The AI trade is broadening beyond chips, and power names now offer the cleaner setup
The bullish AI case does not have to fade for leadership to rotate. This week’s chip volatility, grid-strain headlines, and transformer shortages all point to the next leg of AI monetization moving toward electrical equipment, cooling, and grid capacity.
The AI trade is not ending; it is getting more physical. That matters because the market spent the last year treating AI exposure as mostly a semiconductor question, while the real bottleneck is increasingly whether enough power, cooling, switchgear, and transmission capacity can be built to support the next wave of data-center demand. This week’s setup made that shift hard to ignore: renewed volatility in chips landed at the same time as fresh reporting on grid congestion and equipment shortages. Our view is straightforward: if AI capex is still coming, the cleaner risk-reward now sits with the infrastructure names that make that spend usable in the real world.
The key change is that AI demand is no longer just a compute story; it is a capacity story. Public reporting this week showed generator step-up transformer lead times stretching past 160 weeks, up from 143 weeks in 2024. That is not a sentiment problem or a valuation debate. It is a hard supply constraint, and hard constraints are where pricing power and backlog visibility tend to show up first. If hyperscalers and developers still want to build, they cannot skip the electrical layer.
The demand side is also becoming too large to dismiss as a niche issue. Consensus-linked research now points to U.S. data-center power demand rising from 31 GW in 2025 to 66 GW in 2027, with data centers taking a much bigger share of peak summer load over that span. Pair that with reports of northern Virginia power prices spiking above $2,000/MWh amid congestion, and the market’s old framing starts to look incomplete. The next debate is not whether AI needs chips; of course it does. The more important question is which part of the stack is now scarce enough to command the cleaner earnings setup.
That is why we would rather reframe AI exposure around the picks-and-shovels than chase another straight-line semiconductor move. NVDA still looks exceptional on operating quality, with 63.0% net margins and 65.5% revenue growth, and AVGO is still printing huge AI-linked growth. But those are now the most crowded expressions of the theme, and they sit at the center of every capex-payback argument each time the tape gets shaky. By contrast, the power-and-cooling basket is tied to a simpler proposition: if the buildout continues, the physical bottlenecks have to be funded.
A quick look across the group shows why the setup is broadening rather than merely rotating out of chips:
Those numbers do not say semis are expensive and industrials are cheap in some simplistic way. They say the market is still paying a massive premium for pure compute leaders, while several infrastructure beneficiaries offer more direct exposure to the new bottleneck at far lower sales multiples. ETN at 5.55x sales and PWR at 3.28x are not bargain-bin stocks, but they are being valued against visible real-world constraints rather than against the next twist in AI chip sentiment. Even VRT, after a huge run, is easier to justify as a cooling-and-power beneficiary than as just another momentum trade, given that thermal management is becoming as essential as compute density.
Yes, bulls on the chip side will fairly argue that the biggest economic rent still sits with the silicon layer. They are not wrong to point to AVGO posting 286.6% EPS growth or to NVDA still combining elite margins with extraordinary growth. But that comparison misses where the market debate is actually moving this week. The issue is not whether semis remain great businesses; it is whether the next incremental dollar of AI enthusiasm is better expressed through names exposed to transformer shortages, grid upgrades, and cooling demand than through the most consensus-owned mega-cap winners.
That is also why the industrial-and-power basket is not just a defensive sidestep. PWR has already shown the market what backlog-backed demand can look like, and ETN, VRT, and GEV all sit in parts of the chain where lead times and congestion can sustain spending even if investors get more selective about which AI applications monetize fastest. GEV is especially notable here: it carries a TickerSpark Score-friendly profile for this debate because the market is increasingly treating generation and grid equipment as part of the AI stack, not as a separate utility story. Once data centers become a meaningful driver of regional power stress, the old sector boxes stop being very useful.
The historical analogy is less dot-com software than dot-com infrastructure. Early winners captured the narrative; later winners solved the bottlenecks. That is what this moment looks like. AI still needs accelerators, but it also needs substations, transmission work, cooling systems, transformers, and electrical distribution gear. When the constraint shifts from code to electrons, the cleaner setup shifts with it.
The bottom line is that the AI bull case can stay intact while leadership broadens away from semis. We think that is exactly what this week’s mix of chip volatility, grid-strain reporting, and equipment shortages is signaling. The better expression of the theme now is not abandoning NVDA or AVGO; it is recognizing that ETN, PWR, VRT, and GEV are increasingly where AI demand becomes unavoidable rather than optional.
What would change our mind? A clear easing in grid congestion, shorter lead times for critical electrical equipment, or a sharp slowdown in data-center power commitments would weaken the bottleneck thesis. Until then, we would rather own the parts of the AI buildout that the industry physically cannot route around.
Our take, not advice. This is opinion commentary — informational only, not personalized investment recommendations. Markets carry risk. Do your own research and consider your own situation before any trade.
▌The Daily Briefing · Free
A new stock idea, every evening.
One stock worth watching each weekday, plus the analysis behind it. Free, in your inbox.
