Tech firms ration AI as costs outpace returns
Amazon shut down its AI leaderboard after employees gamed it to inflate usage. Microsoft, Meta and Uber have also cut back as token consumption and chip costs spiral beyond projections.
May 30th 2026 · World
Major technology companies are scaling back their AI ambitions as costs spiral beyond initial projections. Amazon has shut down its internal AI leaderboard called Kirorank, which encouraged employees to use AI more frequently, after discovering that workers were engaging in "tokenmaxxing" - making AI complete menial tasks to inflate token usage for leaderboard rankings. The company joins Meta, which forcibly closed its employee AI leaderboard in April after similar gaming issues, and Uber, which burned through its entire 2026 AI budget in just one quarter. Microsoft canceled most of its Claude Code licenses in early May, while Salesforce, DoorDash, and other major firms have moved from aggressive AI adoption to rationing the technology amid soaring costs with disappointing returns. The cost crisis stems from exploding token consumption, with Google reporting that Gemini token usage jumped from 480 trillion tokens per month in May 2025 to 3.2 quadrillion tokens per month in May 2026. Goldman Sachs forecasts a 24-fold increase in token consumption by 2030, reaching 120 quadrillion tokens monthly, driven by agentic AI systems that burn far more compute than basic chatbot interactions. Chip costs are compounding the problem, with a single Nvidia Blackwell GPU in a modern AI cluster potentially costing as much as a new Tesla Model 3. Supply cannot keep pace with demand, as new chip factories cost tens of billions of dollars and take years to build, while production lines are increasingly diverted toward more lucrative AI chips, creating shortages even for non-AI applications. The economic implications extend beyond individual company decisions. High chip costs are raising prices for downstream technology, consumer goods, and automotive products, echoing the inflation spikes of the Covid-era chip shortage. Startups and smaller companies face increasing difficulty acquiring chips, reducing competition and innovation. AI companies now represent a significant share of market capitalization and capital expenditure, creating systemic vulnerability. Industry observers note that companies like Microsoft, OpenAI, Google, and Anthropic have entered circular cross-investment agreements to fund chip spending, with substantial debt financing backed by chip assets as collateral. Gartner's finding that cheaper tokens will not translate into cheaper enterprise AI makes this inequity structural rather than temporary, suggesting that productivity gains from the agentic AI era may accrue only to organizations large enough to absorb escalating compute costs.