Lenore Palladino, an associate professor of economics and public policy at the University of Massachusetts Amherst, argues that ordinary retirement savings are being used to finance the AI data-centre buildout — and that workers could be left holding the risk if demand for those projects falls short. Writing for the Law and Political Economy Project, she points to public pension allocations to private equity and private credit, and to a Trump administration push to open 401(k) plans to similar funds.
Today, Lenore Palladino explains how working people’s retirement savings are being used to fuel the AI bubble.
November 9, 2026
The sell-off she’s writing about
Palladino’s essay, published on 9 November, opens with the previous Thursday’s tech-led rout. That session, 5 November, saw the Nasdaq Composite fall 5.5%, the S&P 500 4.4%, and the Dow Jones Industrial Average 3.4%, according to CNBC — the steepest one-day drops for all three since April 2025. The Guardian put the decline in US stock-market value at about US$3.2 trillion. Reuters put the loss among S&P 500 companies at about US$2.9 trillion. The Magnificent Seven — Nvidia, Apple, Microsoft, Amazon, Alphabet, Meta and Tesla — lost more than US$1 trillion in combined market value that day, CNBC reported, with Nvidia accounting for more than US$440 billion of the drop.
Investors were reacting in part to midterm results that raised the political risk of data-centre slowdowns. Anxiety about AI spending was already visible in the earnings week that ended on 30 October. On 29 October, Alphabet shares fell about 13%, wiping out roughly US$500 billion in market value, after the company reported negative free cash flow of about US$9.1 billion for the third quarter and kept full-year capital-expenditure guidance of US$195 billion to US$205 billion, according to CNBC. Palladino’s claim that the Magnificent Seven lost nearly a trillion dollars on 29 October is not supported by the same session’s reporting; the group’s one-day loss on that scale is the 5 November figure.
How retirement money can reach data centres
Public pension funds have long placed a slice of their assets in private equity and private credit. Americans for Financial Reform Education Fund, in an October report, Tracing the Money, estimates that 78 private equity firms owned or backed 13 to 17 gigawatts of active US data-centre capacity across 670 sites at the end of 2025. That’s 25% to 30% of the 54 gigawatts of capacity then in service, the group says. Private equity-owned or financed companies had also announced or begun at least 26.8 gigawatts of planned US capacity, more than half of a 50-gigawatt pipeline through 2030.
The five largest private equity data-centre owners in that tally, by combined active and planned capacity, are Brookfield Asset Management, Ares Management, Blue Owl Capital, DigitalBridge and Apollo. AFR says they control or hold investments in 17% of existing capacity and 30% of planned capacity. The report also cites a Morgan Stanley estimate that more than a third of the US$2.9 trillion expected to be spent on new data centres from 2025 to 2028 — about US$1.1 trillion — would come from private equity through private credit, joint ventures or equity stakes. AFR writes that this financing is “mostly drawing capital from public pension funds and university endowments.” That is the group’s characterisation, not an official breakdown of every fund’s books.
The pool of retirement money is large. The Investment Company Institute’s second-quarter figures put US employer-based defined-contribution plans at US$15.0 trillion as of 30 June 2026, including US$10.8 trillion in 401(k) plans. Government defined-benefit plans held US$10.4 trillion. IRAs held US$19.9 trillion. Total US retirement assets were US$51.2 trillion.
| Stage | What it shows |
|---|---|
| Retirement accounts | Defined-contribution plans held US$15 trillion in mid-2026, including US$10.8 trillion in 401(k)s |
| Private funds | Public pensions already allocate to private equity and private credit; a 401(k) safe-harbour rule is still under review |
| The buildout those funds can finance | AFR estimates private equity backs 25–30% of active US data-centre capacity, and more than half of the planned pipeline |
What the 401(k) change actually does
Palladino writes that, thanks to recent Trump administration efforts, the US$15 trillion in private-sector retirement accounts is “now, for the first time in history, available to be invested in private funds.” The official record is narrower than that sentence.
On 7 August 2025, President Trump signed Executive Order 14330, directing the Labor Department to re-examine ERISA fiduciary guidance and to propose rules or a safe harbour for plan fiduciaries who offer asset-allocation funds that include alternative assets — private markets, real estate, digital assets, commodities, infrastructure and lifetime-income strategies. The Department published a proposed rule on 31 March 2026. It is process-based and asset-neutral: it does not require plans to offer private funds, and it does not name data centres. The Department of Labor sent a final version to the White House Office of Management and Budget for review on 5 November, according to the Insured Retirement Institute and 401(k) Specialist. Publication in the Federal Register, and an effective date toward the end of the first quarter of 2027, were still ahead. The rule had not yet taken effect when Palladino’s essay appeared.
