Lex

How AI mega-rounds make life harder for VC firms

Lex argues the $3.5tn venture capital industry has a size problem: valuations are ballooning while deal counts shrink. Per CB Insights, rounds above $100mn made up four-fifths of global fundraising in Q3 2026, as AI start-ups suck the oxygen out of the market. That strains the VC “power law” (roughly nine duds per success): AI companies need bigger cheques and repeat fundraisings, forcing investors to put more eggs in one basket or look to other sectors. Money still flows to those who can command large sums: Kleiner Perkins raised $3.5bn across two AI funds this year, and Andreessen Horowitz topped off $10bn-plus of AI funds with a $1.1bn Machine Age fund for AI’s physical build-out. Traditional VCs also share deals with newcomers such as Nvidia, which has taken stakes in Doctolib and Cloverleaf, offering silicon access as well as cash. Smaller firms face thinner returns; options include cutting costs, “solo capitalists” offering fast decisions, using AI to reduce training and legal costs, or backing speculative AI neo-labs. Fear of missing out is amplified, helping founders win fat cheques while eroding investors’ diversification, which looks riskier as assets stay private longer than under the old model.

Why it matters

Core Dealroom thesis on how AI mega-rounds reshape fund construction and portfolio risk: concentration stats, cheque sizes, and the squeeze on diversification — clean versus TI and Daily Deal coverage of individual AI labs.

Executive takeaways

Concentration: CB Insights — rounds >$100mn = ~four-fifths of global Q3 2026 fundraising; AI start-ups “suck the oxygen” out of the market. | • Industry scale: VC industry framed as $3.5tn with a “size problem” (valuations up, deal counts down). | • Power-law strain: classic ~nine duds per success harder when AI needs bigger cheques and repeat rounds; more eggs in fewer baskets or shift to other sectors. | • Who can play: Kleiner Perkins $3.5bn across two AI funds; a16z $10bn+ AI funds topped with $1.1bn Machine Age (physical AI build-out); Nvidia co-investing (Doctolib, Cloverleaf) with silicon access as well as cash. | • Small-firm outs: cut costs; solo capitalists for speed; AI to cut training/legal costs; speculative AI neo-labs. | • Caveat Lex itself stresses: FOMO helps founders; diversification erodes and looks riskier as assets stay private longer than under the old model.

What Financial Times may be missing

No European vs US VC split or named European fund pressure beyond Nvidia’s Doctolib stake — Dealroom’s Europe read-across is largely inferred. No failure-mode sizing if a mega-bet flops (LP clawbacks, DPI hit, fund recycling). CB Insights “four-fifths” is fundraising share, not deal-count share — the piece does not give the complementary deal-count or median-cheque series. Related FT Business Insight “Has AI broken the old VC model?” (Richard Waters) was flagged on-page but not deepread here, so not added.

Read the full article: Financial Times

More top stories