Palantir CEO Alex Karp slammed the artificial intelligence industry as “effing insane” on July 1, 2026, during a CNBC appearance, accusing OpenAI and Anthropic of imposing unsustainable token costs on enterprise customers. “Something has gone completely wrong,” Karp told CNBC’s Squawk Box, arguing that companies are increasingly frustrated with skyrocketing fees for token-based AI pricing and are shifting toward cheaper, open-weight alternatives.
Karp’s core complaint centers on the economics of token-based models: enterprises pay per token consumed, creating unpredictable costs that scale with usage rather than delivering measurable value. He pointed out that many CEOs are now questioning whether the return on investment justifies the expense, with businesses pivoting away from what he calls “tokenmaxxing”—the industry’s obsession with token consumption—in favor of models they can control and own.
The frustration Karp articulates reflects a broader enterprise crisis. Uber, for example, burned through its entire 2026 AI budget in just four months, with per-engineer costs ranging from $500 to $2,000 monthly, according to Fortune. A single two-hour coding session by one Uber executive cost $1,200 in tokens, prompting the company to cap employee AI spending at around $1,500 per person per month. TechCrunch reported in June 2026 that companies across the industry are scrambling to curtail soaring AI costs as token-heavy applications drive up bills faster than leadership anticipated.
Karp’s timing coincided with Palantir’s announcement of an expanded partnership with Nvidia, unveiled on June 29, 2026. The deal pairs Palantir’s software platform with Nvidia’s Nemotron open-weight AI models, designed to let U.S. government agencies build and deploy their own models while maintaining full control over data and compute. Karp framed this as the solution: “What aligns me with Nvidia, and I think is what the technical customers want, is control over their compute, their models, their data stack and their alpha. They want to know they own the means of production.”
The shift toward open-weight models is accelerating across the industry. Chinese open-weight models cost 60 to 90 percent less than Anthropic and OpenAI’s offerings, according to a CNBC report from July 7, 2026, citing data analyst Justin Summerville. Microsoft is reportedly ditching proprietary AI models from OpenAI and Anthropic in favor of its own to reduce costs, while Amazon’s chief technology officer told Fortune in July 2026 that companies worried about mounting AI bills are increasingly switching to cheaper, open-source alternatives.
The token-pricing model has become a critical vulnerability for frontier AI labs. OpenAI was reportedly weighing steep price cuts in June 2026 to compete with Anthropic, according to the Wall Street Journal and Reuters, as both companies face enterprise customers demanding better economics. Anthropic itself shifted from flat-rate enterprise pricing to per-token billing in April 2026, a move that exposed the true cost of AI consumption and sparked backlash from customers accustomed to predictable spending.
Karp’s public criticism amplifies what many executives have been saying privately. The Los Angeles Times reported in July 2026 that the Silicon Data LLM Token Expenditure Index, which tracks what users pay for AI tokens, had fallen nearly 20 percent from a May peak, yet enterprise AI bills were rising—a paradox driven by surging usage that outpaces the modest per-token price declines. Goldman Sachs predicted in May 2026 that token consumption would multiply 24 times between 2026 and 2030, multiplying the cost pressures on businesses unless pricing models change.
Sources
- CNBC — Karp’s July 1, 2026 Squawk Box interview criticizing OpenAI and Anthropic token models; reporting on enterprises shifting to open-weight models and ROI concerns.
- Forbes — Karp’s “effing insane” quote and accusation of a “wealth tax” on businesses via high AI fees and data collection.
- Fortune — Uber burning through its entire 2026 AI budget in four months with per-engineer costs of $500–$2,000 monthly; Amazon CTO commentary on enterprise shift to open-source models.
- TechCrunch — Companies scrambling to curtail soaring AI costs in June 2026; Uber capping employee AI spending after budget exhaustion.
- Palantir investor relations — Announcement of Palantir and Nvidia partnership on June 29, 2026, deploying Nemotron open-weight models for U.S. government agencies.
- Wall Street Journal and Reuters — OpenAI considering drastic price cuts in June 2026 to compete with Anthropic.
- Los Angeles Times — Silicon Data LLM Token Expenditure Index down 20% from May 2026 peak; enterprise AI bills rising despite falling per-token costs.
- Goldman Sachs — Token consumption forecast to multiply 24 times between 2026 and 2030.
