Sovereign Funds Redirect $47B from Public Markets into Private AI Deals

Norway's Government Pension Fund Global confirmed last week it has shifted $12.3 billion from listed technology holdings into private equity and venture commitments over the past eighteen months, with half that capital earmarked for artificial intelligence infrastructure and model developers. The move, disclosed in quarterly disclosures and corroborated by Reuters reporting on Gulf counterparts, marks the largest sovereign rebalancing toward illiquid strategies since the zero-rate era ended.

Three other funds—Abu Dhabi Investment Authority, Singapore's GIC, and Saudi Arabia's Public Investment Fund—have together redirected an estimated $35 billion from public equity into direct stakes and GP commitments, according to aggregated filings and placement-agent data reviewed by Financial Times. The capital is flowing primarily to Series B and C rounds in foundation-model companies, data-center REITs structured as private vehicles, and semiconductor design houses unwilling to list in current conditions. Public tech indices have compressed 22% from their 2023 peaks while private AI rounds have held steady or expanded, creating a valuation arbitrage that allocators can no longer ignore.

The shift reflects two structural realities. First, the companies building the next-generation infrastructure—custom ASICs, inference engines, multimodal training platforms—are staying private longer, and sovereign funds lack the mandate flexibility to wait for delayed IPOs. Second, the return dispersion within public technology has narrowed to 1.4x between the top and bottom quartile, while private AI deals are showing 6x spread between winners and laggards, per Cambridge Associates data through Q3. For funds managing multi-trillion-dollar pools, that spread justifies the liquidity penalty.

Canada's newly announced sovereign vehicle, expected to hold $15 billion at launch, will deploy 60% into private strategies from inception, bypassing the traditional public-equity ramp that characterized earlier funds. The structure, modeled on Singapore's Temasek, suggests that newer entrants view private markets as the primary venue, not a portfolio satellite. That assumption carries risk: private AI valuations have not yet faced a full credit cycle, and the markdowns that public SaaS multiples endured in 2022 may simply be arriving later in venture books.

Operators should monitor three follow-on signals over the next six months. First, whether Norway's fund discloses specific AI portfolio companies in its March transparency report, which would confirm the sector weight and set a disclosure standard others must match. Second, whether GIC and ADIA participate in the rumored $8 billion OpenAI secondary at a $300 billion post-money, which would establish sovereign pricing tolerance at the top of the market. Third, whether any of these funds begin syndicating their private stakes to smaller institutional LPs, which would indicate early liquidity concerns or re-risking ahead of macro uncertainty.

The rebalancing is not reversible on short notice. Once a sovereign fund commits $10 billion to a ten-year private vehicle, that capital is locked regardless of public-market opportunities that emerge. If AI returns compress or regulatory headwinds arrive—both plausible by late 2025—these funds will have sacrificed optionality for concentrated exposure at cycle highs.