Enterprise

Enterprises Cut OpenAI and Anthropic Spending as AI Costs Spiral

Companies are switching to cheaper alternatives and imposing budget caps after burning through annual AI allocations in months.

Omega Editorial· June 26, 2026· 3 min read

Enterprise AI budgets hit the wall

Major enterprises are pulling back from premium AI providers OpenAI and Anthropic as unexpectedly high costs force finance teams to reconsider their generative AI strategies. The shift comes as companies discover their AI spending has outpaced projections by orders of magnitude, prompting some to switch providers entirely while others impose strict usage caps.

AI startup Lindy recently abandoned Anthropic's Claude models in favor of DeepSeek, a Chinese firm offering lower-cost, open-weight alternatives. CEO Flo Crivello told CNBC the move is expected to save millions of dollars over the coming months. "We did it, and you could see that cost curve go down, like, crash to the ground," Crivello said, adding that the decision was "a matter of survival for the business."

The cost crisis extends beyond startups. Uber introduced tiered caps on certain AI tools this month, setting a floor of $1,500 per month after the company depleted its entire annual AI budget in just four months. Jeff Henry, president of consulting at Highspring, reports that multiple clients have paused AI investment until they can demonstrate clear returns, with some unwilling to commit significant resources for another year to 18 months.

Why it matters

The spending crackdown arrives at a critical juncture for OpenAI and Anthropic, both of which filed confidential IPO documents in early June. Gil Luria, an equity analyst at D.A. Davidson, suggests the timing may reflect urgency to go public before enterprise budget rationalization creates headwinds. The shift also exposes a fundamental inefficiency in how companies deploy AI: Darren Kimura of AISquared notes that organizations routinely use expensive frontier models for routine tasks that cheaper alternatives could handle equally well. Glean CEO Arvind Jain estimates that roughly 95% of enterprise AI workloads still run on premium frontier models, leaving substantial room for cost optimization.

CFOs scramble for visibility

Eric Glyman, co-CEO of expense-management company Ramp, says the steep invoices have caught finance departments off guard. "Most CFOs not only didn't plan for this in their annual plans — the steep growth — but don't have great tools to manage this," Glyman said.

Model routing—directing each task to the least expensive model capable of completing it—has emerged as one potential solution, though adoption remains limited. In response to mounting pressure, both OpenAI and Anthropic have introduced new administrative features that give enterprise clients better visibility into consumption patterns and the ability to set usage caps. OpenAI CEO Sam Altman has acknowledged costs as "a huge issue" and the company is reportedly considering price cuts to its token-based services.

Neither OpenAI nor Anthropic provided comment when contacted. Anthropic declined to respond, while OpenAI did not reply.

These details were first reported by CNBC.

#enterprise ai#openai#anthropic#ai costs#deepseek#model routing

This is an original analysis by the Omega editorial team. Source reporting: AI Watch.

Want systems like this working for your business?

Book a Call

More in Enterprise

Enterprise· 3 min read

Cara Uses Amazon Bedrock to Automate Insurance Brokerage Workflows

The AI startup saves agents 10 hours per week by applying domain-specific LLMs to back-office tasks in the $8 trillion insurance sector.

Via AI Watch · Jun 26, 2026
Enterprise· 3 min read

OpenAI May Delay IPO to 2027 as AI Sector Faces Market Turbulence

Sam Altman reportedly seeks a $1 trillion valuation as rising costs and investor skepticism challenge the industry's recent momentum.

Via AI Watch · Jun 26, 2026
Enterprise· 3 min read

AI Content Governance Drives New Demand for DAM Platforms

Enterprise marketers struggle to control AI-generated assets as traditional automation fails to address brand consistency and compliance risks.

Via Automation Watch · Jun 26, 2026