BlackRock Shifted Investment Focus to Physical AI Assets
The firm now prioritizes data center and grid infrastructure as software AI models face potential commoditization.
Updated on Sept. 29, 2026 in Data Centers
Live Poll
Is now a good time to prioritize hardware manufacturers over software companies for AI investments?
As of September 29, 2026, BlackRock shifted its investment recommendation toward physical infrastructure stocks rather than software AI models. The firm identified persistent resource bottlenecks, including limited memory and base-load energy availability, as the primary drivers of this transition.
Why it matters
This realignment reflects a broader market transition where AI value is migrating from software toward the scarce hardware required to support extreme computational density. The shift underscores the physical constraints currently limiting the pace of the global AI buildout.
System memory capacity remains sold out through 2026, while transmission lines for the power grid require up to a decade to complete. These constraints limit the density of AI data centers, which currently suffer from thermal throttling that impedes optimal processing speeds.
The players
BlackRock
A global investment management corporation that oversees massive asset portfolios and provides market analysis on infrastructure and technology sectors.
Micron Technology
A semiconductor company specializing in memory and storage solutions that serves as a key supplier for high-density computing.
Vertiv Holdings
A provider of critical digital infrastructure and cooling systems designed for data centers and power management.
Constellation Energy
A major American energy company and the largest producer of carbon-free electricity in the U.S., focusing on nuclear power base-load supply.
The details
The AI buildout relies on specialized HBM (High Bandwidth Memory — a type of high-speed computer memory) that is currently supply-constrained. Hyperscalers are responding by signing Power Purchase Agreements (contracts to buy electricity from specific generators) with nuclear operators to secure reliable base-load energy, bypassing traditional market volatility. These infrastructure-heavy investments seek to mitigate the heat-related compute bottlenecks that limit processing performance in modern data centers.
Timeline
First seven months of 2026: Chinese electronic-device maker profits rose 110%.
Through 2026: HBM memory capacity remains sold out.
September 29, 2026: BlackRock released its updated investment recommendation.
The Tech Race
This pivot reflects a transition similar to the historical industrialization of power grid expansion, where physical limitations dictate the pace of digital growth. It marks a departure from pure-play software investment models in favor of the physical capital required to host large-scale compute.
Investors and industry participants should anticipate potential shifts in sector valuations as capital migrates from software firms to hardware and energy providers. The prolonged lead times for grid infrastructure suggest that power capacity will remain a defining constraint for high-density AI operations for years.
The takeaway
The valuation of the AI market is increasingly tied to energy stability and memory supply rather than pure software development. Watch for future capital expenditure reports from data center providers to confirm if this infrastructure-first approach accelerates performance benchmarks in 2027.
Further reading
For more information on the infrastructure requirements of modern computing, visit the Data Centers section.
Live Poll
Is now a good time to prioritize hardware manufacturers over software companies for AI investments?
