AI capex lifting industrials
The AI boom is bleeding into physical infrastructure — companies that build power, glass, heavy equipment and grid gear are rallying as data-center demand grows. (youtube.com) Big-ticket commitments like Meta’s reported $21 billion AI infrastructure expansion with CoreWeave are the kind of capital spending that props up names beyond chips. (finance.yahoo.com)
The money wave around artificial intelligence is no longer stopping at chipmakers. On April 9, Meta expanded its cloud infrastructure deal with CoreWeave to $21 billion through 2032, and Yahoo Finance reported that total Meta commitments with CoreWeave now reach about $35 billion. (finance.yahoo.com) That kind of contract turns a software race into a construction race. CoreWeave said the Meta deal supports its plan to spend $30 billion to $35 billion on capital expenditures in 2026, which means more buildings, more power gear, more cooling systems, and more fiber. (finance.yahoo.com) A modern artificial intelligence data center is closer to a factory than an office park. The Financial Times reported that some new facilities coming online from 2026 will each draw at least 1 gigawatt of electricity, roughly the output of a nuclear reactor. (ft.com) That is why industrial stocks have joined the trade. Bloomberg reported in February that orders jumped at Vertiv, Eaton, and General Electric Vernova as data-center demand lifted sales of the equipment that connects these buildings to the grid. (bloomberg.com) Vertiv sells the plumbing and air conditioning of the artificial intelligence boom. Its products include power systems and cooling equipment that keep densely packed servers from overheating when thousands of graphics chips run at once. (bloomberg.com) Eaton sells switchgear, which is the industrial version of a circuit-breaker room for giant buildings. On April 8, Eaton said it would open a new Nebraska plant to make more switchgear because utilities, factories, and data centers are all competing for the same electrical hardware. (markets.ft.com) General Electric Vernova is benefiting from the same squeeze from the power-generation side. The company said hyperscale data centers were about 10% of its order book in 2024 and could rise to 25% of orders in 2026 as technology companies lock in electricity supply. (gevernova.com) Even glass gets pulled into the story because artificial intelligence clusters need huge amounts of fiber-optic cable to move data between servers. Corning said in January that 2026 would bring more “scaling up and across” by hyperscalers, meaning denser networks inside and between data centers. (corning.com) The bottleneck is no longer just chips. Bloomberg reported on April 1 that shortages of transformers, switchgear, and batteries are forcing some United States projects to rely on imports and slowing construction schedules. (bloomberg.com) Wood Mackenzie told Bloomberg in March that United States data-center development slowed as grid limits tightened, with new projects added in late 2025 using about 25 gigawatts of electricity, roughly half the prior quarter’s pace. That is the part of the artificial intelligence boom the stock market is now pricing in: not just smarter software, but a scramble for turbines, breakers, cables, and land with enough power to run them. (bloomberg.com)
