A Key Signal for the AI Boom Is Quietly Flashing Red

The Metric Wall Street Has Been Watching Is Turning

For the past year and a half, one of the most closely watched gauges of the artificial intelligence boom has been quietly ticking in the background — not a stock price, not a quarterly earnings report, but the price the market pays for a single unit of AI output. Now, that gauge is heading in the wrong direction, and investors are paying close attention.

The Silicon Data LLM Token Expenditure Index, which tracks what users pay for AI tokens, is down almost 20% from a high in May after nearly doubling since its inception in December. The gauge is considered the cleanest read anyone has on the $700 billion-plus capital expenditure boom that has done the sector's heavy lifting. For markets already growing nervous about whether that tidal wave of spending will ever translate into real profits, a declining index is an uncomfortable signal.

What Tokens Have to Do With Trillions of Dollars

To understand why this matters, it helps to know what a "token" actually is. Every time someone uses an AI chatbot or AI-powered tool, the underlying model processes text in small units called tokens — roughly three-quarters of a word each. The Silicon Data index is a spending-weighted metric that measures the average price paid per million tokens across the market and serves as a proxy for the market's marginal willingness to pay for AI. Since major providers such as OpenAI, Anthropic, and Google predominantly charge customers based on token consumption, token expenditure directly links AI usage to demand for GPUs, DRAM memory, and data centers.

The marginal change in token expenditure directly impacts the capital expenditure expectations of Nvidia, memory chip manufacturers, and cloud service providers through the transmission chain of GPU computing power, DRAM memory, and data center demand. In other words, when the price of a token dips, the financial logic underpinning hundreds of billions in AI infrastructure investment starts to look shakier.

A Price War Is Already Underway

For stock investors, that could be flashing a warning that AI companies are losing pricing power with increasingly cost-sensitive customers, and that expectations for an eventual AI bonanza could prove misplaced. The problem lies in the fact that the supply of computing power is growing faster than the ability of enterprises to create applications that justify high subscription prices.

Rich Privorotsky, head of Goldman Sachs' One-Delta division, believes that price wars are being triggered by easing infrastructure bottlenecks, citing DeepSeek's 75% price cut and Xiaomi's MiMo model's nearly 99% reduction. Open-source models like Llama are also squeezing the margins of closed-source competitors. The result is a market in which more AI is being consumed than ever before, but the revenue generated per unit of that consumption is shrinking.

Not a Crash — But a Clear Warning

A softer index doesn't necessarily mean AI is getting cheaper in a straightforward way. The gauge blends prices and usage, meaning a dip can imply very different scenarios: either list prices are falling, or demand is shifting toward cheaper models — or it could point to a genuine softening in what buyers are prepared to pay. Comments from Silicon Data suggest the recent pullback may signal a slowdown in the pace of migration toward high-end closed-source models.

Token prices have collapsed more than 90% since 2023, yet total spend has roughly doubled since last year — cheaper tokens have expanded the market, which means an index pause could simply reflect healthy digestion while demand remains real and capital expenditure is money well spent. But the more bearish read is harder to dismiss. Macro strategist Andreas Steno Larsen warned on June 9 that if token pricing continues to weaken, transactions across the cycle — from memory to broader hardware and data centers — could come to an end. With hundreds of billions in AI infrastructure bets still on the table, the direction of this one quiet index may matter more than almost anything else the market is watching right now.