``` Tokens can’t be burned anymore? This might be the most important chart in the entire market. ```

```

Token expenditure growth is showing signs of fatigue, and the market’s core focus on AI is quickly shifting from "whether the technology is feasible" to "whether the costs are bearable."

On June 9, macro strategist Andreas Steno Larsen stated on social media that the Silicon Data LLM Token Expenditure Index is currently the most important chart for the entire market to watch.

This index has more than doubled since last December and rose sharply before May 2026, but recently has declined. Andreas Steno Larsen warned that if Token pricing continues to weaken, this cycle's transactions—from memory to wider hardware and data centers—may come to an end.

At the same time, tech giants are urgently trying to rein in out-of-control internal AI computing power consumption.

Wall Street News previously mentioned, Amazon and Microsoft are cutting back on internal AI tools or stopping projects that track usage to combat staff "Tokenmaxxing"—wasteful compute consumption for the sake of improving internal rankings. On the service side, GitHub Copilot switched from request-based billing to Token-based billing on June 1, causing some users' monthly bills to soar tenfold or more, sparking widespread skepticism about the sustainability of AI subsidy models.

These signals are reshaping investors’ perception of risk in AI infrastructure deals. Marginal changes in Token expenditure transmit through chains of GPU power, DRAM memory, and data center demand, directly affecting capital spending forecasts by Nvidia, memory chip manufacturers, and cloud providers.

Index Peaks: Hardware Transaction Logic Faces a Test

The Silicon Data LLM Token Expenditure Index is a weighted indicator measuring the price paid per million LLM Tokens in the market, seen as a proxy for willingness to pay at the margin for AI. Since major providers like OpenAI, Anthropic, and Google charge clients based on Token consumption, Token expenditure directly links AI usage to demand for GPU, DRAM, and data centers.

The index’s recent stagnation has triggered concerns about the hardware cycle in the capital markets. Silicon Data commented that the recent decline may signal a slowdown in migration toward high-end closed-source models. If Token spending continues to weaken, the marginal revenue available to fund incremental GPU, DRAM, and data center procurement will dwindle, altering the risk profile of companies whose capital spending plans are tied to Token-driven growth.

While a single drop does not constitute a trend, as a leading indicator for hardware cycles, this data suggests enterprises’ dependence on costly cutting-edge models may be systematically declining.

Billing Crisis: Tech Giants Halt "Ineffective Consumption"

The enterprise AI boom is facing its first real billing crisis.

According to Axios, citing an AI consultant, one of their corporate clients recently spent $500 million on Claude in a single month simply because employee usage had no cap.

Internally, using AI usage as a performance metric has backfired. Reports indicate Amazon’s developer platform Kiro once had an internal leaderboard, "Kirorank." Meta also experienced similar situations, with staff attempting to boost Token consumption to climb the rankings.

Amazon SVP Dave Treadwell admitted that employees executing meaningless AI tasks to game the leaderboard drove up operating costs. He explicitly instructed staff "not to use AI for the sake of using AI," after which the beta dashboard was taken offline. Amazon now tracks AI-generated code’s actual value with a "normalized deployment" metric instead of Token consumption.

Pricing Rebound: Subsidy Era Nears End

On the supply side, the long-standing "growth through subsidies" business model in AI is nearing its limit.

On June 1, GitHub Copilot officially switched to Token-based usage billing. Users on Reddit reported their monthly bills are expected to jump from under $45 to more than $847.

GitHub CPO Mario Rodriguez previously stated that with the rise of agentic AI, old pricing models are unsustainable. Gartner analyst Arun Chandrasekaran, in an interview with Business Insider, noted that as advanced reasoning models push up compute consumption, more companies will adopt usage-based billing.

Investor Tommy Shaughnessy warned of the systemic risk of such subsidy models. He pointed out that major AI companies currently have deeply negative profit margins, and once enterprises face true usage-based pricing, actual consumption will far exceed expectations—for example, Uber is projected to spend its entire 2026 AI budget in just four months. If investors lose confidence in expected returns, capital flows supporting GPU purchases and model training will reverse.

Cost Restructuring: Cheap Models May Dominate

Facing soaring inference costs, the market is seeking low-cost alternatives.

Goldman Sachs One-Delta head Rich Privorotsky believes that with DeepSeek dropping prices by 75% and Xiaomi MiMo by nearly 99%, easing infrastructure bottlenecks is triggering a price war.

Wall Street News previously mentioned,Coinbase CEO Brian Armstrong predicts that 80% of AI workloads will migrate to models that are 99% cheaper within 12 to 18 months, leaving only 20% of cutting-edge tasks on frontier models. He noted that energy and computing power will be the true bottlenecks.

Hugging Face CEO Clement Delangue, citing Stanford data, confirmed this trend: local models have achieved 71.3% accuracy on real-world queries at very low cost. Micro1 CEO Ali Ansari called this a "healthy swing" from overuse to rational use.

There is a deep divide on Wall Street about real investment returns from AI. According to Goldman’s Jim Schneider, agentic AI will drive Token consumption up 24-fold by 2030, and cloud providers’ gross margins will turn positive in the short term. JPMorgan’s economic research shows that explosive growth in Python packages on PyPI proves genuine productivity gains.

However, the bearish camp is equally adamant. Goldman semiconductor analyst Jim Covello reported that the current industry boom comes at the expense of upstream consumption, with nearly all the value flowing to semiconductor firms—a situation he sees as unsustainable.

Boosted.ai CEO Josh Pantony stressed that enterprise concerns over data openness are undermining the efficacy of AI agents. Under the multiple considerations of cost, returns, and security, how much real value the next AI bill generates will be the ultimate verdict of the market on this technological investment.

Risk warnings and disclaimerThe market has risks, and investment must be cautious. This article does not constitute personal investment advice and does not take into account individual users’ special investment goals, financial circumstances, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article suit their specific situation. Invest accordingly at your own risk. ```