AI Doubts: Tesla, Meta, Nvidia Slip; China Stocks Lose $2.4T

Chinese stocks entered a bear market on 22 October 2026, with US$2.4 trillion lost since a peak in May, Bloomberg reported. It put the fall down to “a protracted slump in consumption and mounting doubts over the profitability of artificial intelligence (AI) companies”, which it said have made Chinese stocks one of the world’s worst performers.

Hours later in New York, Nasdaq 100 futures were down 1.72% before the bell, Reuters reported, after earnings from Tesla and Meta “did little to dispel doubts about the AI outlook.” Tesla fell 3.6%, Meta 2% and Nvidia 1.9% in premarket trading. This is a follow-up to our AI-fear dashboard from this morning: it adds the company-level moves, Bloomberg’s $2.4 trillion and a demand-side number Axios published the same day on falling token prices.

Business Pill · THE JEVONS PARADOX

A short Business Engineer explainer on the Jevons Paradox: why cheaper AI can mean a bigger bill. It teaches the general idea only and says nothing about any company or market move in this piece.

What Was Reported in China

The CSI 300 Index closed 1.8% lower on Thursday, bringing its total losses from a high in June to 20%, according to Bloomberg’s report as published by The Edge Malaysia. “The rout in Chinese onshore shares has erased US$2.4 trillion in market value from a recent peak, according to data compiled by Bloomberg.”

• The causes Bloomberg names. A slump in consumption, doubts over the profitability of AI companies, and the lack of “a broad-based spending package”, with economic growth that’s “barely meeting national targets”.
• Hong Kong. The Hang Seng Tech Index slumped as much as 5.3%, “the biggest intraday drop since July 17”, which Bloomberg ties to regulatory and business challenges facing China’s e-commerce industry rather than to AI.
• State buying. Turnover in a basket of nine ETFs “known to be traded by the national team” reached 52 billion yuan (US$7.79 billion) by 2.45pm local time, the most since 15 October.
• One fund manager’s view. “This slide is too rapid and may just herald further declines ahead,” Fu Zhifeng, chief investment officer of Shanghai Chengzhou Investment Management, told Bloomberg, citing “uncertainty over how to justify hefty AI spending” among the overhangs.

What Was Posted on X

Bloomberg’s own post from its Evening Briefing Asia: Chinese stocks entered a bear market, losing $2.4 trillion since their May peak, as doubts over the profitability of their AI companies and a slump in consumption weighed on sentiment. It names two causes, not one, which is why our table lists both.

Axios’s post links its report on falling token prices, with a chart of two token price indexes. It matters here because, as we read it, token spending is the demand-side counterpart to the capex that investors questioned the same morning.

The Premarket Table: Twelve Moves and Their Stated Reasons

Reuters’ premarket report names a reason for most of the US moves, and it is mostly AI spending. The table lists each move with the reason the outlet gives, including the one factor that was not about AI: oil.

US moves are premarket figures Reuters gave at 7:10 a.m. ET on 22 October 2026, before regular trading opened. The China figures are Bloomberg’s, published by The Edge Malaysia. The selection of rows is ours; each reason is the outlet’s own.

An earlier Reuters story put Meta’s premarket fall at about 3%, with about $50 billion of market value at stake if the losses held; we covered that in our piece on the Meta share move. Reuters’ markets report, the source for this table, gives 2% at 7:10 a.m. ET.

The Demand-Side Number: Token Prices

Axios published a separate data point on 22 October 2026: the price of tokens, the units AI models bill in, is falling. “There’s no token price that captures all activity, but several available public indexes show similar declines,” Axios wrote.

• Ramp. The index from the corporate spending company, based on what businesses spend on OpenAI and Anthropic, “is down 48% from its late February peak of $1.22 per million tokens.” By our arithmetic, that puts it near $0.63.
• Silicon Data. Its LLM Token Expenditure Index “has broken the buck in recent days, falling below $1 per million tokens.”
• Spend versus use. “For the first time, we’re seeing token usage go up and token spend hold flat,” Ramp’s lead economist Ara Kharazian told Axios. “And it’s now occurred for two months straight.”
• The capex on the other side. The largest US hyperscalers “are on track to spend $800 billion on capital expenditures this year and are forecast to spend another $1.1 trillion in 2027, according to Goldman Sachs,” Axios reported.