Credit-market warnings, and the limits of the comparison
In a September commentary, S&P Global Ratings wrote that “the credit quality of hyperscalers is gradually weakening.” The agency assumed the top six US hyperscalers — Amazon, Microsoft, Alphabet, Oracle, Meta and SpaceX — would spend more than US$7 trillion on data centres and AI-related capital expenditure from 2025 through 2030, with financings “becoming more complicated and less transparent” and returns taking years to show. It asked, if demand disappoints, chip values fall, or utilisation declines, “do the losses sit with the operator, the sponsor or the guarantor?”
Naveen Sarma, an S&P analyst who co-authored the report, told Axios the risks were not “existential” for most of those firms, which remain highly rated borrowers with strong cash-flowing businesses — Oracle being the notable exception. A separate S&P Global Market Intelligence note, The Price of Intelligence, said the build-out had reached historically high investment intensity but had “not yet developed the financing characteristics that made the dot-com build-out fragile.”
Credit default swaps have become a widely watched gauge of unease. The Financial Times reported that average weekly trading in CDS tied to bonds issued by companies including Oracle, SpaceX, Alphabet, Amazon, Meta, Broadcom and Nvidia exceeded US$9 billion in September, up from under US$2 billion in July, according to Morgan Stanley Research. Mohamed El-Erian flagged that FT reporting, and later described the early-November Nasdaq sell-off as a reminder of how much is levered on the AI trade — for “good reasons (productivity promise, investment scale, IPO prospects, etc.) and less good.”
Good morning.
November 6, 2026
The broader market impact of yesterday’s selloff in the NASDAQ, the largest since March 2020, is a reminder of how much is levered on the AI trade — for both good reasons (productivity promise, investment scale, IPO prospects, etc.) and less good
From the FT:
October 20, 2026
“Investors are racing to buy credit default swaps to protect their portfolios in the event the AI boom unravels….
In a sign of investors’ increasing unease over the borrowing binge, the cost of buying CDS on many AI-linked companies has hit record highs this month.”
Lisa Abramowicz of Bloomberg Surveillance quoted Apollo chief economist Torsten Sløk saying credit spreads on “all things AI” were widening enough to slow some projects: “Credit markets are beginning to vote with their feet.”
Credit spreads on all things AI are widening out enough to slow some AI projects: Apollo’s Torsten Slok. “Rates are maybe not important for AI but yields are important for AI:” Slok. “Credit markets are beginning to vote with their feet.”
October 24, 2026
The midterm shock to the buildout
Palladino is on firmer ground when she says voters pushed back on data centres. CNBC, drawing on Ballotpedia, counted 41 local ballot measures in 10 states. Voters backed restrictions or rejected pro-development measures in every contest with results then available, except Harwood, North Dakota. Gina Hinojosa defeated Texas Governor Greg Abbott, and she has pledged to keep a pause on new data-centre approvals until the legislature passes stronger protections, the Texas Tribune and AP reported. Democrats who campaigned on pauses or tighter rules also won governors’ races in Georgia and New Mexico. Elon Musk, after the results, posted:
Data centers in space looking pretty appealing right now
November 4, 2026
What trustees can, and can’t, do from here
Palladino wants pension trustees to refrain from “risky, unregulated financial assets linked to the data centre boom,” and she warns that simply moving back into public index funds would not escape AI exposure, as those indexes are already heavy in large technology companies. SpaceX, which she mentions in that context, began trading in June 2026 but, under S&P’s existing rules, was not due to be considered for the S&P 500 before June 2027, Reuters reported.
Some funds have already tightened. Pensions & Investments reported that Florida’s Jacksonville Police and Fire Retirement System withdrew from a digital-infrastructure fund, citing data-centre risks, and that the Wyoming Retirement System had paused new private-credit commitments. Reuters, in April, also found that many large North American pension funds were sticking with private credit. A 401(k) safe harbour, if and when it is published, would still leave the prudence decision with plan fiduciaries; it would not require a data-centre allocation.
Palladino’s alternative is to steer more retirement capital toward housing and renewable energy. In April, New York City Comptroller Mark Levine announced a US$4 billion, four-year Housing Investment Initiative across the city’s five pension funds, intended to finance affordable and mixed-income housing while still seeking risk-adjusted returns for beneficiaries.
The official starting points are Palladino’s LPE Project essay, AFR’s Tracing the Money report, S&P’s hyperscaler credit outlook, the White House executive order, the Labor Department’s proposed rule, and the ICI’s second-quarter retirement market data.
Discover more from Sulat AI & Tech News
Subscribe to get the latest posts sent to your email.
