First a message from our friends at Stansberry Research (sponsor)
Buy this AI Stock?
Dear Reader,
Whitney Tilson – the former hedge fund manager CNBC once dubbed "The Prophet" – just unveiled a new breakthrough.
It's a proprietary stock-grading engine that can analyze thousands of securities in real time... and uncover what even Wall Street's best analysts miss.
It's so complex, Tilson says, not even an army of MIT quants could replicate it.
Now, he's giving away one of his system's highest-rated stock ideas... and the name might surprise you.
It's not Nvidia.
It's not Amazon.
It's not Palantir, Oracle, or any other AI darling in the headlines.
This company is quietly partnering with major universities to roll out a new "intelligence education" platform, and Tilson believes it could be a much smarter way to play the AI boom.
To see the ticker symbol, and get a free demo of the system behind it all...
Regards,
Matt Weinschenk
Director of Research, Stansberry Research
Featured Article
AI Infrastructure vs. AI Chips — Where Is the Real Money Now?
For most of the last two years, “AI investing” has been treated like a one-word answer: chips.
If you owned the best accelerators, the best CUDA moat, the best networking—congratulations, you owned the narrative. And the market paid you for it.
But the tape has a new tell in 2026: the market is starting to reward a different kind of edge—constraint power.
Not processing power. Constraint power.
The real money in the next leg of AI isn’t necessarily in the companies that make AI smarter; it’s increasingly in the companies that make AI possible to deploy at scale. That means the “unsexy” layer: power delivery, cooling, grid gear, racks, and data-center physical build-out.
Today’s market proved it—loudly—with Vertiv.
Vertiv (VRT), a critical data-center infrastructure supplier, reported Q4 results with net sales of $2.88B (+23% YoY) and organic orders up 252% YoY—and the stock ripped about 25% in a single session.
That kind of move isn’t a normal earnings beat. It’s the market admitting:
The bottleneck is no longer “can we get enough GPUs?”
The bottleneck is “can we power and cool them fast enough?”
So here’s the unique position I’m taking—clearly:
The next 6–12 months of “AI alpha” will likely come from the bottlenecks: power, cooling, and grid capacity—more than from mega-cap AI chips.
Chips still matter, of course. But the market has already priced chips like a coronation. Infrastructure is being repriced like a discovery.
And that difference—priced perfection vs. priced skepticism—is where active traders get paid.
Let’s break it down with numbers, then turn it into a practical trading playbook.
1) Chips aren’t slowing—chips are already priced like they can’t fail
Start with the undisputed king: Nvidia.
Nvidia’s Q3 fiscal 2026 numbers were objectively monstrous: $57.0B revenue (+62% YoY) and Data Center revenue $51.2B (+66% YoY).
That’s not “growth.” That’s industrial-scale demand.
But here’s the trading issue: everyone knows that. When a story becomes consensus truth, the question for price becomes:
How much better can reality get than expectations?
What happens when the next quarter is “only” great, not biblical?
What happens if hyperscalers shift spend from chips to the facilities needed to run them?
This doesn’t mean Nvidia is “done.” It means the asymmetry changes:
chips can keep rising, but they’re vulnerable to valuation air pockets,
while infrastructure can rise because the market is still catching up to the constraint.
2) Infrastructure just delivered the “proof print” traders wait for
Vertiv is the poster child, but it’s not alone.
Vertiv’s demand shockwave
Vertiv guided 2026 net sales of $13.25B–$13.75B, adjusted EPS $5.97–$6.07, and adjusted free cash flow $2.1B–$2.3B.
The market didn’t just like the results—it liked what the results imply: a sustained wave of data-center buildouts, not a one-quarter spike.
Eaton: “data center momentum” is now explicit
Eaton reported record results and highlighted that in Electrical Americas, order acceleration was up 16% driven by “data center momentum,” and Electrical backlog grew strongly (the release cites 29% YoY backlog growth in Electrical).
Investing.com summarized the data-center slice more bluntly: data center revenue grew ~40% vs Q4 2024, with orders up ~200% in the quarter and backlog up ~31% YoY.
That’s not “AI hype.” That’s industrial demand.
Hubbell: grid + data centers = a real growth cocktail
Hubbell’s earnings transcripts/coverage show data center sales in Electrical Solutions exceeded 60% growth for the quarter and data centers are now over 10% of that segment’s sales; 2026 guidance calls for 7–9% total sales growth and adjusted EPS $19.15–$19.85.
Hubbell is not a chip company. It’s the kind of company institutions used to own for “steady utility grid work.” That’s changing.
3) The most underappreciated AI reality: electricity is the ultimate choke point
AI is not just a compute arms race. It’s an energy procurement arms race.
Reuters reported Duke Energy raised its five-year capex plan to $103B and has signed agreements to supply 1.5 gigawatts of power to new data-center customers since the last quarter—bringing contracted data-center capacity to 4.5 GW, with 9 GW more in its pipeline.
Read that again.
Not megawatts. Gigawatts. And that’s one utility.
This is why I think the “real money” shifts toward infrastructure:
AI chips are the brains.
Infrastructure is the body.
The grid is the oxygen.
And oxygen is becoming scarce.
Let’s put the AI stack into layers and highlight tradable names in each.
