BlackRock is advising investors to target physical infrastructure rather than software as the artificial intelligence buildout speeds up. The firm notes that value is shifting toward computing power, data centers, and electricity.

The thesis puts a spotlight on bottleneck equities in the semiconductor, data center infrastructure, and power sectors, such as Micron Technology Inc. (NASDAQ:MU), Vertiv Holdings Co. (NYSE:VRT), and Constellation Energy Corp. (NASDAQ:CEG).

Commoditizing the Model Market

BlackRock warns, in its weekly market commentary, that the software side of AI may face pricing pressures. The firm states that cheaper models “could accelerate AI adoption while commoditizing the model market.”

As a result, BlackRock predicts this dynamic “could shift value toward the physical infrastructure needed to support greater AI use computing power, data centers and electricity.”

The firm refers to this shift as “physical AI.” BlackRock notes this transition could “amplify demand for power electronics and batteries while broadening demand for components like motors and actuators.”

Targeting AI Bottleneck Opportunities

Instead of broad market exposure, BlackRock recommends a selective approach. The firm is currently “overweight U.S. and EM equities” with a specific directive to “focus on AI bottleneck opportunities: power, chips and data centers.”

Hardware suppliers in this space are already seeing margin expansion. BlackRock highlights that “profits at Chinese electronic-device makers rose 110% year over year in the first seven months of 2026, supported by global AI demand.”

This data reinforces BlackRock’s strategy of favoring areas where “growth is translating into stronger margins and returns – particularly in AI hardware and select advanced manufacturing.”

The Scarcity Investment Theme

BlackRock ties this hardware demand to its broader “scarcity investment theme.” The firm observes that the global AI buildout is “speeding up, making bottlenecks binding.”

While the firm acknowledges China’s “manufacturing and supply-chain advantage” in physical AI components, BlackRock remains neutral on Chinese equities due to intense domestic competition.

Furthermore, BlackRock cautions that “export controls and efforts to reduce reliance on Chinese technology could limit market access.”

Why MU, VRT, and CEG Represent ‘Physical AI’ Bottlenecks

• Micron Technology – The Memory Scarcity: As AI shifts from training to continuous inference, system memory has become a critical constraint, specifically High Bandwidth Memory (HBM) required for AI accelerators. With HBM capacity largely sold out through 2026 and manufacturing expansion requiring multi-year lead times, memory suppliers like Micron are positioned at the center of a structural supply-demand imbalance.
• Vertiv Holding – The Thermal Limit: The extreme computational density of AI data centers generates localized heat that throttles processing speed, making thermal management as crucial as chip supply. Vertiv targets this precise bottleneck by providing the essential liquid cooling systems and power management infrastructure required to safely operate hyperscale AI facilities.
• Constellation Energy – The Base-Load Power Deficit: AI data centers consume massive amounts of electricity and require 24/7 reliability that intermittent renewable sources struggle to provide. Because building new power grid transmission lines can take up to a decade, existing nuclear operators like Constellation possess significant pricing power, prompting hyperscalers to sign decades-long Power Purchase Agreements (PPAs) to secure reliable base-load energy.

Note: Performance data as of Sept. 28, 2026