On Monday, Anthropic filed to go public. The filing, and Bloomberg's reporting last week, show who gets paid. Menlo Ventures put in more than $1 billion for a stake now worth about $14 billion. Spark Capital's $75 million check from 2023 is worth about $7 billion. It's the clearest proof yet of the strategy behind this year's biggest venture funds.

Those funds are bigger than ever. In January, Andreessen Horowitz closed $15.075 billion, the largest raise in its history, spread across growth, infrastructure, apps, American Dynamism and bio. It has since added more vehicles, taking its 2026 total past $20 billion. A month later, Thrive Capital closed Thrive X at $10 billion, double its prior fund, and reportedly turned away billions more. Kleiner Perkins followed with $3.5 billion in March. Menlo Ventures raised $3 billion in June.

Almost all of it is pointed at one category. By one live tracker's count, 88% of venture capital raised in 2026 is going into AI-focused or AI-adjacent strategies, and three firms (a16z, Thrive and General Catalyst) account for 43% of all tracked commitments. The money going out looks the same. In the third quarter, AI startups took $102 billion of the $159 billion invested globally, 64% of the total, across a record 27 rounds of $1 billion or more, according to Crunchbase.

That's the top of the market. The other end looks nothing like it.

Mega-funds exist because frontier AI eats capital. The three largest AI rounds of early 2026, OpenAI at $110 billion, Anthropic at $30 billion and xAI at $20 billion, together topped $160 billion. Only a handful of firms can write into rounds that size, and the ones that can have built their franchises around a single lab.

Thrive is the clearest case: OpenAI is described as the largest position in its portfolio. Menlo bet the firm on Anthropic, first investing in 2023 when the company was pre-product and pre-revenue, then following into later rounds. It also launched the Anthology Fund with Anthropic in 2024, which has deployed closer to $250 million across more than 60 companies.

That's the mega-fund model in 2026: own a frontier lab, then build a platform of funds, credits and access around it.

Now that model is about to pay out. Anthropic's S-1 sets up a listing as soon as November, after a last private valuation of about $965 billion. Bloomberg reports Iconiq's stake is worth about $17 billion, Menlo's about $14 billion and Spark's about $7 billion. Returns that size are what fund the next round of mega-funds.

Katie Haun left a16z to start her own fund because she knew crypto better than any generalist firm could. AI now has its own version of that move: people who built, trained or deployed AI systems leaving to run small, focused funds.

Air Street Capital. Nathan Benaich, a researcher turned investor and co-author of the annual State of AI report, closed a $232 million Fund III in March. It's the largest solo GP venture fund ever raised in Europe, meaning one decision-maker and no investment committee. Checks run from $500,000 to $15 million, and the portfolio includes Synthesia, Black Forest Labs and Poolside.

Conviction. Sarah Guo left Greylock to start Conviction in 2022 with a roughly $101 million first fund, then closed a $230 million second fund in January 2025. Portfolio names include Harvey, Sierra and Mistral.

Zero Shot. Three OpenAI alumni, including Evan Morikawa, who led applied engineering through the launches of ChatGPT and DALL-E, and Andrew Mayne, OpenAI's original prompt engineer, have held a first close toward a $100 million fund. Early checks went to Worktrace AI and Foundry Robotics.

Operator funds sell what a generalist can't: technical diligence from people who know which problems are genuinely hard and which the next model release will make obsolete, plus networks inside the labs where the best founders come from. Solo GPs add speed. With no committee, a decision can take hours instead of weeks.

The tradeoff is concentration. One person's judgment replaces a partnership's checks, and one person's career moves can disrupt the fund. That risk isn't theoretical: Meta hired Nat Friedman and Daniel Gross in 2025, two of the best-known operator investors in AI. The talent war we covered in Issues 01 and 02 reaches investors too.

In AI venture, the middle is disappearing. Know which end you're raising from, and why.

Who's writing AI checks at each end of the barbell, and what they're built for.

Fund sizes as publicly reported. Not an endorsement of any fund.

Nothing here is legal or investment advice.