2026 was meant to be the year Wall Street’s trillion-dollar artificial intelligence (AI) bet started paying off.
asserted
bet → mean → ?
Instead, it’s turning into a test of nerves.
asserted
it → turn → nerves
In the first half of the year, many AI stocks hit record highs as US hyperscalers like Google's Alphabet, Meta and Microsoft recorded bumper earnings.
asserted
hyperscalers → hit → earnings
Two of the largest AI players — Anthropic and OpenAI — announced plans for public listings, following SpaceX's much-hyped IPO in June, which briefly made Elon Musk the world's first trillionaire.
AI models, meanwhile, got noticeably better at difficult tasks in the race to beat human intelligence.
asserted
models → announce → intelligence
AI doubts trigger market turbulence
asserted
doubts → trigger → turbulence
The second half of 2026, however, has brought a lot more uncertainty.
asserted
half → bring → uncertainty
While the stock of Nvidia, a chipmaker, pushed even higher last month, many hyperscalers and chip or memory firms have seen their stock prices gyrate wildly.
asserted
prices → push → Nvidia
Then last week, after the Democrats secured control of the United States Congress in the midterm elections, technology stocks sold off dramatically, with the Magnificent Seven — including Meta, Apple, Amazon and Tesla — losing more than $1 trillion in value in a single trading day.
asserted
Seven → secure → day
In a further twist, the AI sector took another hit this week as Nvidia shares plunged nearly 10% over two days.
asserted
shares → take → days
The sell-off followed news that Nvidia's board — alongside the FBI — is investigating CEO Jensen Huang’s ties to a woman suspected by US intelligence of being a Chinese spy.
asserted
board → follow → intelligence
On Thursday, Huang resigned from the chipmaker when a second alleged marital affair was revealed, sending Nvidia's shares lower still.
asserted
affair → resign → shares
Amid huge spending, big returns remain elusive
Wall Street is growing increasingly cautious about the sheer size of the AI outlay, amid warnings that future earnings may not justify the huge spending.
uncertain
earnings → remain → spending
Growing pressure on US lawmakers to curb the expansion of AI data centers has added to those doubts.
asserted
pressure → grow → doubts
In September, investment bank Goldman Sachs estimated that hyperscalers need around $300 billion in annual AI revenue to break even on their spending, leaving a gap of roughly $230 billion.
asserted
hyperscalers → estimate → billion
"We're not in Kansas anymore," Dennis Dick, chief market strategist at the Canada-based Stock Trader Network, told DW, meaning that times have changed.
asserted
times → base → DW
Now, he said, investors are telling companies: "show me the money."
AI stocks needed a 'healthy correction'
Other analysts think it's too soon to expect the AI bubble to burst, including Kit Rees, investment strategist at London-based wealth manager Quilter, who thinks last week's rout was a "healthy correction."
"We’re still in the very early innings of the AI build-out," Rees told DW.
asserted
Rees → say → DW
"There’s still a lot of excitement around this technology and investors will want to have a slice of the pie."
asserted
investors → ’ → pie
In addition to the roughly $800 billion invested in capital expenditures this year, AI hyperscalers plan to spend a further $1.2 trillion in 2027 and $1.4 trillion in 2028, according to Goldman Sachs estimates.
uncertain
hyperscalers → invest → estimates
Questions over timing of Anthropic IPO
While OpenAI has delayed its public listing until next year, Claude-maker Anthropic still plans its IPO later this month.
asserted
Anthropic → delay → IPO
On Tuesday, the company said it would offer 825 million shares at between $70 and $78 per share.
asserted
it → say → share
Dick said Anthropic’s decision to go ahead amid the market sell-off left him scratching his head.
asserted
decision → say → head
“Is this the best timing?” he asked.
asserted
he → ask → ?
While Anthropic had originally targeted a valuation of over $2 trillion, it now thinks the firm could be worth around $1.2 trillion to $1.3 trillion.
uncertain
firm → target → trillion
Dick said the valuation depends on whether Anthropic can deliver on a forecast by Goldman Sachs of over $600 billion in revenue by 2030.
asserted
Anthropic → say → 2030
"If you get to a company with $600 billion in revenue, [a] $1.3 trillion [valuation] doesn’t sound that bad," he told DW.
asserted
he → get → DW
"But when you’re talking about a company with currently $52 billion in revenue, it sounds like it’s overvalued."
Other analysts, like Vivek Seth, chief executive officer of Shaires Holdings, are more bullish about the Claude-maker's prospects.
asserted
analysts → talk → prospects
“Anthropic is a once-in-a-generation company,” Seth told Bloomberg this week, noting how it has grown rapidly and has strong institutional backing.
asserted
it → tell → backing
“Its IPO will establish a market-clearing valuation as conditions stabilize.”
asserted
conditions → establish → valuation
Signs to watch, not reasons to flee
With many investors already running for the hills, Rees isn’t convinced the AI bubble will burst in the way skeptics expect.
asserted
skeptics → watch → way
She thinks they should watch for stronger signs that the AI trade is maturing.
asserted
trade → think → signs
This could be "a slowdown in AI-related revenues for companies, or signs of less demand or less appetite for financing companies’ AI plans, be it through debt or through equity offerings," Rees told DW.
uncertain
Rees → relate → DW
There are early signs that investors have less appetite for AI-related debt.
asserted
investors → be → debt
Borrowing costs for AI infrastructure firms like CoreWeave and Oracle have soared recently.
asserted
costs → soar → CoreWeave
Even the cost of insuring Nvidia’s debt against default has risen 76% in a month, CNBC reported last week.
asserted
CNBC → insure → month
"They [Oracle] have been spending, some people say, like drunken sailors," Dick told DW.
asserted
Dick → spend → DW
"They’ve compromised their balance sheet to a point where the stock’s fallen 65% from the highs.
asserted
stock → compromise → highs
That has not stopped Oracle, along with Meta, Microsoft and Alphabet, from continuing to issue new AI bonds.
asserted
That → stop → bonds
Since the end of last month, the four AI giants have announced deals worth $67 billion.
asserted
giants → announce → deals
Rees said another sign could be any new regulations that make it harder to build out US data centers.
uncertain
it → say → centers
This summer, New York and Texas temporarily paused new data center approvals.
asserted
York → pause → approvals
…and 7 more, not listed.
The Aporia
