AI Jitters and Rate-Hike Fears Combine for Worst Day for Stocks Since July
Worries about interest rates spurred by a surprisingly hot inflation reading collided with growing jitters in the technology sector on Wednesday. And a session that began with rising bond yields and expectations of an October rate increase culminated with the stock market’s worst day in months. (…)
The worst of the selling Wednesday was concentrated in companies that have borrowed heavily to finance data center build-outs and have signaled plans to keep doing so, making lenders uneasy. Shares of CoreWeave tumbled more than 15% while some of the bonds it issued earlier this year dropped to fresh lows in the secondary market, underscoring investor concerns about its credit quality.
Momentum-chasing strategies that bid up the fastest-rising stocks and sell when prices fall likely contributed to the accelerating selloff Wednesday afternoon, said Art Hogan, chief market strategist at B. Riley Wealth Management. (…)
Semiconductor companies, many of which count CoreWeave and Oracle among their biggest customers, were among the other worst performing stocks on Wednesday. (…)
Among the concerns: additional competition from one of China’s largest memory-chip makers, which announced it would roll out a new kind of DRAM chip earlier this week. (…)
Heading into Wednesday, investors had grown increasingly confident that the Fed would pause this month after having just lifted rates in September—its first-rate hike in three years. A series of Fed officials had encouraged that thinking in recent weeks, arguing that inflation pressures weren’t severe enough to warrant a rapid string of hikes.
Wednesday’s inflation print dented that confidence—even if it still didn’t spur a sea change in investors’ outlook.
As expected, the data showed a big jump in gasoline prices spurred by the continuing Iran conflict. But it was increases elsewhere that were most concerning to investors since the Fed tends to focus on so-called core prices, which exclude volatile food and energy categories.
Those rose 0.3% from the previous month, above the 0.2% increase anticipated by economists surveyed by The Wall Street Journal. Inflation in services continued to be sticky. But there were also surprising increases in the prices of goods like apparel, stirring concern that high energy prices were starting to drive up inflation elsewhere by increasing transportation costs. (…)
- Broad Pickup in Prices Lifted Inflation to 3.7% in September Gasoline drove up prices, but other costs also rose when energy was stripped out of the measure
Recent CPI trends are not friendly with core CPI and core Services having both accelerated to 4.0% annualized rates …
… threatening to soon bring real labor income growth near zero on a YoY basis with likely adverse consequences on spending.
Anthropic’s Quarterly Revenue Jumped 45% to Nearly $17 Billion
Anthropic’s revenue climbed 45% to $16.8 billion in the third quarter, according to people familiar with the matter (…).
Anthropic had posted revenue of $11.6 billion in the second quarter, and just under $5 billion in the first quarter. Investors expect Anthropic to reach more than $110 billion in annualized revenue by year-end.
Bloomberg has other key info:
- Anthropic is generating positive free cash flow, had adjusted gross margin of 85% and an over 11% positive adjusted operating margin as of the third quarter, according to the people. (…)
- The company’s net revenue retention, a measure of its ability to keep its existing customers and increase their spending over time, was 900%, some of the people familiar with the update said. Its compute margin, or the revenue left over after subtracting the cost of its computing power and other necessary infrastructure, is more than 40%, some people said..
It also chose to quote some cautious views:
- “I am cautious about Anthropic’s revenue outlook,” said Jung In Yun, chief executive officer at Fibonacci Asset Management. “Pricing could fall further as competition intensifies, so even strong usage growth may not translate into equally strong revenue growth or sustainable profitability.” (…)
- “This news is somewhat negative,” said Matt Maley, chief market strategist at Miller Tabak + Co. “It’s something that will likely raise some concerns in investors’ minds. We also have problems in the credit market surrounding the tech sector in the US. We have already seen the CDS prices for the hyperscalers rise significantly in recent weeks.”
Anthropic revenues:
2025: $386M.
2025: $4600M. Operating loss: $8000M
Q1’26: $5000M.
Q2’26: $11600M.
Q3’26: $16800M. Operating profit: $1850M with 40% compute margin and “positive free cash flow”.
“This news is somewhat negative”? Really?
From the FT:
The start-up has also clamped down on “distillation” — a controversial practice whereby rival AI labs train their models on the outputs of more advanced ones.
When rival companies “distil” their models from Anthropic’s models, that involves heavy usage and spending.
Anthropic told investors on Wednesday distillation accounted for 5 per cent to 7 per cent of revenue and that they had cut it back “materially” in the third quarter of the year, said four people with direct knowledge of the matter.
One investor said Anthropic’s revenue growth was impressive in the context of a price war with rivals and efforts to cut back on distillation, which had to date been a meaningful source of the company’s revenue. They added that the company has already seen demand increase as efficiency has improved.
Taiwan Semiconductor Manufacturing Co. delivered a better-than-expected revenue outlook, sustaining a torrid pace of growth yet failing to dispel rising concerns about the sustainability of a global AI buildout.
