Enterprise AI Bills Surge as Usage-Based Pricing Replaces Subscriptions

Enterprise AI spending is spiraling beyond initial projections as technology providers shift from predictable subscription models to usage-based pricing, leaving companies scrambling to control costs and reassess their AI strategies.

The transition has caught businesses off guard. Uber exhausted its entire 2026 AI budget within four months after employees rapidly adopted AI coding tools, forcing management to impose usage caps, according to reports first published by USA Today. Harold Byun, CEO of BlueRock, a startup focused on safe AI system deployment, noted that the licensing model change surprised many organizations, with customers reporting 20% to 30% budget overruns immediately following the shift.

The cost equation changes

While token prices—the units measuring AI usage—have declined, the actual expense of completing tasks has increased. Modern AI workflows now involve more processing steps, larger datasets, and longer input sequences than early adopters anticipated. Gartner projects that AI coding costs will exceed the average developer's salary by 2028, and three-quarters of executives surveyed expect technology budgets to rise this year, with nearly half forecasting double-digit increases.

This financial pressure is reshaping how companies deploy AI resources. Businesses are increasingly adopting routing tools like OpenRouter, an AI marketplace that assigns tasks to the most cost-effective model while reserving premium systems for complex work. Open-source tokens processed through OpenRouter jumped from 34% in January to 65% in June, according to Citi analysis.

Why it matters

The enterprise AI market is reaching an inflection point where cost predictability matters as much as capability. This shift threatens the revenue models of leading AI providers preparing for public offerings while creating opportunities for open-source alternatives. The dynamic could fundamentally alter which companies dominate enterprise AI adoption and reshape competitive positioning across the sector.

Chinese models gain ground

Chinese AI models are narrowing the performance gap with U.S. counterparts while offering dramatically lower pricing—as little as 18 cents per million tokens compared to an average $4 for top-tier models. The four most popular models on OpenRouter are all Chinese, with DeepSeek holding the leading position. BlueRock's Byun estimates that open-source models, once more than a year behind cutting-edge systems, now trail by roughly four months.

Security concerns remain a barrier to widespread enterprise adoption of Chinese models, particularly in sensitive sectors like cybersecurity. Industry executives including Microsoft's Satya Nadella, Palo Alto Networks' Nikesh Arora, and Coinbase's Brian Armstrong have publicly stated that smaller, less expensive models can satisfy most corporate requirements.

Palo Alto Networks' Arora urged AI laboratories on X to adopt "forward pricing" that charges customers today at the lower rates expected in future years. OpenAI is reportedly considering significant price reductions, including token usage cuts, anticipating similar moves from competitor Anthropic.

Val Bercovici, chief AI officer at WEKA, which optimizes AI model performance, summarized the emerging consensus: open-source models deliver "90% as good at 10% of the price," eliminating the need for premium tokens on routine tasks.

Christopher Brown, a financial adviser at Synovus Securities, noted that price competition between OpenAI and Anthropic will intensify as both companies position for initial public offerings, though the shift toward cheaper models could constrain their revenue growth.

These details were first reported by USA Today, with reporting by Aditya Soni.