'Something Has Gone Completely Wrong': Palantir's Alex Karp Goes Ballistic On OpenAI, Anthropic

by Tyler Durden

On Wednesday, Palantir CEO Alex Karp delivered a blistering critique of frontier AI labs, accusing them of having an "effing insane" business model that leaves enterprises paying escalating token costs for limited value while risking their proprietary data and intellectual property. He said that top AI labs such as Anthropic and OpenAI were misleading corporate partners - "overselling" the risks of AI while at the same time offering their most powerful models to companies and governments worldwide. 

When CNBC host Becky Quick said "You sound pretty angry," Karp replied: "This is the voice of American business that is being channeled through me," he told CNBC, joking later that he might "get kicked out of the room."

Karp's comments come amid a "revolt" against U.S. AI labs - driven by a combination of high costs, questionable ROI, and increasing regulatory headwinds which have driven some major U.S. companies to cheaper Chinese alternatives, creating multiple pressure points for OpenAI, Anthropic, and others.

"Something Has Gone Completely Wrong"

Karp slammed the token model used by Anthropic and OpenAI - saying "I'm not throwing shade at them, but something has gone completely wrong," he said. "The basic view among enterprises in this country is I'm going to chillax and waste my time with tokens."

He argued that that labs were overselling risks while simultaneously pushing powerful models, and that companies were effectively paying a "wealth tax" that transferred their "alpha" (competitive advantage) to third parties. On national security, he warned: "Are we really going to outsource the battlefield of this country to the consensus view in Silicon Valley? That is effing insane."

The remarks aligned with a 9-point "AI sovereignty" manifesto Palantir posted on X the day before, which criticized "tokenmaxxing" for incentivizing disposable scripts over robust systems and urged institutions to retain control over their data, model weights, and competitive edge.

Via X: 

"Controlling your weights is controlling your fate," the manifesto stated. "If you let others control your weights, you are allowing them to migrate the alpha of your business to theirs."

Karp noted Palantir's expanded Nvidia partnership, which enables custom, sovereign AI deployments where customers retain control over compute, models, data, and weights - a direct counter to the metered frontier API model.

Watch CNBC's full interview with Palantir CEO Alex Karp

The Cost-Driven Shift To Chinese Models

As we've been noting, high token prices and mixed returns have prompted several U.S. companies to adopt or explore Chinese open-weight models:

• Microsoft is considering a Microsoft-hosted, fine-tuned version of China's DeepSeek V4 (or another open-source model) as a lower-cost engine for its Copilot Cowork agentic tool, as it moves toward usage-based pricing.
• Coinbase CEO Brian Armstrong revealed the company cut internal AI spending by nearly 50% by defaulting engineers to Chinese open-weight models (Zhipu AI's GLM 5.2 and Moonshot AI's Kimi series) via an internal gateway, while maintaining high usage.
• Cursor, a fast-growing AI coding startup, built its Composer 2 model on top of Moonshot AI's Kimi K2.5 (backed by Alibaba).

Data from OpenRouter shows Chinese models capturing a rapidly growing share of global token consumption - in some periods exceeding 60% among top models - as enterprises seek cost relief without fully sacrificing capability.

Karp had warned against underestimating China's progress; these examples illustrate the trend in real time.

Watch the entire interview below: