The world's largest state investors are quietly rotating out of public AI stocks and into the physical infrastructure that makes AI run, a move that tells you something important about where the smart money thinks the returns actually accrue.
When Abu Dhabi's MGX, BlackRock's Global Infrastructure Partners, and the AI Infrastructure Partnership , a vehicle that counts Singapore's Temasek and Kuwait's sovereign wealth fund among its backers , agreed to buy Aligned Data Centers for roughly $40 billion, it wasn't just a deal. It was a statement. The largest data center acquisition on record, beating Blackstone's $16.6 billion purchase of AirTrunk in 2024 by a factor of nearly two and a half, signals a consensus forming among the funds that manage entire national balance sheets: the infrastructure layer is where this cycle's real money lands.
The numbers behind this rotation are striking. According to data from Global SWF, allocations to private credit by sovereign wealth funds rose 22% year-over-year to $180 billion, with capital flowing into data centers, fiber networks, and the power grids that feed them. Funds overseeing more than $12 trillion in combined assets are making this pivot at the same time, not coincidentally. They're looking at public markets and seeing a problem.
The top 10 stocks in the S&P 500 now control roughly 40% of the index by market cap, a level that exceeds the dot-com peak of 2000, when that figure was closer to 27%. During a 28-session rally between late March and early May 2026, analysts at Nomura found that just 10 stocks drove 69% of the index's gains, with Nvidia, Alphabet, Amazon, Broadcom, and Microsoft carrying most of the weight. What you effectively own in a passive S&P 500 fund right now is a directional bet on AI adoption, concentrated in a handful of companies, at valuations that price in a lot of things going right.
Sovereign funds don't have the luxury of ignoring that. They have mandates, they have political accountability, and they can't easily exit a $50 billion position in Nvidia if sentiment turns. Private infrastructure assets don't mark to market in the same way. A contracted data center lease generating inflation-linked escalators isn't going to drop 30% because a competitor releases a cheaper model. That predictability is exactly what these funds are buying , and it's why the rotation is structural, not tactical.
There's a second driver that gets less attention: export controls. National security restrictions on cross-border technology investments have made it genuinely complicated for Gulf and Asian sovereign funds to take large stakes in US AI companies without triggering review. Chips, software, and frontier model companies are increasingly off-limits or politically sensitive. Physical infrastructure , power grids, fiber, cooling systems , sits in a different regulatory category. Concrete and kilowatts cross borders more cleanly than code.
Where the buildout money is actually going
The AI Infrastructure Partnership's first investment, the Aligned deal, gives you a picture of the playbook. AIP has an initial target of mobilizing $30 billion in equity, with potential to reach $100 billion including debt. McKinsey projects SWF investments in AI-related private assets could exceed $500 billion by 2028. That's a staggering number, and it implies a view: that the returns from the AI cycle will be captured not primarily by the companies writing the models, but by whoever owns the substrate those models run on.
It's worth sitting with that thesis for a moment, because it's not obvious. The conventional narrative positions Nvidia, Microsoft, and OpenAI as the defining beneficiaries of the AI buildout. But the sovereign fund bet is essentially that this is more like the railroad era than the software era. In the railroad era, the companies that made fortunes weren't always the ones running the trains , they were the ones who owned the land, the right-of-way, the stations. Data centers, fiber routes, and power purchase agreements are the 21st-century equivalent of that land.
Retail investors don't have access to most of this. The Aligned deal wasn't offered on a public exchange. The private credit funds channeling capital into fiber and power infrastructure have minimum commitments that exclude all but institutional allocators. So while passive investors in index funds are concentrated in the model makers , the publicly traded AI companies that dominate the S&P 500 , the sovereign funds are buying the picks and shovels layer at negotiated valuations with contracted cash flows. It's a bifurcation of access, and it matters.
Frankly, what this rotation tells you is that the funds with the longest time horizons and the most sophisticated risk management have already decided the public AI trade is crowded. They're not shorting it. They're just doing something smarter with their next dollar: building the roads while everyone else argues about which car company wins the race.
Also read: Momenta's Hong Kong IPO prices at HK$295.60 as Chinese autonomous driving bets on software margins over profits • Micron Technology briefly overtook Meta and Tesla in market value after revenue quadrupled on AI memory demand • Baidu's Kunlunxin is chasing a $50 billion Hong Kong IPO with a condition investors have rarely seen