Spark, Gigafund, Greenoaks buy AI stakes without control

A quieter shift is playing out on AI cap tables, and this week the Wall Street Journal put names to it: firms like Spark Capital, Gigafund and Greenoaks are writing very large late-stage checks into companies like Anthropic and SpaceX and, per the reporting, buying those stakes without seeking influence. As companies stay private for well over a decade, the old venture bargain of capital in exchange for board seats and governance rights is not what the biggest checks are buying anymore.

The Anthropic story is the cleanest example. Spark's Yasmin Razavi, reportedly the only outside investor on Anthropic's board, led a $75 million investment in early 2023 that the firm says was the largest check it had ever written. Investors elsewhere in Silicon Valley had passed at a $4 billion valuation on a company with no product or revenue. That stake is now reportedly worth about $7 billion at Anthropic's latest $965 billion mark. Greenoaks, led by Neil Mehta, first invested earlier this year and co-led the round at the $965 billion valuation, according to the WSJ, and now owns a multibillion-dollar position. Gigafund, launched in 2017 by ex-Founders Fund partner Luke Nosek, has taken a similar posture with SpaceX, putting more than $1 billion in.

Why this matters if you are not a growth investor: it tells you where the leverage actually sits in this cycle. Founders of scarce frontier assets can raise checks measured in billions and hand out very little in return, because there is more late-stage capital chasing Anthropic-shaped companies than there are Anthropic-shaped companies. Firms willing to accept passive economics get into the room; firms that insist on the old terms do not.

The honest caveat is that the WSJ piece is a snapshot of a valuation regime that only works if today's marks hold. Late-stage buyers writing nine and ten figures for economic exposure without governance rights inherit the downside if a frontier lab stumbles on safety, regulation, or unit economics, and it is not obvious what those investors' LPs are being paid venture-tier fees to do beyond sourcing allocation. What the reporting doesn't spell out is the fine print: the liquidation preferences, ratchets, or information rights that presumably compensate for the missing board seat.

The forward-looking question is who this pattern benefits next. If passive late-stage money is now table stakes to buy into a frontier lab, that flow eventually has to find AI companies that are not Anthropic or OpenAI, and those are the checks worth watching.