More than 75% of limited partners intend to allocate to AI in the next 12 months, four times the rate for blockchain. AI venture funds are posting double-digit returns. And yet, according to a new S&P Global Market Intelligence report, those returns are “built almost entirely on paper valuations and mark-to-market gains, not cash distributions.”
The exit window is closed. The capital keeps coming in. Something has to give.
The Numbers Tell a Specific Story
Over the past three quarters, deal counts have fallen 23% while total funding has more than tripled, according to the S&P report. Average investment sizes are climbing rapidly as capital concentrates in fewer companies. 82% of new capital flows into mega-deals above $1 billion, primarily U.S.-based platforms.
“We’re witnessing unprecedented investor conviction colliding with a closed exit window,” Ilja Hauerhof, New Product Development Director for Private Markets at S&P Global Market Intelligence, told Benzinga. “Fresh capital is moving off the sidelines, supporting large AI funding rounds. The structural imbalance means capital is entering faster than it is leaving.”
Aside from xAI’s $250 billion acquisition by SpaceX, overall exit activity remains subdued. No significant AI IPOs. Limited M&A. LPs are leaning on interim performance signals instead of waiting for liquidity events.
Capital Export From Europe to the U.S.
The geographic concentration is accelerating. Crunchbase data shows U.S. companies captured $39 billion of April’s $56 billion in global venture funding, approximately 70%.
“The AI investment boom is triggering a capital export from Europe and the UK directly into U.S. markets,” Hauerhof said. “Investors are crossing borders to chase American mega-deals.”
This is not a general market rotation. It is capital flowing toward a specific cluster of five hyperscaler-backed companies that absorbed 60% of all 2026 venture funding through April, according to Crunchbase. The rest of the ecosystem competes for the remaining 40%.
What Breaks First
The structural question is not whether AI deserves investment conviction. It is whether conviction alone can sustain a market where capital enters faster than it exits. Several pressure points are visible:
LPs evaluating fund performance are seeing double-digit returns that exist only as markups on paper. If exit markets stay closed through 2026, those markups face revaluation pressure as fund life cycles mature. The companies sitting at $10-15 billion valuations (Anthropic, Sierra, others in this tier) need IPO windows or strategic acquisitions to convert paper to cash.
The paradox for agent infrastructure companies is sharper. Crunchbase reports that nearly half of U.S. GDP growth in Q1 2026 came from AI buildout. Enterprise spending on AI is real. Revenue is growing. But the traditional liquidity path (IPO) requires public market appetite for companies burning capital to capture market position, and that appetite has not materialized for AI-native companies at current valuations.
The bridge from “AI intent” to “AI liquidity” depends on exit markets reopening, as S&P notes. Until they do, the sector runs on conviction and paper performance. That works until it doesn’t.