AI Startups Slash Spending as Model Costs Squeeze Margins

The cost crunch hits AI-first companies

AI startup Lindy recently made a dramatic shift: CEO Flo Crivello moved 100% of the company's traffic off Anthropic's Claude models to DeepSeek, a Chinese provider offering cheaper open-weight alternatives. The result was immediate—costs plummeted, saving the 25-person company millions of dollars within months.

"It's a matter of survival for the business," Crivello said, noting that Lindy still expects to spend more on AI than payroll despite the cuts.

The move reflects a broader reckoning across the AI industry. After years of unconstrained spending—fueled by the belief that AI would deliver transformative returns—companies are now demanding clearer ROI before continuing to pour money into expensive frontier models from OpenAI and Anthropic.

Uber implemented spending tiers on AI tools this month, capping baseline access at $1,500 per month after the company burned through its entire annual AI budget in just four months, according to details first reported by CNBC.

Why it matters

The spending pullback arrives at a critical moment for OpenAI and Anthropic, both of which filed confidentially for IPOs in early June. Anthropic reported a $47 billion annualized run rate in May, up from roughly $10 billion for all of 2025. OpenAI's run rate was pacing near $25 billion earlier this year, up from $13.1 billion in 2025. But analysts warn these growth rates may represent a peak rather than a sustainable trajectory.

"Current growth rates for Anthropic and OpenAI are the fastest they will ever be, which is mostly a matter of basic math," said Gil Luria, an equity analyst at D.A. Davidson. "That is a good reason to go public now, as is the concern that some of their largest enterprise customers may start limiting their out-of-control token spend."

The rationalization phase begins

Jeff Henry, president of consulting at Highspring, said clients are pulling back until they "can really start to prove an ROI," with some waiting another 12 to 18 months before making major spending decisions.

The issue isn't just total spend—it's inefficient spend. Darren Kimura, CEO of enterprise AI company AISquared, said companies are overusing expensive frontier models for simple tasks that cheaper alternatives could handle. Model routing, which matches tasks to appropriate models, remains rare: Glean CEO Arvind Jain estimates that roughly 95% of enterprise AI usage still runs on frontier models.

Both OpenAI and Anthropic have responded by rolling out enterprise controls. OpenAI launched analytics and budget tools this month, while Anthropic introduced spending limits and usage provisioning in August. Expense management startup Ramp reports that CFOs are scrambling to manage a spending category that didn't exist in their annual plans.

New competition from deep pockets

The cost pressure coincides with intensifying competition from tech giants. Microsoft unveiled new low-cost models this month and emphasized model routing in GitHub Copilot. CEO Satya Nadella wrote in June that "the last thing any of us want is a world where every company across every sector is ceding value to a few models that eat everything they see."

Amazon's top AI executive, Peter DeSantis, told CNBC the company hopes to compete with frontier models within the coming year, leveraging in-house chips to undercut rivals on price. Google showcased Gemini 3.5 Flash at its developer conference last month, pricing it at half to one-third the cost of comparable frontier models.

"Microsoft and Google have the infrastructure and capability—the entire stack—where they can come in and stiff-arm both OpenAI and Anthropic," said PitchBook analyst Harrison Rolfes.

The New York Times reported Thursday that OpenAI is leaning toward delaying its IPO until next year, though neither company has provided a firm timeline. With traditional venture capital sources tapped out and their biggest backers now competing against them, the public markets may represent the best path to the capital both companies need.

These details were first reported by CNBC.