Everything About AI
NextFin News - BlackRock's AI trade is no longer a bet on the Magnificent Seven. The world's largest asset manager is steering clients toward the power lines, data campuses, and private infrastructure that sit behind the screen - a shift that turns artificial intelligence from a technology bet into an energy and capital-allocation story. And after a blistering rally in 2025, the second wave has already had its momentum moment: the trade now demands conviction, not chase.
In its Fall 2026 Investment Directions, BlackRock reported that AI infrastructure companies grew earnings 54% year over year in the second quarter, compared with 14% for the S&P 500 excluding AI infrastructure. The gap captures a trade that has migrated down the value chain: from the model builders and cloud platforms that dominated 2025 to the physical assets that keep the AI economy running - and to the private funds that increasingly own them.
The Pioneers Gave Up Their Easy Alpha
The first wave of the AI trade was simple to express: buy the companies building the models and the chips. In 2025, capital flooded toward platform builders and foundation-model designers. That phase delivered extraordinary returns but also extraordinary concentration. At its peak, dispersion within the U.S. technology sector reached roughly 71% between the best- and worst-performing level-two subsectors, according to BlackRock's Autumn 2026 Investment Directions. A broad technology index had become a blunt instrument for expressing a view on AI.
Rick Rieder, who leads BlackRock's fixed-income platform, put the divergence in blunt terms earlier this year: what had been one trade for a decade - software, semiconductors, and the Nasdaq - diverged by 60 percentage points in four months. When a theme matures, returns migrate from the headline names to the constraints.
BlackRock's own client surveys show the rotation in real time. In its January 2026 Investment Directions report, which surveyed 732 EMEA-based clients, only one in five respondents said the largest U.S. technology groups remained the most compelling AI opportunity for the year ahead. By contrast, 54% backed providers of the power needed by data centers, and 37% named infrastructure as their top AI investment choice. Just 7% believed the AI theme was a bubble. Investors were not leaving AI. They were leaving the crowded front of the trade.
"It's increasingly important to risk-manage megacap and AI exposure while also capturing differentiated upside opportunities," said Ibrahim Kanan, BlackRock's head of core U.S. equity.
The earnings data confirm that the opportunity has broadened rather than faded. BlackRock's Fall 2026 outlook notes that the AI economy continues to expand as frontier-model providers grow revenue, adoption spreads, and an increasing share of compute shifts from training toward inference and real-world usage. That shift changes which companies capture the value: inference runs continuously, close to end users, and consumes power around the clock.
Power Became the Binding Constraint
The mechanism behind the rotation is physical, not financial. AI systems run continuously and require constant cooling; a data center without power is a warehouse full of idle silicon. BlackRock's research describes AI as a structural component of global power consumption rather than a short-term spike - a framing that matters because it turns a cyclical demand story into a multi-year buildout.
The scale of the constraint is large. Goldman Sachs Research raised its forecast in April 2026: global data center power demand is now expected to grow 220% from 2023 levels, reaching a record 1,350 terawatt-hours by 2030, up from the firm's earlier estimate of 175% growth. The United States will account for roughly 60% of incremental demand, with U.S. data center capacity projected to more than double to about 95 gigawatts by 2027. Data centers are projected to consume as much as 11% of U.S. electricity by 2030 in upside scenarios, compared with about 3% in 2022.
This is where the investment opportunity shifts asset classes. The beneficiaries are no longer only software multiples. They are regulated utilities with rate bases, independent power producers with merchant generation, transmission builders, and nuclear operators offering low-carbon baseload. BlackRock remains constructive on nuclear, citing strong earnings and order-book growth, and sees European infrastructure as a key beneficiary of a multi-year cycle modernizing critical assets and expanding electrification.
The capital commitment behind this buildout is already visible in corporate guidance. The four largest hyperscalers - Amazon, Microsoft, Alphabet, and Meta - have collectively guided toward roughly $700 billion to $725 billion in capital expenditure for 2026, up about 75% from the roughly $410 billion spent in 2025, with the bulk directed at AI infrastructure: GPU clusters, custom silicon, and data center construction.
The Trade Has Already Had Its Momentum Moment
Here is the uncomfortable fact that the structural thesis must survive: the utilities leg of the AI trade has already run hard and then given most of it back. The Morningstar U.S. Utilities Index rose 19% in 2025 and has surged roughly 70% from its late-2023 low. Utilities ranked as the third best-performing S&P 500 sector in the first quarter of 2026, behind only Energy and Materials. Yet as of mid-September 2026, the Utilities Select Sector SPDR Fund was up only about 2.4% year to date, with a one-year return near flat.
That reversal is the clearest test of whether this is a structural shift or a crowded rotation. A cyclical trade mean-reverts once the narrative gets priced; a structural buildout digests valuations while the underlying demand compounds. The distinction matters because it determines whether investors should buy the dip or avoid the sector entirely.
The rate channel explains part of the stall. Regulated utilities carry bond-like duration, so their 2025 rally rode both the AI power narrative and falling interest rates. When the rate path turned uncertain in 2026, the multiple compressed even as the demand story remained intact. The second wave of the AI trade, in other words, correlates with the bond market as much as with the Nasdaq - a linkage most investors did not underwrite when they bought the pioneers.
Capital Is Moving Offshore - Into Private Markets
The second structural change is where the money is being deployed. The largest AI infrastructure deals are no longer happening in public equity markets. In October 2025, a consortium including BlackRock, Microsoft, Nvidia, and Elon Musk's xAI agreed to acquire Aligned Data Centers for $40 billion - the largest global data center deal to date. The transaction was backed by BlackRock's Global Infrastructure Partners and members of the Artificial Intelligence Infrastructure Partnership, or AIP, which was created by BlackRock, MGX, Microsoft, and Nvidia in September 2024 with a goal of deploying $30 billion of equity capital. The Kuwait Investment Authority, Temasek, and xAI joined as additional participants.