Layer A: Power + Cooling (the new “picks and shovels”)
Vertiv (VRT): explosive order growth and raised 2026 guidance.
Eaton (ETN): backlog, order acceleration, and data-center revenue +40% cited in coverage.
Hubbell (HUBB): grid demand + data-center exposure, with 2026 EPS guidance $19.15–$19.85.
These are not “cheap” after the move, but they are trend-capable because they’re tied to multi-quarter build cycles.
Layer B: “AI factory” integrators (high growth, but watch margins)
Super Micro (SMCI): reported $12.7B net sales in Q2 FY2026 vs $5.7B a year ago, but gross margin fell to 6.3%.
SMCI is the classic active-trader stock: momentum can be incredible, but margins and concentration risk can turn it into a trap if the market shifts to profitability discipline.
Layer C: The sleeper that matters more than people admit—memory & networking
As inference scales, the “AI factory” becomes bandwidth-hungry.
Micron: investor materials show data center was 56% of FY2025 revenue and cited gross margins of 52% (in that FY2025 context), highlighting mix shift and profitability benefits.
Broadcom: Reuters quoted AI revenue growing 63% to $5.2B in fiscal Q3 2025 and projected $6.2B for Q4.
This is where “chips vs infrastructure” becomes a false binary. The next alpha pocket may be in enablers:
memory (HBM),
switching (Ethernet scaling),
and custom silicon ecosystems.
5) The coming weeks: the 3 most likely trading regimes
Regime 1: “Infrastructure leadership” continues
This happens if:
rates stabilize,
hyperscalers reaffirm capex,
and the market keeps hunting non-mega-cap AI exposure.
Watch for follow-through in VRT/ETN/HUBB—if they hold gaps and build tight bases, they can trend.
Regime 2: “Chips reclaim leadership” on a risk-on burst
If the market shifts risk-on hard, mega caps can rip simply because they’re liquid and heavily owned.
But that doesn’t kill infrastructure—it just changes beta.
Regime 3: “AI digestion”—the most dangerous regime
If yields spike or macro data shocks, high-multiple AI names can de-rate quickly. In this regime:
infrastructure sometimes holds up better than chips,
but everything gets more volatile.
6) Active Trader Playbook: how to actually express this view
Here are actionable frameworks (suggestive, not definitive).
Strategy A: The “Constraint Basket” (trend-following, not chasing)
Build a watchlist basket:
VRT, ETN, HUBB
and trade them only after they set a post-event base.
Rules that matter:
prefer tight consolidation after big earnings gaps,
buy breakouts with defined risk,
cut fast if the breakout fails.
Strategy B: The “Barbell” (chips + bottlenecks)
Instead of choosing sides:
hold a smaller “chip anchor” (e.g., NVDA exposure via spreads),
and pair it with infrastructure leaders.
This works when you believe AI spending persists but leadership rotates.
Strategy C: Pair trade (advanced)
If you think the next move is rotation out of perfection:
long infrastructure (ETN/HUBB/VRT),
hedge with a small short/put spread on a chip-heavy index (QQQ or SOXX).
You’re not betting AI dies—you’re betting the market reprices where the bottleneck profit sits.
Strategy D: Options: use spreads to avoid IV traps
After earnings gaps, IV can stay elevated. Consider:
call spreads on breakouts,
put spreads as gap insurance if you chase (better: don’t chase).
Strategy E: The “Utility Tell”
Keep one utility name on your dashboard as a signal. Duke’s capex/data-center power agreements are a reminder that the grid isn’t optional.
If utilities tied to data-center demand strengthen during tech chop, it’s a clue the market believes the buildout is real and long.
7) The risks that can break the infrastructure thesis
This is not a free lunch.
Risk 1: Capex digestion
Hyperscalers can pause, optimize, and renegotiate timelines. Infrastructure names can get hit if “order pull-forward” becomes the narrative.
Risk 2: Rates/valuation compression
These are not deep-value names. If yields jump, they can sell off with the market.
Risk 3: Execution risk
Delivering high-density cooling/power gear at scale is operationally complex. Any supply chain issues can compress margins.
Risk 4: Crowding (the trade gets discovered)
The moment everyone agrees “infrastructure is the new chips,” the easy part is over. That’s why we trade setups, not slogans.
The bottom line
The AI trade is evolving:
Chips are still the engine—and Nvidia’s numbers prove demand is not fading.
But the market is waking up to a bigger truth: the constraint is physical.
Power, cooling, grid gear, and deployment speed are becoming the gatekeepers of AI capacity.
Vertiv’s order explosion and guidance reset were the clearest “tell” that capital is moving down the stack.
Eaton and Hubbell show it’s not a one-company story—it’s an ecosystem repricing.
And utilities like Duke are effectively confirming that the grid is being rebuilt for this wave.
My position: Over the coming weeks, the best active-trader risk/reward is likely in the AI bottlenecks—especially infrastructure and electrification—because chips are already priced like destiny, while bottlenecks are still being priced like a surprise.
Trade it like an adult: wait for bases, define risk, and let the market prove your thesis.
Editorial Disclaimer
This commentary is for informational and educational purposes only and does not constitute investment advice. All market strategies involve risk, and past performance is not indicative of future results. Readers should conduct their own analysis or consult a licensed financial professional before making investment decisions.