# Five stocks riding AI's supply chain shocks > As Big Tech spends trillions on data centers, bottlenecks in power, fiber, memory, and chips are creating temporary pricing power for suppliers. - Title: Five stocks riding AI's supply chain shocks · Kiplinger - Summary: As Big Tech spends trillions on data centers, bottlenecks in power, fiber, memory, and chips are creating temporary pricing power for suppliers. AI… - Keywords: artificial intelligence, capital expenditure, data centers, supply chain, ge vernova, corning, micron technology, nebius group, astera labs, semiconductor shortage, infrastructure spending, demand shocks - Source: Kiplinger — https://kiplinger.com/investing/stocks/stocks-to-buy-before-the-next-ai-demand-shock-hits - Published: 2026-09-24T16:12:33+00:00 - Read time: 3 min - Topics: artificial intelligence, capital expenditure, data centers, supply chain, ge vernova, corning, micron technology, nebius group, astera labs, semiconductor shortage, infrastructure spending, demand shocks ## The AI spending explosion Big Tech's combined capital spending on data centers jumped from $142 billion in 2022 to around $600 billion this year. Alphabet, Microsoft, Amazon and Meta alone will spend $5.3 trillion on capital expenditures between 2025 and 2030. ## Decades of underbuilt capacity The US share of global semiconductor manufacturing collapsed from 37% in 1990 to 12% by 2020. Over three decades, America outsourced power generation, manufacturing, packaging, and precision optics capacity to other countries. ## Demand shocks move through supply chain When trillions in fresh demand meets decades of underbuilt capacity, bottlenecks move sequentially through the supply chain. Each bottleneck temporarily gives pricing power to whichever supplier sits in it. ## GPU bottleneck already passed Nvidia's data center revenue hit $194 billion, up 68% year-over-year, with gross margins above 75%. The shock then moved to server assemblers like Dell and Super Micro. ## Power is the current bottleneck Constellation Energy signed long-term contracts at $30 to $40 above regional electricity rates, with adjusted earnings up 25% last quarter. AI data centers need gigawatts of new capacity that take years to build. ## Memory is next in line Micron's high-bandwidth memory is sold out through year-end, with gross margins rocketing to 85% from below 40% a year ago. Samsung and SK Hynix expect the memory shortage to persist into 2027 or later. ## Optical connectivity emerging The shock is now moving into optical connectivity and specialized chips that shuttle data between GPUs inside server racks. This bottleneck is just getting started. ## Spending will keep rising Unlike prior infrastructure booms, AI data center spending requires equipment replacement every four to six years indefinitely. PwC projects annual spending climbing from $800 billion this year to $1.8 trillion by mid-century. ## GE Vernova: power equipment maker GE Vernova makes gas turbines and grid equipment utilities need to add power capacity. Power segment orders jumped 134% and Electrification orders up 66%, with order backlog expected to cross $200 billion by end of 2027. > Microsoft has already lined up seven of the company's gas turbines for a Texas data center project alongside Chevron. ## Corning: fiber optic supplier Corning invents more fiber-optic cable than any company on Earth. AI data center sales nearly doubled last quarter, and the company locked in a $6 billion multiyear deal to wire Meta's next-generation AI data centers. ## Micron: memory chip leader Micron's quarterly revenue hit $41.5 billion, up 346% year-over-year, with gross margin at nearly 85%. High-bandwidth memory capacity is sold out through year-end and the shortage is expected to persist into 2027. > No sector illustrates "sell everything you make at whatever price you want" better than memory chips right now. ## Nebius: GPU rental platform Nebius rents GPU clusters to companies that need them, a model Wall Street calls a neocloud. Revenue jumped 454% to $582 million last quarter, and Nvidia invested $2 billion directly into the company. > The constraint isn't finding customers, it's building fast enough to keep up with them. ## Astera Labs: chip interconnect Astera Labs makes chips that connect GPUs inside server racks as they pack in more capacity. Quarterly revenue hit $392.4 million, up 104%, and management guided next quarter 30% above analyst expectations. ## Valuations already reflect good times All five stocks have already had big runs and trade at valuations assuming no hiccups. Micron, Nebius and GE Vernova are particularly exposed to this risk if spending slows. ## Buy in stages, not all at once The case for owning these names is that $31.6 trillion in AI spending over 25 years is nowhere near finished. Each new leg of buildout crowns a fresh winner, but demand shocks eventually resolve as new supply comes online. ## Own stocks while imbalance favors suppliers Buy in stages rather than all at once, keep position sizes reasonable given volatility, and remember that demand shocks resolve as new supply gets built. The goal is to own these while the imbalance still works in their favor. ## Key takeaway AI infrastructure spending will reach $31.6 trillion through 2050, creating a series of supply bottlenecks that reward different suppliers in sequence. Buy into these shocks while imbalances favor suppliers, not after they resolve.