Goldman Sachs Sees $765 Billion AI Buildout as Power, Memory and Optical Gaps Widen
Updated
Updated · The Motley Fool · Jun 22
Goldman Sachs Sees $765 Billion AI Buildout as Power, Memory and Optical Gaps Widen
3 articles · Updated · The Motley Fool · Jun 22
Summary
$765 billion in AI infrastructure spending is projected for 2026, with Goldman Sachs saying the biggest outlays will target power supply, memory chips and optical networking bottlenecks.
945 TWh of data-center electricity demand by 2030 underpins the power push, and Goldman highlights GE Vernova as a beneficiary after its power unit booked $10 billion of Q1 equipment orders and built a $163 billion backlog.
Micron stands out on memory because high-bandwidth memory is sold out through 2027 and the company is locking in multiyear strategic customer agreements, including its first five-year deal.
Optical networking could expand ninefold into a market worth more than $150 billion, Goldman says, favoring Marvell after it launched a 102.4-terabit-per-second switch and posted 28% revenue growth last quarter.
The forecast reflects how AI spending is shifting from raw compute expansion toward fixing the infrastructure constraints that determine how fast new data centers can scale.
With AI's thirst for power growing, can clean fuel cells solve the data center energy crisis faster than traditional turbines?
Is the industry's massive bet on HBM memory a costly mistake, with a superior optical alternative on the horizon?
The $7 Trillion AI Infrastructure Surge (2026–2031): Bottlenecks, Risks, and the Global Race for Compute Power
Overview
Between 2026 and 2031, global investment in AI infrastructure is set to reach an unprecedented $7.6 trillion, averaging $765 billion per year. This surge is focused on building compute power, data centers, and the electrical infrastructure needed to support them. The investment marks a fundamental shift in the economy, driving up borrowing costs and reshaping financial strategies. Projections are based on current chip sales, using NVIDIA’s data center revenue as a key indicator for future deployment of AI accelerators. These figures help estimate the scale of data centers and power generation required, highlighting the massive transformation underway.