"With this investment in Aligned Data Centers, we further our goal of delivering the infrastructure necessary to power the future of AI, while offering our clients attractive opportunities to participate in its growth," said Larry Fink, BlackRock's chief executive and chairman of AIP.
Aligned operates 50 campuses with more than 5 gigawatts of operational and planned capacity across North and South America. The transaction, expected to close in 2026 subject to regulatory approvals, illustrates a broader truth: the AI buildout is becoming a private-markets story. Public-market investors can own the utilities and the equipment makers, but the highest-value infrastructure assets are being absorbed by infrastructure funds and sovereign capital before retail investors ever see them.
This creates an asymmetry that defines the next phase. Public-market participants get the second-order beneficiaries - the power sellers and the grid builders - while private capital captures the direct infrastructure returns. For asset managers, the alpha in AI is migrating away from stock selection within the Magnificent Seven toward asset origination and long-duration project finance. BlackRock's answer is selectivity and active management: broad technology exposure has become a less precise way of expressing conviction in the theme, the firm wrote in its Autumn 2026 outlook, favoring strategies that identify the companies, sectors, and parts of the value chain best positioned to benefit as leadership shifts.
BlackRock's thematic ETF lineup makes the value-chain approach explicit. Its 2026 Thematic Outlook groups exposures as compute and conflict, pipes and power, and capture themes, alongside digital assets. Over the past decade, U.S. thematic funds have grown more than eleven times, reflecting investor demand for targeted expressions of conviction rather than broad index exposure.
The Second-Order Trade: AI Now Correlates With Bonds
The first-order effect of the AI buildout is higher demand for chips, servers, and data centers. The second-order effect - the one the market is still working through - is that the AI trade now transmits through the bond market. Rate cuts lower the discount rate on long-duration infrastructure cash flows, lifting utilities and transmission assets. But rate cuts also signal slowing growth, which can depress the very AI capex that drives power demand. The net effect depends on whether the market reads monetary policy as preventive or reactive.
That linkage changes portfolio construction. A portfolio heavy in AI-exposed utilities is no longer a clean hedge against a technology selloff; it is a levered bet on both AI adoption and a benign rate path. This is the practical meaning of BlackRock's push to manage AI concentration "in, around, and beyond" the theme: the same macro driver can lift the pioneers and the physical layer for different reasons, and it can knock both down for a third reason entirely.
The Counter-Thesis: A Crowded One-Way Trade
The strongest case against this rotation is that it has become crowded. Utilities and infrastructure names rallied sharply in 2025 on the AI power narrative, and much of the expected earnings growth is embedded in current valuations. If hyperscaler capital expenditure fails to convert into revenue - if AI adoption slows or monetization disappoints - the infrastructure buildout could stall, leaving rate-sensitive utilities exposed to a double hit: lower power demand and higher real rates. The 2026 stall in utility performance is the market's first hint that this risk is real.
There is also a timing risk. Building data centers and transmission networks requires steel, cement, and energy-intensive manufacturing, with permitting timelines that can delay the revenue payoff for years. The gap between upfront investment and eventual productivity gains can stretch across rate cycles, and in a higher-rate environment that gap compresses valuations. BlackRock's own survey found that only 7% of respondents view AI as a bubble - but belief in the destination is not the same as a clear path to returns.
The counter-thesis is serious but rests on one condition: that AI demand proves cyclical rather than structural. The falsifying signal is concrete. If hyperscaler AI-related capital expenditure growth decelerates for two consecutive quarters from its current trajectory, or if data center power-demand growth fails to track the projected 220% increase from 2023 levels by 2030, the structural-buildout thesis breaks and the rotation reverses. Until that signal prints, the physical constraint remains the binding fact.
What Comes Next
The evolution of BlackRock's AI trade points to three time horizons, and they do not all point the same way. In the short term, sentiment and the interest-rate path will drive the rate-sensitive utilities and infrastructure names - a liquidity trade as much as an AI trade, which is why the sector can stall even while the thesis holds. Over the medium term, earnings will separate the winners: companies with contracted power sales to hyperscalers and visible rate-base growth will outperform those trading on narrative alone. Over the long term, the structural shift is toward the physical layer - power, grids, cooling, and the private infrastructure that owns it.
Scenarios matter more than a single line. In the base case, AI adoption continues to spread from training to inference and real-world usage, power demand tracks the upward trajectory, and the infrastructure buildout compounds - the 54% earnings growth premium that AI infrastructure companies posted in the second quarter narrows but persists. In the upside case, faster-than-expected AI monetization accelerates capex beyond the current $700 billion-plus hyperscaler guidance, lifting both the pioneers and the physical layer. In the downside case, a capex slowdown or a sharper-than-expected rise in real rates re-rates the rate-sensitive infrastructure complex, and the 2026 stall becomes a trend rather than a pause.
What to watch: quarterly hyperscaler capex guidance, data center power-demand updates from grid operators, and the interest-rate path. The single most telling metric is whether AI infrastructure earnings growth sustains its premium over the broader market, and whether the Morningstar utilities index can reclaim its 2025 highs on earnings rather than multiple expansion.
The AI trade has not ended. It has moved underground - into the grid, the data campus, and the private funds that own them. The pioneers built the demand; the next winners will be those who own the constraint.
Explore more exclusive insights at nextfin.ai.
