A Fund That Almost Broke Just Doubled Down On Chips

A Fund That Nearly Broke Just Made Its Biggest Private Bet

Three weeks ago Situational Awareness, the AI-focused hedge fund run by former OpenAI researcher Leopold Aschenbrenner, was selling. Sharp declines across its publicly traded technology holdings triggered margin and collateral calls from its bank counterparties at the end of July 2026, and the fund moved the bulk of its public stock book off its own balance sheet to meet them. That scare is now a matter of record, covered in detail elsewhere. What happened next is the part worth sitting with.

In the first week of August, Bloomberg reported that Situational Awareness had quietly put an additional 400 million dollars into Source Foundry, a stealth semiconductor-manufacturing startup. TechCrunch's independent reporting the following day confirmed the figure and added that the fund's total stake in the company now stands at roughly 500 million dollars. A fund that had just been forced to sell what it could sell went looking for more of what it could not.

Four Hundred Million More Into A Stealth Chip Startup

Source Foundry was founded last year by two Stanford researchers, Abdulmalik Obaid and Joe Burg, and counts Sequoia Capital among its backers. The company is reportedly valued at around 5 billion dollars. Its pitch is narrow and specific: a semiconductor-manufacturing process designed to be simpler, cheaper, and faster to bring online than the fabrication methods the industry already runs on.

That pitch targets a gap that has become the loudest complaint in AI infrastructure circles - demand for AI chips keeps climbing faster than the industry's capacity to actually manufacture them can expand. Design houses can draw a faster chip long before a fab can be built or retooled to produce it at volume. Source Foundry is betting that the bottleneck sits in manufacturing, not design, and that whoever shortens the path from blueprint to wafer captures a disproportionate share of the value AI chip demand creates over the next decade.

Why Liquid Positions Get Sold And Illiquid Ones Do Not

The distinction that matters here is not chips versus power, or one company versus another. It is liquid versus illiquid. Situational Awareness sold its public equity book at the end of July because it was public: shares trade every day, a bank can mark them to market every day, and a margin call has to be met with something that can be turned into cash before the market closes. A private stake in a two-year-old chip startup cannot be sold that way at any price, on any timeline that would matter to a margin clerk.

That is precisely why the size of this new commitment is informative. Adding 400 million dollars to an illiquid position is not something a fund does reflexively in the days after a liquidity scare. It is a decision made with the knowledge that the money will very likely be locked up through whatever comes next. Reading the two events side by side, the public sale looks less like a loss of conviction and more like a fund protecting its most liquid, most easily priced assets first, while its private conviction moved in the opposite direction.

Conviction Signal Or Concentrated Risk - Both Readings Hold

There are two honest ways to read this. The first is that Situational Awareness has genuine, well-researched conviction that chip-manufacturing capacity is where the AI buildout's next bottleneck will resolve, and that Source Foundry is positioned to capture a meaningful share of that resolution. Doubling down on a private bet within days of a public near-miss is what real conviction looks like when it is not performative.

The second reading is less comfortable. A fund that just proved its public book can be forced to unwind in a matter of days has now added half a billion dollars to a position that, by its nature, cannot be unwound in a matter of days at all if trouble returns. Concentrating exposure in the one asset class that offers no exit ramp is not obviously safer than the exposure the fund just walked away from, even if the underlying business case is sound. Both readings can be true of the same trade, and only Situational Awareness knows which one is doing more of the work.

What Owners With AI-Linked Exposure Should Ask Next

The practical lesson is not about this one fund. It is about what a single public headline actually tells you. A forced sale, a margin call, a near-collapse that makes the news is visible by design - the trade happens on an exchange, the price is public, and the story writes itself. A 400 million dollar top-up to a private, illiquid position is nearly invisible by comparison. It surfaces only when a reporter finds it, weeks after the money moved.

If your business depends on a vendor, a fund, or a counterparty whose capital comes from a concentrated, leveraged AI-focused pool, a single bad-week headline should prompt a more specific question than whether they survived it. Ask about the shape of their commitments: how much sits in positions they could sell tomorrow if they had to, and how much sits in positions that would stay locked in place no matter what the market does next. The answer to that question tells you more about long-run counterparty stability than any one week's stock chart ever will.

Read next: Valar Atomics Raises $1 Billion for Reactor Fleets  |  A 10 Kilowatt Reactor Just Raised a Billion Dollars