Alex Karp was blunt. The Palantir CEO told CNBC on Wednesday that plenty of American businesses are paying for AI services that deliver almost nothing in return — while handing over their most valuable data. “Every single enterprise in this country, these people are livid,” Karp said.
“They are paying for tokens that create no value. These people are stealing the weights and alpha of my business.” His warning did not arrive in a vacuum.
By the time Karp spoke, a quiet but decisive shift was already underway inside thousands of companies. The timeline of that shift tells the real story. It started with the bills.
Across corporate America, CFOs watched AI spending balloon month after month. The token-based pricing models from OpenAI and Anthropic — pay per query, pay per output — were eating budgets.
The math was brutal: a company could spend six figures a month on premium frontier models and still struggle to show a clear return. CTOs began asking a hard question. Was the marginal improvement in model performance from the most expensive systems worth the cost?
For many, the answer came back no. That opened a door. A new generation of Chinese AI models had been quietly improving. They were not as flashy as the American frontier models.
They did not have the brand recognition. But they were cheaper.
Much cheaper. And for a growing number of enterprises, cheap enough to matter more than a few percentage points on a benchmark test. The pricing battle that followed was not a sudden war.
It was a slow grind. One company after another started testing the cheaper alternatives.
They ran internal pilots. They compared results. What they found was that for many routine business tasks — customer service, data sorting, internal documentation — the lower-cost models performed just fine.
Tiezhen Wang, the former head of Asia-Pacific at Hugging Face, told Rest of World that the pattern is predictable. Companies build their first products using proprietary models. That gets them to market fast.
But as they scale, the costs explode. That is when they start looking at open-source alternatives.
“Switching later could allow firms to reduce spending on AI tokens substantially,” Wang said. The open-source approach has a clear advantage. It cuts out the per-token meter.
A company can run a model on its own servers, pay a flat infrastructure cost, and never see a bill spike because some department had a busy week. That predictability is a powerful lure for any CFO.
Karp’s CNBC appearance crystallized the frustration. He said what many enterprise customers were already feeling but had not said publicly. They were livid.
They felt trapped. They were paying premium prices and getting back tokens — digital units of computation — that created no measurable value for their business. At the same time, they were feeding their proprietary data, their “weights and alpha,” into systems they did not control.
The timeline of this shift is not dramatic. There was no single tipping point.
No one day when everything changed. It was a steady accumulation of cost-conscious decisions. A CFO here, a CTO there, running the numbers and finding that the premium models did not justify their price tags.
By the time Karp spoke, the market was already moving. The cheaper Chinese models were gaining traction.
The pricing battle across the industry was already underway. The question was no longer whether enterprises would switch. It was how fast.
{“@context”: “https://schema.org”, “@type”: “NewsArticle”, “headline”: “Palantir CEO warns US businesses are paying for AI tokens that create no value”, “description”: “Alex Karp was blunt. The Palantir CEO told CNBC on Wednesday that plenty of American businesses are paying for AI services that deliver almost nothing…”, “datePublished”: “2026-07-02T07:15:41+00:00T09:30:00+00:00”, “author”: {“@type”: “Person”, “name”: “Sophie King”}, “publisher”: {“@type”: “Organization”, “name”: “InfoPulse Today”, “url”: “https://newsfeed.infopulsetoday.com/”}}
