Artificial intelligence is costing large companies more than many expected, pushing finance leaders into an unfamiliar choice between AI compute spend and future hiring, according to CNBC (Deirdre Bosa and Jasmine Wu; May 29, 2026).
Interviews with enterprise AI executives Arvind Jain of Glean and Matan Grinberg of Factory AI describe Fortune 500 buyers burning through annual AI budgets in weeks, overusing premium models, and entering a new phase of cost discipline—raising questions about whether the AI investment boom can sustain current valuations if price sensitivity rises.
Annual Budgets Exhausted in One or Two Months
Jain told CNBC that overspending on AI is the top enterprise concern right now.
“Companies are telling us that their AI budgets are getting exhausted in one month or two months, and these are annual budgets,” he said.
Costs have not fallen as buyers hoped. Each new frontier model release is roughly twice as expensive per token as its predecessor, Jain said—putting enterprise AI on what he called “an unsustainable path right now.”
For the first time in his experience, technology spend is comparable to people costs: “choose tech or people.” Historically, software was a small fraction of operating expense; now AI budgets are increasingly coming instead of planned headcount growth.
Corporate Finance & AI Capital Allocation Careers 2026
With enterprise tech costs matching human salaries and driving CFOs to actively trade headcount budgets for AI token allocations, the corporate demand for Corporate FinOps Directors and Strategic Capital Allocators is surging. Join the leaders of Multi-Jurisdictional Corporate Resource Optimization & Capital Strategy, Asymmetric Token FinOps Governance & Model Routing Architecture, and Hybrid Workforce Financial Engineering & Margin Leakage Mitigation.
Explore High-Stakes Roles →Three Phases: Mandate, Tokenmaxxing, Reckoning
Grinberg described how leadership teams moved through three stages in roughly a year:
- Phase 1: Boards press CEOs to “do something” about AI.
- Phase 2: Tokenmaxxing—using frontier models aggressively regardless of cost.
- Phase 3: Reassessing whether premium intelligence is needed for every task.
“Companies say, hey, if we could optimize one thing, is it the number of employees that we have, or is it the AI spend per employee?” Grinberg said.
His question for teams: “Do we need to be using Opus-level intelligence for every single task? You just don’t need to.”
Powerful but Inefficient—and Often Over-Provisioned
Jain said AI today is “very powerful, but very inefficient.” Value delivered still trails what businesses pay.
A major driver: roughly 95% of enterprise AI usage still runs on the most expensive frontier models—even for work cheaper tiers could handle, Jain said.
The fix he highlighted is model routing—sending simple jobs to lower-cost models. With the right routing up front, Jain said companies can achieve roughly 10x savings.
That is also the pitch behind Factory AI, which routes engineering tasks across frontier models automatically. Grinberg compared incremental upgrades between top-tier releases to two seasoned professors: “Opus 4.7 versus Opus 4.8 is like the difference between a professor who’s been a professor for 13 years versus 15 years.” For most users, the gap is hard to detect.
What This Means for the Broader AI Trade
Public markets have rewarded AI infrastructure and chip names as demand soars—CNBC noted record highs and trillion-dollar valuations in the sector. But inside Fortune 500 budget meetings, the story is tightening.
The bull case assumes demand stays enormous and buyers stay relatively indifferent to unit cost. Enterprise accounts suggest the opposite may be emerging: buyers are becoming price-sensitive, rebalancing tokens against humans, and questioning premium model usage at scale.
CNBC linked the trend to longer-term questions for OpenAI and Anthropic, whose businesses rely heavily on premium pricing—if enterprises optimize spend, growth and margin assumptions tied to unconstrained token consumption could face pressure.
Hiring the Architects of AI Observability & Enterprise FinTech?
As enterprise CFOs re-engineer corporate balance sheets, trading traditional human headcount budgets for metered AI token consumption and deploying real-time financial observability layers to protect operating margins, securing elite talent fluent in advanced algorithmic FP&A, cloud-compute cost optimization, and high-velocity technology procurement is the ultimate competitive edge. Find the experts building a resilient future without the “click tax.”
Bottom Line
Enterprise AI is moving from experimentation to economics. CFOs are no longer asking only whether AI works—they are asking whether each token is worth more than the next hire. Smarter routing and tiered model use offer immediate savings, but the deeper shift is strategic: AI budgets are now competing directly with headcount plans, and that trade-off may define the next phase of corporate AI adoption.
Frequently Asked Questions
Q: What does “tokens or humans” mean for companies?
A: CFOs are weighing whether to spend on AI tokens/compute or preserve budget for future headcount—a trade-off executives say is new because AI costs now rival people costs in some workflows.
Q: Why are AI budgets running out so fast?
A: CNBC cited rising per-token prices on new frontier models (roughly 2x per release), heavy use of premium models for simple tasks (~95% of usage), and aggressive early adoption (“tokenmaxxing”).
Q: How can enterprises cut AI spend without stopping adoption?
A: Leaders interviewed pointed to model routing—matching task complexity to the cheapest capable model—with potential savings around 10x when routing is done systematically.
