The race to power artificial intelligence is shifting from hype to hardware, according to Jim Thorne of Wellington-Altus Private Wealth. In a recent TV appearance, the chief market strategist said the AI boom is set to be a long, capital-heavy expansion rather than a quick sprint.
Thorne framed the next phase as an infrastructure story that will reshape spending by tech giants, utilities, and manufacturers. He argued that the push to train and run AI models will demand new data centers, stronger power grids, and a deeper supply of chips and networking gear. The message was clear: this wave will take years and a lot of money.
From Software Buzz to Concrete and Steel
Investors have spent the past two years focused on AI software and chip leaders. Thorne urged a wider view. He said AI will need massive physical buildouts to reach its promise. In his words, the trend resembles a long cycle rather than a short fad.
Thorne called it “a decade-long infrastructure cycle,” describing it as an A.I. supercycle that will require steady investment.
The pitch echoes past shifts. The dot-com era pushed fiber networks. The smartphone wave expanded wireless towers. Cloud computing led to huge server farms. AI may combine parts of all three, with higher energy needs and stricter reliability demands.
What a Long Buildout Could Involve
The core of the thesis is simple. AI models need fast chips, fast networks, and cheap, steady power. Those needs cascade through the economy.
- New data centers with advanced cooling and power systems.
- More high-bandwidth networking within and between facilities.
- Upgrades to local grids, transmission lines, and substations.
- Supply of chips, memory, and power equipment.
- Skilled labor for construction and maintenance.
This buildout could support semiconductor makers, equipment suppliers, industrials, and utilities. It could also affect energy markets as operators seek reliable sources and backup plans.
Winners, Bottlenecks, and Trade-Offs
Thorne’s framing points to a broad group of potential winners. Chip designers and manufacturers may benefit from steady orders. Electrical equipment makers could see long backlogs. Network suppliers may gain from higher traffic and tighter latency needs.
But bottlenecks loom. Transmission projects can take years to approve and build. Transformers and switchgear already face long lead times in many regions. Water use and siting rules could slow construction. Costs can rise when supply chains stretch.
There are also trade-offs. Data centers pull heavy loads from local grids. That can crowd out other users during peak times. Communities may push for stricter standards on energy mix, water use, and noise. These debates could shape where and how fast facilities are built.
Industry Impact and the Next Phase of Spending
Thorne’s remarks fit with signs from major tech firms, which have guided for higher capital spending tied to AI. Utilities and independent power producers have begun to cite data center demand in their plans. Construction firms report interest in high-spec facilities.
If the buildout sticks, spending may shift in phases. Early dollars focus on chips and flagship sites. Later, funds move to grid upgrades, regional redundancy, and efficiency gains. Over time, software and services that run on this base may see steadier growth.
Risks and What Could Go Wrong
The long-cycle view carries risk. AI progress could slow if model gains taper or costs stay high. A downturn could delay projects. Policy shifts on energy or data privacy could alter plans. Competition might compress margins in some segments even if revenue grows.
Investors also face timing risk. Infrastructure returns often lag the headline story. Stocks can run ahead of cash flows, then correct as projects meet real-world hurdles.
What to Watch
Key signals will help test the thesis over the next year:
- Capital spending updates from top cloud and chip firms.
- Grid interconnection queues and transmission approvals.
- Lead times for transformers, switchgear, and high-voltage gear.
- Power purchase agreements tied to data center clusters.
- Local resistance or incentives that affect siting.
Thorne’s take puts hard assets at the center of the AI story. If the “decade-long infrastructure cycle” view holds, the next leg of the boom may be written in permits, steel, and power lines, not just code. For investors and operators, watching supply chains, construction timelines, and energy costs may matter as much as model updates. The coming months should reveal whether spending plans match the ambition—and how fast the grid and supply base can keep up.
