Bottom line: Corporate AI spending has spiraled out of control; OpenAI promises more efficient models, while the Jevons Paradox could drive renewed demand growth over the long term.
OpenAI CEO Sam Altman has confirmed that enterprise customers increasingly view exploding AI token costs as a business problem — and has pledged to optimize model efficiency. The phenomenon of “tokenmaxxing” has inflated budgets without delivering proportional added value.
At the “Intelligence at Work” event, Altman acknowledged that growing AI spending among enterprise customers has now become a central topic of discussion. This is the first time customers have raised the issue proactively. Altman cited a recurring scenario: “My company burned through the entire 2026 budget in the first quarter.” OpenAI is now working to deliver customers more performance per dollar spent, though Altman offered no concrete measures.
The practice of “tokenmaxxing” has contributed to the problem: companies increase their AI token consumption expecting to generate productivity gains and justify high spending. Nvidia CEO Jensen Huang publicly called on developers to consume tokens worth at least half their annual salary. An extreme example: startup OpenClaw consumed 603 billion tokens in one month and paid $1.3 million for it — without demonstrating clear business value.
Meanwhile, it is becoming clear that the hoped-for benefits often fail to materialize. At Amazon, AI agents were apparently deployed to rank well on internal leaderboards rather than for genuine added value. Microsoft has scaled back its Claude Code licenses due to costs. A structural problem is exacerbating the cost situation: agent-based AI systems, in which models autonomously handle multi-step tasks, consume many times more tokens than simple queries.
Altman remains optimistic in the long term. He relies on the Jevons Paradox — the economic observation that cheaper resources lead to disproportionately increasing demand. Altman illustrates this with internal OpenAI data: 6.5 years ago, the highest user consumed 100,000 tokens monthly. Today, the global average per employee sits exactly there. The current top user at OpenAI consumes 100 billion tokens monthly. If this trend continues, the global average could eventually reach this order of magnitude as well.
Source: www.it-daily.net · Published June 5, 2026
Lumi AI News — AI-assisted curation according to Art. 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.6.0.