LPs have bought the wrong multiple

BAIN & COMPANY, the consultancy, calculates that sustaining the current AI investment trajectory will require roughly $500 billion in annual capital expenditure to generate about $2 trillion in revenue — a four-times revenue multiple on capital deployed. That ratio has not been demonstrated at any scale. Nor is it close to being. The combined annualised revenue of the three frontier labs — OpenAI at around $20 billion, Anthropic at $14 billion, xAI at a few billion more — sums to under 2% of the figure Bain's math requires. Yet limited partners, the pension funds and endowments and sovereigns who write the checks at the back of the venture chain, have responded to this asymmetry by accelerating their commitments to AI-focused funds, not slowing them.

In the first quarter of 2026, US venture funds raised $47.8 billion in new commitments — more than half of the total raised in any of the previous three years, according to PitchBook. A single vehicle, Thrive Capital Management's $9 billion growth fund, accounted for nearly one-fifth of the quarterly total. Behind those headlines is a quieter consolidation. Capital is concentrating into a small group of established managers with explicit AI mandates, while emerging managers face their most selective fundraising environment in years. The concentration goes further than AI itself. The capital is flowing into the few firms whose existing positions in OpenAI, Anthropic, xAI, Databricks, and Cursor make them appear to be hedged against the asymmetry their commitments are funding.

Returns to scale

Take the math an LP is implicitly underwriting when it commits to one of these vehicles. The benchmark for exemplary venture performance is three times paid-in capital — a TVPI of 3x — the multiple that justifies the asset class's illiquidity to LPs in the first place. For a $1 billion fund, that requires $3 billion in distributions. For a $9 billion growth fund, $27 billion. Power-law dynamics imply that one or two portfolio companies must carry the entire return: in the typical fund of 20 to 30 names, between one and three investments generate 50% to 80% of total proceeds, according to PitchBook. The fund-makers have to be enormous.