The AI trade is losing one of its key signals

AT A time when markets are growing uneasy over whether the enormous sums being poured into artificial intelligence will ever pay off, the prices the sector commands for each unit of usage are drifting lower.

The Silicon Data LLM Token Expenditure Index, which tracks what users pay for AI tokens, is down almost 20 per cent from a high in May after nearly doubling since its inception in December 2025.

The gauge is the cleanest read anyone has on the US$700 billion-plus capex boom that has done the sector’s heavy lifting.

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.

“There are increasing reports that users of AI solutions, priced in tokens, are having to restrain unlimited use due to high costs,” said veteran investor Louis Navellier.

“The chatter that OpenAI is pushing back its initial public offering to next year is seen as a sign that, currently, profitability remains a problem.” 

Just to clarify, a softer index doesn’t mean AI is getting cheaper. 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.

It could also point to a genuine softening in what buyers are prepared to shoulder. 

Each of these possibilities carries different implications. Silicon Data, which built the index, has warned people to stop reading it as a price tag. The firm calls it a proxy for marginal willingness to pay.

SEE ALSO

AI development may be progressing too fast to manage risks effectively: UN experts

Singapore startups double down on AI usage, running multiple platforms: Aspire report

Cheaper AI is better: Soaring bills are reshaping how businesses choose models

The world is burning power for AI – Singapore wants to do it differently

More

Let’s start with a benign read: While token prices have collapsed more than 90 per cent since 2023, total spend has roughly doubled since 2025.

Cheaper tokens have expanded the market. This means than an index pause is simply digestion, while demand is real and capex is money well spent.

The bull case for Nvidia, memory makers and data centre names rests here.

Now for the interpretation that is keeping people up at night: Bears warn that sustained weakness in the index could end the trade that saw nearly the entire AI cohort rally hard this cycle. 

It is token spending that justifies the next capex order, and the bill is already looking stretched.

Allianz Research said there’s nearly a 46 per cent growth gap between AI investment and sales. That is worse than the 32 per cent divergence measured during the 2001 telecom bust. 

Fortunately for bulls, the downward trend has paused. It is too early to call a bottom after one flat week, but it is enough to keep the case for a rebound alive.

“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,” said David Miller, senior portfolio manager at Catalyst Funds.

“The net use of AI delivers a positive return on investment for companies, at least over the long term.”

There is also a more recent, demand-side reason why the bearish read may have legs. Washington has a newfound willingness to exert control over a pivotal industry.

The US government only this week removed foreign access restrictions on Anthropic PBC’s Fable 5 model, days after regulators requested OpenAI to stagger the roll out of an upcoming release.

Meanwhile, the European Union’s AI Act targets frontier models for mandatory evaluations and stringent transparency requirements.

None of these cap prices directly, but they do create a deployment-and-compliance burden on top platforms that lesser-but-still-useful systems do not carry.

This consideration may just offer companies’ financial chiefs a rational reason to route workloads to cheaper models.

To be sure, this is not a chip-glut call. Top-end graphics processing units (GPU) and high bandwidth memory are sold out through 2026, with no real relief arriving until 2028.

The hardware tell is more subtle in that it points to a demand-mix shifting away from top-end training GPUs, toward inference-optimised parts. This changes the mix of winners, but does not hand you a short.

Still, “unbridled” market enthusiasm, intensifying competition from China and price sensitivity makes DWS strategists led by chief investment officer Vincenzo Vedda cautious. “We are monitoring areas where valuations may look stretched,” they said.

The conclusion is that the token chart cuts both ways, and one should hold both reads at once.

If the late-June flattening holds and the dip was just mix-shift digestion, cheaper tokens will keep on expanding the markets, meaning capex spending stays justified, leaving the bull case intact.

If instead, this is the point where customers’ willingness to pay peaks just as regulatory headwinds nudge demand down-market, then the most expensive part of the trade is also the first to crack.

That is because it is a pricing-power story, not a silicon story, that is funding the march toward US$1 trillion of capex in 2027. BLOOMBERG