Semafor made the same link in its morning report, quoting Apollo’s chief economist that a slowdown in spending on compute tokens suggests “productivity returns for firms on their AI investments are either arriving far slower than projected or failing to materialize at all.” Axios also gives the counter-argument: the price drop “has been largely shrugged off with frequent references to the so-called Jevons Paradox”, and AI usage, measured in tokens, “continues to rise incredibly quickly.”

Connecting the Coverage

• Our AI-fear dashboard logged market moves and counterpoints from 21 and 22 October, from China’s CSI 300 to US Treasuries and SoftBank’s market value; this piece adds the company-level premarket moves and the token-price data.
• Our piece on the Meta share move covers the analysts who raised their targets while the stock fell.
• Our piece on Nadella’s cost-to-outcome frontier covers his letter’s line that “customers will increasingly measure the cost of tokens against the outcomes they produce.”
• Semafor cited an Australian data centre operator’s profit warning that drove its stock down as much as 44%; our piece on NEXTDC’s guidance cut covers the ASX operator whose plunge Bloomberg’s headline put at 44%.

The Structural Read

As we read it, the market moves sort into two kinds. Tesla and Meta fell on their own spending news, Reuters reported, while SpaceX and Super Micro fell as investors focused on “the soaring costs of investments and unclear return timeline.”

As we read it, falling token prices cut both ways in the sources. Axios reports the decline reflects customers switching to cheaper open-weight models and price cuts by OpenAI and Anthropic, while usage keeps rising.

As we read it, China’s fall is not a pure AI story. Bloomberg names a consumption slump and policy disappointment alongside AI profitability doubts, and ties the Hang Seng Tech drop to e-commerce.

Three Implications

FOR AI MODEL MAKERS Axios writes that token spending by businesses is closely related to revenue for AI model makers; as we read it, that makes the flat spend line a number to watch.

FOR THE CAPEX PLANS Axios cites Goldman Sachs: the largest US hyperscalers are on track to spend $800 billion on capital expenditures this year and are forecast to spend $1.1 trillion in 2027.

FOR CHINA’S MARKET Bloomberg Intelligence analysts wrote that the national team might need to buy more and for longer than in its two most recent rescues, according to Bloomberg’s report.

The Business Engineer Lens

This story maps onto the Business Engineer framework The AI Capex Map.

The framework sets out to map the financing of AI: “How the money flows, where it cracks, what it buys, who can absorb it, and what happens when revenue can’t catch up.”

As we read it, today’s reports touch the last clause: Ramp’s data shows business token spend flat while the hyperscalers’ capex, per Goldman Sachs via Axios, keeps rising. The framework was published in May 2026 and does not discuss these figures.

What Is Not Established

The US moves are premarket figures given before regular trading opened, and the final moves may differ. The reasons in the table are the outlets’ attributions, not measured causes: Bloomberg names a consumption slump and policy disappointment alongside AI doubts for China, ties the Hang Seng Tech drop to e-commerce, and Reuters says rising oil added to the jitters. Reuters gives no separate reason for Nvidia, Amazon or the chip stocks. Ramp’s index reflects spending by businesses on OpenAI and Anthropic, and Axios says no single token price captures all activity.

The Bottom Line

Bloomberg data shows US$2.4 trillion erased from Chinese onshore shares since a May peak, and Bloomberg names doubts over AI profitability among the causes. In New York, Reuters reported Nasdaq 100 futures down 1.72% before the bell, with Tesla down 3.6% and Meta down 2% after results that it said did little to dispel doubts about the AI outlook. The same day, Axios reported Ramp’s token price index down 48% from its February peak, with business token spending flat for two months.