For most of last week, AI investors had a puzzle. Leopold Aschenbrenner's hedge fund Situational Awareness, days after margin calls nearly destroyed it, had put $400 million into a private company that nobody involved would name. Bloomberg reported the existence of the deal without identifying the recipient. On Friday, The Wall Street Journal named it: Source Foundry, a startup founded in 2025 by two Stanford researchers that is building equipment to compete with ASML.

The Journal reported that the fresh money brings Situational Awareness's total position in Source Foundry to $500 million, on top of an earlier $100 million, and that the year-old company was valued at about $5 billion in a recent funding round. Sequoia Capital had already backed it. Bloomberg later confirmed that Source Foundry was the recipient and noted that the Journal named the company first.

The timing is the reason anyone cares. CNBC first reported the forced sale of the fund's entire public stock book to Ken Griffin's Citadel, at a discount, after prime brokers issued margin calls on a portfolio built around AI infrastructure names. Assets fell from roughly $45 billion at the start of July to about $10 billion. The fund lost about 67% for the month.

Aschenbrenner had signaled strain before that. In a letter to investors dated July 24, he wrote that the fund had not been immune to market swings, particularly in Asia, and opened an August 1 window for backers to add money, as the firm sought new capital after the AI selloff. He also told them his long-term view of AI had not changed.

What the reveal changes is the category of the bet. The fund's known private holdings have been AI companies: Anthropic, Fluidstack, MatX. Source Foundry sits a layer beneath all of them, in lithography, the manufacturing step where light is used to print circuit patterns onto silicon wafers. The startup is "developing new lithography and manufacturing equipment aimed at challenging ASML." The old bet was on the firms that build AI models and the chips that run them. The new one is on the machine tools without which none of those chips get made at all.

What Source Foundry is aiming at

ASML, based in Veldhoven in the Netherlands, is the only company that sells extreme ultraviolet lithography systems, the tools required to produce the most advanced chips. One machine can cost more than $400 million, roughly the size of the check Situational Awareness just wrote. ASML's full-year 2025 results showed €32.7 billion in net sales and €9.6 billion in net income. The company has guided to between €34 billion and €39 billion in sales for 2026.

Sequoia partner Stephanie Zhan has framed the target in supply chain terms. Tracing the chain upstream from models to chips to the machines that manufacture them, she said, each layer becomes more critical and more constrained. Source Foundry, in her description, goes after "the tightest bottleneck: tooling for semiconductor manufacturing, starting in lithography."

Three other startups are chasing the same target

Source Foundry is not alone. Substrate, based in San Francisco, has raised $100 million for lithography tools and plans to run its own fabs rather than sell equipment, aiming at mass production by 2028. Inversion Semiconductor and xLight are both working on light sources built around shrunken particle accelerators. None of the three has shipped a production tool.

xLight has the most official backing. Former Intel chief executive Pat Gelsinger is its executive chairman, and the US government agreed to take an equity stake worth up to $150 million under the CHIPS Act, a deal The Wall Street Journal first reported. ASML chief executive Christophe Fouquet has said his company is working with xLight on technology demonstrations. That is a useful reminder of how this industry usually resolves: ASML bought Cymer, its EUV laser supplier, in 2012 rather than compete with it.

Source Foundry has been in stealth since it was incorporated in California in July 2025, and the reporting does not establish that it has a working product. It is not public which manufacturing step the company is attacking first, whether any chipmaker has agreed to evaluate its tools, or whether last week's $400 million was new money into the business or a purchase of shares from an existing holder. Neither firm has commented.

Building a commercially viable lithography system takes years, and the field's history is one of long timelines and absorbed challengers. The more interesting thing about this check is not the $5 billion valuation but the shape of the fund behind it: a vehicle that just lost two thirds of its value in public markets and is now concentrating what remains in private positions that do not get repriced on a bad Tuesday. That solves a liquidity problem. It does not solve a physics problem.