Companies Slam Brakes on AI Spending After Skyrocketing Token Costs

Corporate America has discovered that unlimited AI access comes with a price tag nobody anticipated. After months of encouraging employees to use artificial intelligence tools without restraint, companies now face unexpected bills that have forced executives to impose strict spending limits and question whether the technology delivers sufficient value.

The shift marks a dramatic reversal from earlier enthusiasm. Coinbase executive Rob Witoff, who oversees the crypto exchange’s infrastructure, watched usage explode after Anthropic’s Claude launched its improved coding model Opus 4.6 in February. “Our internal usage started to go parabolic across the company,” he explained. The company now operates a sophisticated system of weekly price caps ranging from $500 to $5,000 based on each employee’s job level and role, representing a middle ground between unlimited access and complete restriction.

Witoff captured the evolution in corporate AI strategy succinctly. “Once people understand what’s possible, usage takes off on its own. Then the focus shifts from ‘Are people using AI?’ to ‘Are they using it well?’ That’s where we are today,” he said. The question has shifted from adoption to optimization as finance departments grapple with costs they cannot accurately predict.

The Token Economy Creates Budget Chaos

Companies struggle because AI pricing operates fundamentally differently from traditional software licensing. The basic unit of AI use – the token – represents a fragment of text, code, or data processed by a model when it reads a prompt or produces an answer. Tokens do not map neatly onto single words, and some can be cached by AI models so they are not charged again while others must be processed as new. The result creates uncertainty that often becomes clear only when the bill lands at month’s end.

Only one in four companies say they have a comprehensive view of what artificial intelligence costs them, according to an unreleased KPMG survey reported by The Wall Street Journal. About half have some visibility into AI use costs. One in five have no visibility, or only see the damage once the bill arrives. “It’s a new resource that needs to be managed that didn’t exist quite that way, and we’re seeing exponential growth,” Steve Chase, KPMG’s global head of AI, told the Journal.

The firm works with companies that have blown through annual token and cloud-computing budgets in months, Chase reported. Sam Ransbotham, professor of analytics at Boston College’s Carroll School of Management, confirmed the widespread frustration. “People are getting these massive bills,” he observed.

Pricing Models Shift as Companies Retreat

Executives and developers now rethink their recent approach by switching models, imposing limits, and prioritizing projects. AI giants including OpenAI and Anthropic risk losing market share to cheaper models at a time when they race toward stock-market-altering IPOs. The high stakes prompt unprecedented scrutiny across industries as companies attempt to balance innovation with fiscal responsibility.

Salesforce CTO Parker Harris told Business Insider that his company has fully opened the floodgates for spending on Anthropic tools, but that likely will not last forever. They must find a balance that does not divert too much money to the rising startup. “We gotta run a business, we’re a public company,” Harris said. “We can’t tell our investors like, ‘Yeah, sorry, we gave half of our upside this year to Anthropic so they can go public’.”

The corporate restraint represents a significant shift in how businesses approach artificial intelligence adoption. Companies that encouraged employees to “tokenmaxx to their hearts’ content” now discover that freedom carries financial consequences they cannot ignore. Without comprehensive tracking systems in place, organizations face difficulty understanding where money goes and which use cases deliver genuine value versus which merely generate impressive-looking outputs that fail to improve productivity.

CFOs Demand Accountability From AI Investments

Chief financial officers increasingly demand justification for AI expenditures that previously received automatic approval. The technology’s novelty initially shielded it from normal budget scrutiny, but that protective shield has dissolved as costs mount. KPMG data reveals the scale of the problem: organizations routinely deplete budgets intended to last twelve months in just a few quarters, forcing mid-year recalibrations that disrupt planned initiatives.

The situation forces companies to ask uncomfortable questions about AI’s actual return on investment. Many organizations adopted the technology driven by fear of missing out rather than clear business cases. Measuring whether increased AI usage translates into meaningful productivity gains or competitive advantages proves difficult when the tools integrate into workflows in ways that make their specific contributions hard to isolate and quantify.

Coinbase’s approach offers a template other organizations may follow. By setting role-based spending limits, the company maintains access to AI tools while preventing runaway costs. The system acknowledges that different positions require different levels of AI assistance – engineers building products need more capacity than administrative staff handling routine tasks.

The Path Forward for Enterprise AI

Companies face a delicate balancing act moving forward. Restricting AI access too severely risks losing productivity gains and falling behind competitors who manage the technology more effectively. Maintaining unlimited access, however, creates unsustainable cost structures that drain resources from other strategic priorities. The solution likely involves more sophisticated usage monitoring, clearer guidelines about appropriate use cases, and ongoing education about effective prompting techniques that maximize value per token consumed.

Organizations that navigate this transition successfully will likely emerge with more sustainable approaches to AI integration. They will understand which tasks genuinely benefit from artificial intelligence assistance and which represent expensive distractions. This maturation process mirrors earlier technology adoption cycles where initial enthusiasm gave way to measured implementation focused on demonstrable business outcomes rather than technological novelty.

The current adjustment period, while painful for companies facing unexpected bills, represents a necessary evolution in enterprise AI adoption. As the technology moves from experimental toy to essential tool, normal financial discipline must apply. The companies that survive this adjustment will develop frameworks that allow them to harness AI’s capabilities without surrendering budget control to unpredictable token consumption patterns that make long-term planning impossible.