AI infrastructure stocks lose $4.2 trillion as the "picks-and-shovels" boom reverses

Deep News
6小時前

Global manufacturers, utilities, and mining companies have seen a combined $4.2 trillion in market value erased since June, a sell-off that has wiped out nearly half of the $9.4 trillion in gains these companies accumulated over the previous 12 months, as the once-popular AI "picks and shovels" trade rapidly reverses.

According to an analysis by the Financial Times, the more closely a company is linked to data center construction in media coverage, the more severe the hit it has taken. Investor confidence in the AI boom is wavering, with everything from makers of gas turbines for hyperscale data centers to the high-end fiber optic companies needed to connect AI server racks falling out of favor. The industrial sector has become the second-worst performer in the S&P 500 since the summer, trailing only technology.

The market shift reflects broader concerns: hyperscale cloud companies continue to push capital spending higher through debt financing, compounded by growing political resistance to data center construction in the United States, leading investors to question whether the building boom can continue.

Market value of leading companies halved

Caterpillar, a major supplier of gas turbine generators for data centers, has fallen more than 30% over the past six months, and its market value has shrunk by $21.4 billion, or 44%, since its stock hit a record high on June 30.

In addition, nuclear power company Oklo, despite signing a power supply contract with Meta in January, has seen its shares plunge 49%. HVAC system manufacturer Comfort Systems and fiber optic maker Corning have each lost about half their market value over the same period.

According to the Financial Times, Deutsche Bank senior U.S. economist Brett Ryan said a large amount of institutional money had "blindly rushed into" these trades, "potentially contributing to overly optimistic valuations for some companies." He also noted that the stock market is "converging toward signals the credit market had already sent" — the latter having long warned about the growth of risk financing in data center construction.

Venture capitalist and MIT researcher Paul Kedrosky put it bluntly: investors treating traditional industrial companies like semiconductor stocks is a mistake in itself.

"In discussions with hedge funds and family offices, the argument that 'everything is a growth stock' is everywhere — which is, of course, absurd."

Still, despite the pressure on share prices, the operating fundamentals of many companies remain solid. The latest U.S. government data show that in the 12 months through September, U.S. data center construction spending (excluding IT equipment and chips) reached $96 billion.

Third-quarter results also confirm the resilience of demand. Caterpillar reported a record backlog of $78 billion, with sales in its power generation business up 75% year over year, driven mainly by data center demand. Siemens' backlog of electrical equipment orders from data centers grew 60% in 2026 to 7 billion euros; Eaton's comparable orders doubled year over year. Several gas turbine manufacturers said they had secured billions of dollars in orders in the first nine months of 2026.

"In terms of the revenue fundamentals and potential earnings outlook for these companies, we still see very strong expectations," said Dell'Oro Group analyst Alex Cordovil. He added that broader trends such as electrification and supply chain restructuring continue to support the core businesses of many industrial companies. "Construction has not slowed... backlogs are at historic highs, which means that at least through 2027, revenue growth is largely locked in."

Richard S Warzala, chief executive of industrial control equipment manufacturer Allient, said on a Nov. 5 earnings call that demand for AI will not decline.

"Even if a moratorium is imposed on new data center construction, projects that have already broken ground must be completed."

In addition, Kedrosky is skeptical of backlog data, arguing it is a lagging indicator that reflects earlier expectations rather than real current demand. He pointed to "double or even triple ordering" — companies stockpiling excess inventory to secure a place in line amid a perceived supply shortage.

Brett Ryan's assessment is more direct: the stock market is completing a belated valuation correction, aligning with risk signals the credit market had already conveyed. This adjustment, which has lasted several months and erased more than $4 trillion in market value, is essentially a systematic repricing by capital markets of the "everything is a growth stock" narrative.

Regulatory uncertainty after the midterm elections

Another important backdrop to the sell-off is the uncertainty over data center regulatory policy after the U.S. midterm elections. In this month's midterm elections, proposals to restrict data center construction won broad support among American voters, but their actual impact remains unclear.

In a post-election report, Citi Research noted that while stricter regulation is expected to bring more volatility to the data center industry, the possibility of widespread construction restrictions remains low.

Constellation Energy CEO Joseph Dominguez said the post-election market reaction was an "overreaction," stressing that Democrats "care deeply" about protecting blue-collar jobs at data centers.

Deutsche Bank's Ryan believes that even if bans are introduced in some areas, they will only shift construction demand to regions with looser regulation.

"Ultimately, if the market demands these facilities, they will eventually be built somewhere."

免責聲明:投資有風險,本文並非投資建議,以上內容不應被視為任何金融產品的購買或出售要約、建議或邀請,作者或其他用戶的任何相關討論、評論或帖子也不應被視為此類內容。本文僅供一般參考,不考慮您的個人投資目標、財務狀況或需求。TTM對信息的準確性和完整性不承擔任何責任或保證,投資者應自行研究並在投資前尋求專業建議。

熱議股票

  1. 1
     
     
     
     
  2. 2
     
     
     
     
  3. 3
     
     
     
     
  4. 4
     
     
     
     
  5. 5
     
     
     
     
  6. 6
     
     
     
     
  7. 7
     
     
     
     
  8. 8
     
     
     
     
  9. 9
     
     
     
     
  10. 10