Observed Signal · May 29, 2026 · Analysis / Opinion · Source: Gary Marcus · Impact: 2/5 · Sentiment: Neutral
After the Decline of Tokenmaxxing: What's Next?
Gary Marcus published an essay on May 29, 2026, arguing that signs point to a decline in aggressive token-driven AI usage (“tokenmaxxing”). He cites converging evidence: a reported drop in Nvidia H200 rental prices, a Financial Times report that Amazon removed an internal AI usage leaderboard, and commentary from other writers/researchers (Fortune’s Jeremy Kahn and AI researcher Lisan al Gaib). Marcus contrasts his own, more cautious predictions with a more optimistic forecast from Lisan al Gaib and notes the pair agreed to reassess outcomes in 12 months. The piece is an analytical commentary on rising inference costs, shifting ROI expectations for hyperscaler AI investments, and changing signals in AI adoption behavior.
Analysis highlights potential shifts in AI inference economics (GPU rental prices and internal usage controls) that could affect enterprise adoption, model providers, and AI infrastructure spending — relevant but not an industry-altering policy or major platform technical change.
Track Amazon Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Article published on Substack on 2026-05-29.
- Thierry (handle @ThierryBorgeat) posted that the rental price for an Nvidia H200 fell from $7/hr to $4/hr in three weeks (≈ -40%).
- Financial Times reported Amazon removed an internal AI leaderboard to prevent gaming of usage metrics.
- Fortune writer Jeremy Kahn and AI researcher/benchmark creator Lisan al Gaib are cited as commenting on the perceived decline of tokenmaxxing.
- Gary Marcus and Lisan al Gaib agreed to check back in 12 months to evaluate their contrasting predictions.
Connected Companies & Entities
5 Entities mappedOntology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Tokenmaxxing Fade Threatens AI Revenue Boom
Enterprises rapidly spent budgets on agentic coding agents in early 2026, but many are now questioning the return on that token-intensive spending. The piece highlights Salesforce and Uber as major investors in agentic coding; Salesforce reportedly underestimated its initial token budget. The shift away from “tokenmaxxing” — heavy, unoptimized per-token model usage — is creating a broader industry debate about how to measure ROI and whether sustained demand for high-volume model usage will persist. The trend has raised concerns for model providers' revenue growth and has prompted engineering teams to re-evaluate agent deployment and cost controls. (Published 2026-05-28.)
Warnings That OpenAI Could Trigger an AI Market Crash
Bloomberg reports SoftBank recently had difficulty securing a margin loan of about $6 billion backed by its OpenAI stake after scaling back an earlier $10 billion target; talks stalled and SoftBank shares fell nearly 10%. The story reinforces warnings (e.g., Gary Marcus) that lenders’ reluctance to underwrite debt against OpenAI equity and other market pressures could increase the risk of an AI‑sector correction. SoftBank has reportedly committed roughly $60 billion to OpenAI, faces about $40 billion of bridge refinancing due by March 2027, and may rely on potential IPOs for OpenAI or Anthropic to monetise holdings. AllianceBernstein’s Hua Cheng described the margin loan as one part of a broader financing puzzle.
Firms Pull Back on Costly 'Tokenmaxxing' Trend
Companies are rolling back the practice known as "tokenmaxxing"—aggressively increasing AI token consumption without proportional productivity gains—after reports revealed extremely high internal usage and bills. Sources say Meta halted an internal token-consumption leaderboard after The Information reported about ~60 trillion tokens used in 30 days; Amazon and Microsoft have also restricted internal competitions or access patterns. Examples include Openclaw founder Peter Steinberger reportedly spending about $1.3 million in 30 days (costs covered by OpenAI) and Uber exhausting its annual AI token budget within four months of 2026. Industry observers predict a shift toward "token-minimization" and stricter internal limits as firms seek better ROI and cost controls for LLM usage.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.