- AI affordability is becoming a top enterprise priority as agentic AI workloads drive token usage — and costs — far beyond early expectations
- Hyperscalers including Microsoft, Google and others are racing to prove enterprise AI can scale responsibly, with cheaper models and cost controls
- Enterprise AI spending is still set to surge in 2026, but Gartner warns poor cost visibility could turn rapid adoption into wasted budget
Tokenmaxxing was all the rage just a few months ago, but cooler heads are starting to prevail and AI affordability is now the name of the game.
“We are not yet in a full AI price war, but competitors to Microsoft, including OpenAI, Anthropic and Google, are pushing a more cost-effective approach to heavy AI usage with their own cost controls, particularly as agentic AI capabilities proliferate,” J. Gold Associates Founder and Principal Analyst Jack Gold wrote in a note to investors.
The hyperscalers are responding to apparent strain on enterprises, which are increasingly grappling with a sudden deluge of AI costs as AI workloads and agents begin to proliferate. A lot of the problem has to do with the rise of agentic operations. These workloads happen at machine speed and each AI agent can spin up additional agents, that in turn can generate “tens to hundreds of thousands more tokens per transaction, and do that hundreds to thousands of times per day,” Gold wrote.
This, of course, means the cost of agents can quickly outrun the cost of employees, he noted.
Gartner recently predicted that spending on AI models and platforms will jump 63.4% year on year to $64 billion in 2026. The biggest growth (210%) is expected in domain-specific models and other specialized generative AI models. That growth is directly tied to enterprise priorities.
“Enterprise AI budgets are coming under greater scrutiny, with increased focus on usage efficiency, cost control and measurable outcomes,” said Arunasree Cheparthi, Senior Principal Research Analyst at Gartner.
But AI isn’t the only line item in the budget.
Gartner Distinguished VP Analyst John-David Lovelock in July pointed out that inflation, supply shortages, rising hardware and component costs, AI initiatives and shifting priorities are all straining enterprise technology budgets.
Hyperscalers hop into action
The growing enterprise cost problem – which also encompasses issues around appropriate model selection and content ingestion – has rapidly become a problem for hyperscalers, which are counting on business AI adoption. This tension was readily apparent on recent earnings calls, with executives highlighting low-cost and cost-efficient model options.
Alphabet CEO Sundar Pichai said on the company’s Q2 2026 earnings call it is seeing “tons of demand” for its low-cost Flash model family, adding “It's very important to us to have the best frontier models out there, as well as models which are very performant and low cost.”
But Gold noted that hyperscalers are starting to offer tools beyond the models themselves to rein in costs.
Microsoft, which has a large enterprise base, just rolled out new controls via its Cost Management Dashboard, which Gold noted allows users to track usage at the individual level, implement policy-based limits, set alerts and leverage tiered billing options.
“Competitive assessments by organizations of which and/or whether to acquire agentic tools now often centers on costs, and not just the fit within traditional application installations,” Gold wrote.
Indeed, the cost question is poised to become even more important in the months and years ahead.
Gartner this week revealed that business leaders dedicated around 12% of their functional budgets to AI in 2025 and 85% said they plan to increase spending in 2026. At the same time, 11% of organizations said they were completely unaware of what their unit spend on AI was in 2025.
“This lack of financial visibility heightens risk as spending accelerates,” said Gartner Distinguished VP Tina Nunno in a statement. “Without disciplined measurement tied directly to business outcomes, organizations risk wasted resources and unmet expectations.”
Read more about enterprise AI here:
Enterprises risk AI sticker shock as token costs pile up
Together AI expects IBM-backed AI capacity to sell out before launch
Enterprises think they’re ready for AI – they’re not