensus rating is a milder Buy from 32 analysts, with an average target near $191.68, below where the stock already trades after a blowout summer.

DA Davidson has kept a Neutral rating and a $180 price target, warning the stock trades near 93 times expected 2026 revenue, a valuation few software peers can match.

Nebius’s own numbers explain the enthusiasm, and the risk

The excitement has real backing. Nebius reported second quarter revenue of $582.3 million, up 454% year over year, with adjusted EBITDA swinging to a $236.2 million profit, according to its earnings release.

Chief executive Arkady Volozh has said the company could sell its entire 2027 capacity today under similar contract terms.

But three clients accounted for 59% of Nebius’s second quarter revenue, and the Palantir arrangement carries no disclosed dollar figure, 24/7 Wall Street noted. For now, the preferred partner label is a credential, not a contract.

Palantir’s own quarter, by contrast, already shows up in revenue. Second quarter sales grew 93% year over year to $1.935 billion and U.S. commercial revenue jumped 149%, the company reported, with results filed with the SEC.

Palantir also raised full year guidance to imply 82% revenue growth, a bar few enterprise software peers are even trying to clear.

More AI:

  • Nvidia just made a move Wall Street wasn’t ready for
  • Microsoft just took sides in AI policy fight
  • OpenAI just disclosed something genuinely alarming

Palantir is building two separate sovereign AI tracks

Stack the two deals together and a pattern appears. Nvidia handles the government track, Nebius now handles the commercial one, and Palantir supplies the software layer for both without owning a single data center of its own.

That positioning matters because sovereign AI is no longer a niche pitch. McKinsey estimates the market could reach $500 billion to $600 billion globally by 2030, as regulated industries and governments balk at handing data to generic cloud platforms.

Only a few dozen countries currently host the advanced compute needed to serve that demand domestically.

Rivals are not standing still. CoreWeave, the closest peer to Nebius, is up roughly 35% year to date and joined the Nasdaq 100 without needing a Palantir style endorsement, a reminder that compute providers have more than one software partner to court, and that Palantir will need more than one preferred label to keep its edge.

The next real test will not be today’s stock move. It will be whether Palantir’s customers actually migrate onto Nebius infrastructure once the integration window closes, converting a preferred partner label into contracted revenue analysts can price with confidence.

Both companies report next in early November, the first checkpoint for that conversion.

Related: Bank of America resets Nebius stock price target after 454% surge