Michael Cembalest Warns AI Data-Centre Buildout Faces Real Hurdles After Midterms

Investors have spent weeks asking whether the recent AI sell-off is a valuation problem, a liquidity problem or an early sign that spending on data centres has raced ahead of reality. In his latest Eye on the Market, J.P. Morgan Asset Management strategist Michael Cembalest tackles all three, moving from the US midterm elections to the financing of AI infrastructure and, somewhat surprisingly, to the collapse of Theranos.

Cembalest starts with politics. Based on New York Times projections available early on November 4, Democrats were on course for 53 Senate seats, with two independents expected to caucus with them, and 257 House seats. That would make 2026 the sixth midterm since 1910 in which the president's party took control of both chambers and then lost them. Cembalest puts those results alongside earlier episodes, from wartime inflation in 1918 to the backlash over the Iraq War in 2006, but his core point is that markets follow the business cycle far more closely than they follow the voting booth.

Interest rates do not drop because Congress changes hands, and the Federal Reserve does not rethink its approach because of election-night maps. The bill for AI infrastructure is just as large under any majority. That is why the real political effect shows up in a narrower place: the local and state fights over where data centres get built.

Here the numbers get specific. Cembalest estimates that 37.5 gigawatts of data-centre capacity planned for 2028 could face construction pauses, and another 15 gigawatts could meet extra regulatory conditions. For a market that has already priced in enormous spending growth, that is a serious complication. Projects that stall or cost more to approve delay the revenue that equipment makers, chipmakers and utilities are counting on.

The money behind the buildout is also getting harder to read. Cembalest notes that nearly $4 trillion in future purchase and lease commitments sits outside conventional balance-sheet liabilities. At the same time, technology valuations have come down and credit spreads have widened, which suggests lenders are growing more cautious. Investors who assumed capital spending was locked in may need to look again at how those commitments are structured and who ends up holding the risk.

The Theranos section is the sharpest part of the piece. Elizabeth Holmes built a story around breakthrough technology, a magnetic founder and plenty of capital, and none of those things could replace transparent accounting, independent oversight or results that could be checked. Cembalest's warning is that the same ingredients can make AI businesses look more solid than their disclosures justify. For everyday investors, the lesson is simple: ask what a company has actually proven, and how much of its growth depends on borrowed money and projects that have not yet been approved.

Read the full analysis at Investing.com.

Related: UP Governor Anandiben Patel tells youth: build startups that fix local problems · Qualcomm shares gap down 13% after results, then claw back some losses