The chipmaking linchpin is projecting sales of as much as $54.3 billion in the December quarter, handily beating expectations. That should translate into a roughly 44% rise in 2026 revenue in US dollar terms, and followed a better-than-expected 69% jump in profit for July to September. (…)
Hsinchu-based TSMC serves as a key barometer of global AI demand as every major industry player including Nvidia, Advanced Micro Devices Inc. and Alphabet Inc.’s Google relies on the Taiwanese chipmaker to turn their most cutting-edge designs into reality. (…)
TSMC’s outperformance follows a spate of positive results from big industry names including ASML Holding NV (…).
But investors were in a foul mood yesterday, particularly the more speculative ones:
(…) The Kospi has fallen about 21% from its September high. (…) A closely watched US gauge of semiconductor stocks has erased $1.6 trillion in market value in a six-day losing streak. Strong earnings reports from Samsung and ASML Holding NV have failed to calm investors. (…)
Adding to jitters over the AI trade, CoreWeave Inc. shares plunged as its bonds fell to levels that typically indicate distress, leading investors to again question the sustainability of large spending on AI infrastructure. (…)
Recent weeks have brought mounting signs of increasing investor concern about the sheer size of AI debt loads. CoreWeave’s bonds due in 2032 have slid to about 78 cents on the dollar, prompting Elon Musk to flag the drop with a “yikes” on his social media platform. That’s after just late last month, SoftBank Group Corp. had to pay record yields in an effort to fund its massive AI ambitions. (…)
The selling has made Kospi’s valuation more attractive at less than 5 times forward earnings estimates, less than half the level a year ago. Yet with the bearish sentiment prevailing, money managers are calling for caution.
“The South Korean stock market’s valuation has dropped to an extreme level now,” said Kang DaeKwun, chief executive officer at Life Asset Management. While any pullback from the current level can be considered as an entry point, investors must make sure that problems over tech funding are resolved, he added.
Not helping the mood:
AI Race Puts $3.5 Trillion at Write-Down Risk, Gartner Warns
More than $3.5 trillion of AI infrastructure spending could be at risk of impairment by 2030 as the industry’s investment boom threatens to outpace the revenue required to justify it, according to an analysis by market research firm Gartner Inc.
AI data center owners would need to generate a cumulative $8.5 trillion in revenue through 2030 to cover capital spending, financing costs and achieve a 12% return on invested capital. Yet Gartner forecasts no more than $3.6 trillion in attributable enterprise and consumer AI revenue over that period — less than half the required amount. (…)
The implications are especially stark for specialist AI infrastructure providers such as CoreWeave Inc. and Nebius Group NV. Their business models are more directly exposed to demand for computing capacity than diversified hyperscalers like Microsoft Corp, Alphabet Inc. and Amazon.com Inc.
Gartner predicts that cost pressures and limited differentiation will force more than 80% of neocloud providers to cease operating independently by 2030.
Will Sommer, senior director analyst at Gartner, said many of these neocloud providers are working to develop the higher-margin software and services of cloud providers, though few may succeed.
“All of them want to be the full-stack. They can’t all be,” he said. “For every dollar that’s spent on compute, only a fraction of that is going to the owners of compute.”
A Columbia Business School professor published his own in-depth analysis on September 24 (Financing the AI Buildout), calculating that
Under the central assumptions, the 2025–2032 buildout requires approximately $3.7 trillion in annual revenue by 2032 to earn a 10% unlevered return at a 50% operating cash-flow margin.
This is equivalent to roughly $5.5 per installed GB300 GPU-hour at full utilization, or about $6.9 per billed GPU-hour at 80 percent utilization, placing the required price in the broad range of current frontier-GPU rental prices.
The underwriting risk is therefore whether demand and pricing power remain strong enough to sustain such prices once 180 GW of compute capacity has been added.
Gartner’s analysis sees a need for $8.5T in revenue through 2030 to achieve a 12% ROIC but it only expects $3.6T in AI revenues.
The Columbia analysis says that $3.7T revenues by 2032 would be enough to generate a 10% return.
But how can we reliably forecast usage and revenues at this time? We have yet to see how AI agents penetrate our lives. The first chart (from JP Morgan) plots token volume through August but the second chart shows how agent demand has been exploding after June.
Unlike the internet bubble, infrastructure spending is following demand and can’t supply enough.
If Anthropic’s Q3 profitability is sustainable, Jevons Paradox could well take care of the current revenue and financing angst.
But it would be nice if bond yields could stop rising…
A few more relevant charts from JP Morgan Asset management:
Remember when Cisco and Norther Telecom P/Es were 100+? I do.
(Koyfin)
CEO confidence drops into negative territory
CEO confidence in the U.S. economy sank into negative territory earlier this month, finds a survey from the Conference Board.
Data: The Conference Board and The Business Council; Chart: Axios/Emily Peck
41% of CEOs expected economic conditions to worsen — up from 19% in the previous quarter.
Despite their dimming views, CEOs have for the most part maintained their spending plans, per the survey.
A separate survey found that 89% of U.S. CEOs said they expected to increase headcount over the next three years as AI creates more business opportunities.
- “As we put more AI into our processes and platforms, the capacity opens up,” Tim Walsh, the CEO of KPMG, which conducted the survey, told Axios’ Nathan Bomey.
- “What you’re seeing there is CEOs being really optimistic that they can actually expand their business, which is going to result in a bigger workforce for them.”
