AI Sticker Shock: Why Companies Are Reeling from Rising Costs

Remember when AI was the tech darling everyone was willing to splurge on? Those days seem numbered. As we hit 2026, companies are waking up to a new reality: AI isn't the cheap buffet it once was. Sticker shock is settling in as big names like Coinbase and Deloitte face ballooning AI expenses. Why now? Because the pricing models that once allowed tech firms to gorge themselves on AI have evolved, leaving businesses to count every dollar.

Coinbase's AI Roller Coaster

Take Coinbase. Their internal use of AI spiked after the release of Anthropic's improved model, Opus 4.6, in February. But with costs rising, Coinbase had to introduce price caps, ranging from $500 to $5,000, based on employee roles. Rob Witoff, who's in charge of infrastructure at the crypto exchange, noted the shift from "Are people using AI?" to "Are they using it well?" Looks like the days of AI excess are fading fast.

So, who benefits from this change? Companies that adapt quickly and learn to optimize their AI usage. But there's a catch. As firms impose limits, giants like OpenAI and Anthropic risk losing business to cheaper alternatives at a critical time. They're both eyeing stock-market-shaking IPOs, after all.

Token Economics: The New Normal

Between February and June, OpenAI, Anthropic, and GitHub shifted to token-based pricing. This change has prompted companies like Walmart and Amazon to curb their AI usage. Meanwhile, Accenture and others are backing a "Tokenomics Foundation" to standardize AI budgeting metrics. The real question is: Will this shift drive efficiency or stifle innovation?

Companies are realizing that throwing money at AI without a clear ROI is a recipe for disaster. A survey by Wakefield Research found that 79% of executives worry their AI budgets could be slashed if spending doesn't yield new revenue.

The End of AI Subsidies

OpenAI's Sam Altman acknowledges the rising concern. AI budgeting wasn't even on the radar earlier this year but has now become a hot topic. The reason for these price hikes? Tokens have become cheaper thanks to Nvidia's chip innovations, but AI's popularity means providers can't subsidize heavy users anymore.

It's ironic. The very tech that's supposed to drive efficiency is prompting companies to reevaluate their priorities. Ahmad Awais of Command Code saw 10,000 new customers in 30 days, driven by demand for cheaper models. It turns out, not every task requires a Ferrari of AI models.

Ultimately, this is a story about power, not just performance. As companies navigate these choppy waters, one thing's clear: the era of AI as an all-you-can-eat buffet is over. The benchmark doesn't capture what matters most. The question of "Whose data? Whose benefit?" is more relevant than ever.