Steve Eisman likens Meta’s $27B AI data-center financing to Enron, while another AI investor pushes back

Meta Platforms is facing a renewed debate about how large AI infrastructure spending is financed, after investors weighed in on a reported $27 billion AI data-center arrangement that one analyst characterized as off-balance-sheet. Steve Eisman, known for his Enron-related bets popularized in Michael Lewis’s “Big Short,” said he saw echoes of Enron in the way the deal was structured, according to a discussion circulated by Yahoo Finance through Benzinga’s markets coverage.

Eisman’s point was not that Meta’s business model is the same as Enron’s, but that the financing approach can obscure the true economic exposure of major commitments, at least from the perspective of traditional balance-sheet metrics. In other words, off-balance-sheet financing can shift how liabilities appear in reported accounts, even when the underlying obligations can remain meaningful to a company’s future cash needs.

Joseph Carlson, an investor associated with YouTube, disagreed with Eisman’s framing. Carlson argued that the Enron comparison was unfair, suggesting that the analogy does not hold once the details of what is being financed, how contracts work, and what risks are actually embedded are taken into account. The exchange underscores a familiar divide in market commentary: whether historical cautionary tales should be applied to newer infrastructure financings without a close review of the legal and economic terms.

The reported figure at the center of the debate is a $27 billion AI data center initiative attributed to Meta. The coverage described the arrangement as “off-balance-sheet,” a term investors use for financing structures where some obligations may not be shown as direct debt on the company’s balance sheet. Companies often use such structures for a range of reasons, including capital-efficiency targets and flexibility in how long-lived assets are built and operated, though the trade-offs can include reduced transparency for certain accounting presentations.

Meta did not disclose, in the material referenced by the report, the full set of financing mechanics that would be needed to validate or refute the Enron analogy on purely technical grounds. Details that could matter include whether obligations are truly non-recourse, how termination or restructuring risk is allocated, what future minimum payments exist, and how the cash flows map to performance milestones. Without those specifics in the cited coverage, the disagreement between Eisman and Carlson appears focused on interpretation rather than on a line-by-line accounting audit.

In the broader technology sector, the issue is timely. AI data centers are expensive, long-duration assets, and the companies building them often want to scale quickly while managing near-term capital impacts. That can put pressure on investors and analysts to look beyond headline spending figures and scrutinize the financing structures behind capacity expansion, including leases, special purpose vehicles, partnerships, and other arrangements that can affect how obligations are reflected in financial statements.

What to watch next is whether Meta’s public filings and investor communications, including any disclosures related to long-term commitments, lease arrangements, or off-balance-sheet commitments, provide clarity that either narrows or expands the gap between the competing views. Investors are likely to look for language that quantifies future obligations and clarifies the economic responsibility tied to the AI infrastructure commitments described in the coverage.