75% of Execs See Tech Budgets Rising as Startups Flee Pricey AI Models

For startup founders, every dollar matters. When Uber—a company with deep pockets—blew through its annual AI budget in four months, it sent a warning: if giants are struggling, early-stage companies are on the brink. The move to smaller, cheaper models offers a lifeline, but also forces a rethink of product roadmaps that assumed unlimited AI access.

The unchecked escalation of enterprise AI costs is forcing a fundamental shift in how businesses select and deploy artificial intelligence models. For years, companies treated AI usage as a direct proxy for productivity—a mindset insiders dubbed 'tokenmaxxing'—and premium models like OpenAI’s GPT-4 were the default choice. Now, soaring and unpredictable bills are compelling even the most resource-rich organizations to pivot toward smaller, cheaper, and often open-source alternatives. This is not a marginal optimization; it represents a structural market correction that will redefine the AI value chain, from infrastructure providers to end-user applications.

“

Microsoft CEO Satya Nadella, Palo Alto Networks’ Nikesh Arora, and Coinbase’s Brian Armstrong have all publicly argued that smaller models can handle a large share of corporate needs.

At the heart of the disruption is a pricing model transformation. Major AI vendors have been migrating from flat subscription fees to consumption-based, per-token billing. While headline token prices have declined, the total cost of completing a task is climbing. Tasks now require more steps, longer prompts, and multi-turn reasoning, driving up token consumption per interaction. Companies accustomed to predictable SaaS billing cycles are suddenly confronted with volatile expenses. Uber’s experience is emblematic: the company burned through its entire 2026 AI budget in just four months after employees enthusiastically adopted AI coding tools, forcing management to impose usage caps. If a cash-rich corporation like Uber can overshoot so dramatically, the alarm is even louder for mid-market firms and startups operating on lean budgets.

The financial strain is registering in boardrooms. Gartner’s latest survey found that three-quarters of technology executives anticipate higher budgets this year, with nearly half projecting double-digit increases. AI is the primary accelerant. By 2028, Gartner estimates that the cost of AI coding tools alone could surpass the average developer salary, a tipping point that would force a wholesale reconstruction of software development economics. Such projections are pushing businesses to embrace routing solutions and model marketplaces like OpenRouter, which allow them to assign each task to the most cost-effective model. Complex coding or reasoning may still require a frontier model, but classification, summarization, and simple Q&A can be offloaded to fine-tuned open-source systems at a fraction of the cost.

This recalibration has broad support from influential technology leaders. Microsoft CEO Satya Nadella, Palo Alto Networks’ Nikesh Arora, and Coinbase’s Brian Armstrong have all publicly argued that smaller models can handle a large share of corporate needs. Their endorsement signals that the shift is not merely a cost-cutting fad but an architectural decision that could accelerate AI democratization. Open-source token volumes on platforms like OpenRouter are surging, indicating that developers are voting with their workloads. Harold Byun, CEO of BlueRock, a startup focused on safe AI deployment, noted that the billing model change 'caught a lot of people by surprise,' underscoring that many organizations did not anticipate the financial impact of their AI experimentation.