Steve Eisman’s AI Stock Exits Point to a Capital Raise Problem

On June 2, 2026, Alphabet (GOOG) priced an $84.75 billion equity raise to expand AI infrastructure and compute, the SEC filing shows. Steve Eisman’s reported exits from Alphabet, Meta Platforms (META) and Oracle (ORCL) should be read through that fact first.

The tension is specific: valuable infrastructure can still become a worse stock bargain if shareholders fund the next dollar of compute.

Key facts at a glance

Alphabet’s 2025 earnings release, CNBC’s April 2026 earnings report, Alphabet’s June 2026 offering filing and public market charts for GOOG, META, MSFT and ORCL set up the contradiction.

Fact	What it proves	What it does not prove
GOOG spent $91.45 billion on property and equipment in 2025.	The old AI build was huge, but still internally funded.	That returns were already bad.
GOOG guided to $180 billion to $190 billion of 2026 capex.	The spend roughly doubled.	That the spending is irrational.
GOOG priced an $84.75 billion equity raise on June 2, 2026.	Shareholders are now part of the funding plan.	That Alphabet is distressed.
GOOG closed at $356.18, META at $582.90, MSFT at $390.49 and ORCL at $140.27 on July 2, 2026, according to public market-price data.	The stocks are still liquid, large-cap AI trades.	That the market has settled the return question.

The table does one job: it separates the funding fact from the investment conclusion. The funding fact is visible. The return conclusion is still open.

What happened

Eisman has been skeptical of the AI software trade, and Fortune’s June 27, 2026 report described his argument around capital spending, competition and returns. A YouTube short tied the discussion to three AI positions sold from his portfolio, with the fuller list behind the premium episode.

The important part is not the drama of the sale. It is the sequence. Alphabet moved from $91.45 billion of 2025 property and equipment purchases to 2026 capex guidance near $180 billion to $190 billion, then priced a giant equity raise for AI infrastructure.

That is a different company behavior.

Why it matters

The old hyperscaler bargain was simple. They made enormous cash, spent some of it on growth, bought back stock, and did not need shareholders to fund the machine directly.

That model is under pressure. Alphabet’s offering filing says proceeds are for general corporate purposes, including capital expenditures to scale AI infrastructure and global compute. It also says Berkshire Hathaway agreed to invest $10 billion in the private placement.

So this is not a bankruptcy signal. It is cleaner than that. The bulls say the spending builds the next moat. Maybe. But once a cash machine sells stock to fund compute, the shareholder is no longer just collecting the upside.

The shareholder is helping finance the experiment.

What this does not tell you

This does not prove Eisman is right. Berkshire’s $10 billion anchor investment cuts directly against the lazy version of the bear case. If the returns are good enough, external capital can be smart.

It also does not confirm every ticker-level detail from free public sources. Public reporting supports Eisman’s AI capex concern, and the public clip confirms three AI sales, but the complete position list sits in the premium episode.

The verdict is narrower. Revenue growth alone is not the question anymore. Returns on the spend are.