JPMorgan Says This Is Where You Should Be Investing in AI

JPMorgan Chase (NYSE:JPM) just issued its Outlook 2026 report that frames artificial intelligence (AI) as a multi-decade productivity revolution on par with electricity or the internet, with no evidence of a bubble forming yet. Massive tailwinds are pushing the technology forward as supply constraints remain severe. For example, data center vacancy rates sit at 1% to 2%, new capacity is 75% to 100% pre-leased years in advance, and leading-edge chips are sold out 12 to 18 months ahead. Capital spending by the tech giants has tripled since 2023 and is forecast to top $500 billion annually by 2026 — nearly a quarter of all U.S. corporate capex. Power demand also stands out as the sharpest bottleneck, with AI projected to drive an additional 662 terawatt hours (TWh) of U.S. electricity consumption by 2030, equivalent to adding several new Texases to the grid. Transmission backlogs exceed five years, and 70% of lines are over 25 years old. While knowledge-worker disruption is real, the bank expects net job gains and 30% productivity increases in impacted sectors. JPMorgan’s core advice: overweight large-cap AI leaders, target the physical supply chain (chips and power first), embrace adopters across industries, and use private markets for later-stage application plays. Public equities offer the clearest exposure today through the infrastructure layers that must scale before broader monetization arrives.

Semiconductors: Still the Tightest Link in the Chain

JPMorgan repeatedly stresses that chip supply, not demand, will gate the entire buildout for years. Advanced nodes are capacity-constrained, hyperscalers are locking in multi-year deals, and foundry expansion lags far behind announced demand. Nvidia (NASDAQ:NVDA) remains the undisputed leader — its platforms train and run over 90% of cutting-edge models, Blackwell chips are booked solid through 2026, and it ranks among the handful of companies driving the capex surge. The report specifically calls out Nvidia’s near-monopoly in accelerated computing and its ability to command premium pricing in a supply-starved market. Broadcom (NASDAQ:AVGO) complements this perfectly with custom AI accelerators (Google, Meta Platforms (NASDAQ:META), and ByteDance are major clients) and high-bandwidth networking silicon. JPMorgan points to the explosion in optical and co-packaged optics demand as AI clusters scale to hundreds of thousands of GPUs. Broadcom is the only merchant supplier at scale for many of these critical components, giving it years of secured growth.

Power and Data Centers: The Real Bottleneck

The report calls energy “the binding constraint” multiple times, noting multi-year permitting delays, retiring coal plants, and surging loads from AI clusters. Nuclear restarts, natural gas buildout, and grid upgrades will be required at unprecedented scale. Vistra (NYSE:VST) aligns tightly with this thesis: it is the largest competitive power generator in Texas and the Midwest, with a mix of gas, nuclear, and battery storage located exactly where hyperscalers are clustering data centers and signing 10- to 20-year direct power purchase agreements. Vistra has already announced multiple hyperscaler deals and is restarting units to meet AI-specific load — exactly the type of flexible, merchant exposure JPMorgan says will capture massive pricing power as wholesale rates spike. IREN (NASDAQ:IREN) represents the entrepreneurial edge of the same thesis. By building data centers on stranded or underutilized renewable and gas sites in Texas, British Columbia, and soon other grids, IREN can offer immediate, low-cost megawatts that hyperscalers desperately need while traditional utilities wait in five-plus-year interconnection queues. Its ability to pivot existing Bitcoin (CRYPTO:BTC) mining load to cloud or high-performance computing (HPC) contracts overnight gives it optionality that pure utilities lack, and its 816 megawatts (MW) of secured power capacity (with more under contract) positions it as one of the fastest-scaling pure-play AI data-center power providers in public markets.

Hyperscalers: Safe Scale Exposure Until Infrastructure Catches Up

Although value is gradually shifting downstream to applications and private companies, the mega-cap cloud providers remain central for now. Microsoft (NASDAQ:MSFT) and Amazon (NASDAQ:AMZN) via AWS are cited for their massive AI-infused cloud growth and capex commitments, with pre-leased facilities and vertical integration giving them durable advantages during the supply crunch. These two (along with Google) are absorbing the majority of new chip and power capacity, translating directly into accelerating cloud revenue. Microsoft’s total commercial bookings alone are now running at a $100 billion-plus annual run rate.

Key Takeaways

JPMorgan’s clearest message is that the next several years belong to the physical enablers — semiconductors and especially power — because these layers must expand before software and services can fully capture value. Nvidia and Broadcom are riding the chip shortage with no relief in sight through at least 2028, while Vistra and IREN sit at the heart of the power bottleneck the bank identifies as the single largest hurdle to AI scaling. With hyperscalers committing trillions over the coming decade, chronic undersupply in both chips and electrons, and regulatory timelines measured in half-decades, these infrastructure providers are positioned for multi-year pricing power and earnings growth that most application-layer companies simply cannot access until the physical buildout catches up. For investors, these stocks are the most direct and defensible way to ride what JPMorgan sees as a once-in-a-generation supercycle.