AI token spending index falls 20% signalling loss of pricing power

The Silicon Data LLM Token Expenditure Index – the market’s most‑watched gauge of what customers actually pay for AI usage – slipped almost 20 % from its May peak. The decline comes after the index had nearly doubled since its debut in December, and investors are rushing to decode what the slide means for the sector’s famed “AI trade.”

For stock‑market participants, the drop is a warning bell that cost‑sensitive users may be tightening the taps on token‑based services. If buyers are less willing to pay, the avalanche of $700 billion‑plus capex that has underpinned the recent rally could lose its forward‑looking momentum.

Media additions

Silicon Data, the firm behind the index, stresses that the number is not a simple price tag. It blends the price per token with the volume of usage, making it a proxy for “marginal willingness to pay.” The firm cautions readers not to treat the index as a static price reference.

What the numbers say

• Index down almost 20 % from the May high.
• Token prices have fallen more than 90 % since 2023, while total spend has roughly doubled year‑over‑year.
• All‑in AI capex is still accelerating, but the gap between investment and sales has widened to about 46 % – a divergence larger than the 32 % recorded during the 2001 telecom bust (according to Allianz Research).

These figures paint a mixed picture: cheaper tokens are expanding the market, yet the revenue per unit of compute is flattening.

Voices from the front line

"There are increasing reports that users of AI solutions, priced in tokens, are having to restrain unlimited use due to high costs."

Louis Navellier, veteran investor, via Yahoo Finance

"During the training phase, the cost of AI infrastructure and token generation is extraordinarily high, but in the current inference stage, the economics are significantly better."

David Miller, senior portfolio manager at Catalyst Funds, via Yahoo Finance

From the chip side, Morningstar reports that Nvidia has softened its previously pledged $100 billion investment in OpenAI, while Oracle announced a $50 billion debt‑and‑equity raise to fund its cloud and AI workloads. Richard Windsor, founder of Radio Free Mobile, summed up the risk:

"The business model of compute upon which the AI boom is being built is not viable."

Richard Windsor, analyst, via Morningstar

Windsor notes that each gigawatt of AI compute costs about $50 billion to build and generates roughly $10 billion of annual revenue – a margin that leaves little room for profit or debt repayment.

Regulatory headwinds add to the mix

US regulators have recently requested OpenAI to stagger the roll-out of an upcoming release, and the European Union’s AI Act targets frontier models for mandatory evaluations and stringent transparency requirements. While these rules do not directly cap prices, they raise compliance costs that can tilt customers toward cheaper, open‑source alternatives.

Open‑source models such as Meta’s Llama have already forced closed‑model providers to cut token prices, a trend highlighted by The AI Cronicle.

Market‑wide fallout

In the equity arena, former Bank of America strategist David Woo now bets against the Nasdaq‑100, arguing that the once‑tight link between hyperscaler capex and AI stock performance has fractured. As Woo puts it, “the market no longer trusts higher capex spending to reflect higher returns on investment.”

"The AI trade may not be crashing—but it is clearly stalling."

David Woo, founder of David Woo Unbound, via Benzinga

Woo points to a “negative” correlation between hyperscaler earnings and AI‑related capex for the first time in three years, suggesting that investors now view spending as a cost pressure rather than a growth catalyst.

Meanwhile, Nvidia’s earnings report – a 34 % jump in revenue and a 41 % surge in data‑center sales – has temporarily lifted sentiment.

Investors are left balancing two opposing reads. If the index’s recent flattening proves to be a short‑term digestion of a shift toward cheaper models, the flood of capital – still projected to top $1 trillion in 2027 – may keep the market buoyant. But if token‑price erosion flags a deeper loss of pricing power, the very engine that has propelled the “AI trade” could sputter, forcing a re‑pricing of valuations across the tech megacap cluster.