$600B AI Capex Arms Race: Hyperscalers Rattle Markets as Pick-and-Shovel Stocks Stand Out

AI Capex Just Went Into Overdrive. Here's What Builders, Operators, and Investors Should Watch

Artificial intelligence is forcing a step-change in infrastructure spending. The top hyperscalers - Amazon, Alphabet, Meta, and Microsoft - plan to spend over $600B in 2026, roughly 70% higher than last year's pace. In 2025, these four invested just over $350B.

Markets aren't sure how fast that spend turns into profit. Amazon is down 12% year to date and Microsoft is off 16%. Alphabet is down less than 1% in 2026, while Meta is up about 1%. Still, early beneficiaries are already showing up in infrastructure.

"Early with the heavy spending"

"Basically, the takeaway is that the most competent companies in the world are telling us that we're still early," said Gene Munster of Deepwater Asset Management. Translation: the buildout phase isn't over - it's accelerating.

Paul Meeks at Freedom Capital Markets expects spending to plateau or grow more slowly after this surge, not collapse. "I've talked to the management teams of all the hyperscalers, and they see this as a real competitive advantage for them to be early with the heavy spending." He also cautioned that expecting clear monetization this soon is unrealistic.

Investor Ken Mahoney took a more cautious view: "There's no guardrails. Feels like companies are just spending and spending and spending, and hope on the other side they come in first place." The tension is clear: scale up now or risk being late - but execute with discipline.

Where gains may show up first: "pick-and-shovel" infrastructure

As capex ramps, the near-term winners often sell the essentials that keep AI running. Mahoney's playbook: "Find those companies that are going to be the backbones … have your shopping list … buy them below the market and be more tactical."

• CoreWeave: Cloud infrastructure for AI workloads. Watching for execution and data centers coming online. Shares are up 33% in 2026.
• Arista Networks, ASML, Snowflake: Munster called them out as beneficiaries. Year to date: Arista +7%, ASML +34%, Snowflake -18%.
• Oracle: On watch for returns on its investments; shares are down 19% this year, last trading under $160.
• Vertiv: Provides electrical infrastructure and cooling for data centers. Shares jumped 24% on Wednesday and have doubled over the past year.
• GE Verona: Another energy-focused name Mahoney is watching. Up 26% this year, around $790.
• Monolithic Power Systems, Bloom Energy: Meeks flagged both as derivative energy plays. Up 32% and 79% this year, respectively.

Semiconductors are still center stage

"The Street has 60%-70% revenue growth for Nvidia this year," Meeks said, versus around 10% for the S&P 500. Nvidia shares are roughly -2% over three months and +2% YTD.

• Broadcom, Taiwan Semiconductor: Consensus picks to keep benefiting. Broadcom is -1% YTD; TSM is +23%.
• Micron: A memory name Meeks is watching. Up 44% this year.

What to watch next (signals that matter)

• Capex follow-through: Do 2026 budgets hold through year-end? Any mid-year trims or re-allocations?
• Time-to-rack: Lead times for GPUs, networking, lithography equipment, cooling, and grid interconnects. Slippage here pushes revenue out.
• Utilization and pricing: GPU hours sold, inference pricing trends, and attach rates for AI features inside core products.
• Energy constraints: Data center siting, grid access, and thermal limits are now gating factors. Expect more activity around siting and efficiency standards. For context, see PUE and efficiency metrics.
• Supply concentration: Watch single-source risks in advanced lithography. Background on EUV tech: extreme ultraviolet lithography.
• ROI timeline: Expect a lag between spend, deployment, and monetization. Early revenue often shows up in infrastructure vendors first.

Practical moves by role

• IT & Development
  • Plan around capacity scarcity: multi-cloud options, spot GPU markets, and workload scheduling to cut inference costs.
  • Prioritize network-aware architecture and observability for AI pipelines; design for portability to avoid lock-in.
  • Invest in skills that match the stack being deployed (networking, data engineering, model optimization).
• Finance
  • Track capex-to-revenue, depreciation timing, and lease obligations for data centers; these drive margin optics.
  • Watch backlog quality, delivery schedules, and cash conversion for "picks-and-shovels."
  • Stress-test AI pricing assumptions and utilization; small changes swing payback periods.
• General Management
  • Align AI roadmaps with realistic infrastructure availability; sequence pilots where capacity is secured.
  • Choose vendors with a delivery record over promises. Set clear KPIs for adoption and cost per outcome.
  • Avoid FOMO. Discipline now beats rework later.

Bottom line

AI capex is scaling hard and fast. The near-term opportunity skews to infrastructure - compute, networking, equipment, and the energy-and-cooling stack - while platform monetization takes longer to show through.

If you build, operate, or invest, keep your lens on capacity, delivery timing, and unit economics. That's where the signal lives.

Want a structured way to upskill your team for this buildout? Explore role-specific learning paths here: AI courses by job.