War, Watts, and the AI Buildout That Won’t Stop

The lights went out across lower Manhattan on the evening of Nov. 9, 1965. Within minutes, 30 million people across the northeastern United States and parts of Canada were sitting in the dark. The trigger? A single mis-set relay on a transmission line near Niagara Falls.

That one relay led to a cascading failure.

No, the grid wasn’t broken. But it was never designed to handle the load being placed on it. And the engineers who warned about the mismatch between electricity demand and infrastructure capacity had been saying so for years.

Nobody listened until the lights went out.

Sixty-one years later, we’re staring at the same mismatch… only this time, the load isn’t air conditioners and elevators. It’s artificial intelligence.

The demand for compute is growing so fast that some of the most serious minds in tech are now debating whether the answer is nuclear reactors or data centers floating in orbit. That’s not science fiction. It’s an active engineering and investment debate. And the outcome will determine which companies sit at the center of a multi-trillion-dollar buildout — and which ones get left behind.

Meanwhile, geopolitics is making the energy equation even more complicated. The Iran blockade has injected real uncertainty into global energy markets. Oil is elevated. Inflation expectations are shifting. And the sectors that benefit from sustained energy disruption — defense, energy infrastructure, and AI compute — are not the same ones most investors are watching.

All of this came together on this week’s episode of Being Exponential.

The Iran War: Put It Behind You

We open the episode with the latest on the Iran conflict, and my read is unambiguous: the war is winding down on every front simultaneously.

Israel and Lebanon agreed to a separate 10-day ceasefire this week. Trump announced he will host Lebanon’s President Aoun and Israel’s Prime Minister Netanyahu at the White House — the first direct talks between the two countries since 1983. Pakistan’s Army Chief and an entire Pakistani delegation arrived in Tehran carrying a U.S. framework for second-round peace talks. Iran’s welcome of the delegation (rather than rejection) tells you Iran is still in the deal-making business. Doors that are closed don’t get opened for Pakistani field marshals carrying American peace plans.

The war risk premium is essentially fully priced out of markets. Oil is in the $90s. Stocks are at all-time highs. The ceasefire holds. The negotiations progress. My bottom line: the bullish backdrop for the next several months is intact as oil gradually retreats toward $80, inflation pressures modestly ease, and the Fed begins to contemplate the beginning of the end of its paralysis.

But the geopolitical story doesn’t end with a ceasefire. In this week’s podcast, I walk through how the Iran conflict reshuffled sector rotation in real time — defense stocks, energy infrastructure, and AI compute were the beneficiaries during the fear cycle, and the transition from wartime risk premium to peacetime normalization creates a different set of winners. The episode breaks down what that rotation looks like and how to position around it.

Orbital Compute vs. Nuclear: The Energy Debate That Will Define AI’s Next Decade

This is the section of the podcast that will probably generate the most conversation, and for good reason. If AI models keep scaling at their current trajectory, where does the electricity come from? The hyperscalers are already spending tens of billions of dollars on data center capacity, and the bottleneck is increasingly not chips or networking…

It’s power.

I break down the two competing visions. On one side, next-generation nuclear — companies like Oklo (OKLO), Constellation Energy (CEG), and the Global X Uranium ETF (URA) — is being fast-tracked by hyperscalers who need reliable, high-density baseload energy that can sit adjacent to massive compute clusters. On the other side, orbital compute is moving from theoretical concept to early-stage execution.

And this week gave us a major real-world data point. Amazon’s (AMZN) acquisition of Globalstar (GSAT) is, on the surface, a connectivity and spectrum play to compete with Starlink. But I see something bigger…

The non-obvious, longer-running, and arguably far more consequential play is orbital compute. You can’t run data centers in space at scale without a dense, trusted low-Earth-orbit constellation, licensed spectrum to route traffic, and an anchor customer creating consistent demand.

Globalstar gives Amazon all three — critically, ahead of the SpaceX initial public offering (IPO) and Elon Musk getting tens of billions of dollars to start launching his own orbital data centers.

The investment map I lay out in the episode spans the full stack: compute owners like Amazon, Microsoft (MSFT), Alphabet (GOOGL), and Meta (META); rocket companies including Rocket Lab (RKLB); orbital network operators like AST SpaceMobile (ASTS), Globalstar, Iridium (IRDM), and ViaSat (VSAT); space compute suppliers such as GlobalFoundries (GFS), Microchip Technology (MCHP), Texas Instruments (TXN), and Nvidia (NVDA); space power companies including Redwire (RDW), Boeing (BA), and Lockheed Martin (LMT); and laser/optical interconnect players like Coherent (COHR), Lumentum (LITE), and Ciena (CIEN).

My conclusion: The commercial space era has begun. Both the nuclear path and the orbital path lead to massive infrastructure buildouts. And the companies supplying the picks and shovels for both are positioned to benefit regardless of which vision wins… or, more likely, as both scale simultaneously.

Now is a great time to be investing in the suppliers for space tech.

Prediction Markets: The New Forward-Looking Signal

From energy, we pivot to fintech — specifically, the rise of prediction markets and what they mean for how investors process information.

Robinhood (HOOD) is at the center of this story. The stock surged this week on a bullish note from Bernstein citing asymmetric upside potential from a recovery in crypto markets combined with breakout growth in prediction markets. My take: prediction markets are the new Vegas. And Robinhood is building the front door to this new Vegas.

The broader point I make on the podcast is worth sitting with. Prediction markets are becoming a real-time, crowd-sourced probability engine for geopolitical events, elections, and macro outcomes. They’re faster than polls, more honest than punditry, and increasingly more liquid than anyone expected.

As platforms like Robinhood scale this product, prediction markets could become the new forward-looking signal that financial markets themselves begin to price around.

I see tremendous upside potential for Robinhood through prediction market growth over the next several years. Technically, I’d like to see the stock retake its February high of $86. If it does, I think HOOD could run toward $200 by the end of the year.

The Biggest Headwind Facing the AI Trade

This is where the episode gets contrarian… and where I’m most direct.

A wave of negative AI headlines has hit the tape. Bubble fears. ROI skepticism. Overspending concerns from hyperscaler critics. And I don’t dismiss them. The social sentiment against AI is awful and worsening rapidly. My concern is that populist backlash against AI is next, followed by anti-AI legislation.

In my view, this is the single biggest risk to the AI trade… not valuations, not competition, not a slowdown in model improvement…

But the sentiment cycle is separate from the fundamental cycle. The negative headlines are surfacing now precisely because the AI buildout has reached a scale that is visible to the general public. That visibility creates the conditions for backlash… the same way every major technology wave, from railroads to the internet, faced a populist headwind once the investment reached a certain size.

The question is whether the backlash translates into legislation that meaningfully constrains AI investment, or whether it remains noise within a much larger supercycle.

My read: the fundamentals are too strong, the capital commitments too deep, and the competitive dynamics too urgent for any political headwind to slow this buildout materially. But it’s a risk worth watching, and I walk through why the social sentiment trajectory matters even if the earnings trajectory doesn’t change.

AI Infrastructure: Still Early Innings

I close the podcast with the big picture — and it’s anchored by the most important data point of the week.

TSMC’s Q1 report landed like a thunderclap. Revenue growth exceeded estimates. Full-year 2026 guidance was raised to 30%-plus growth. Capital spending was guided to the upper end of its $56 billion range. Management’s characterization of AI-related demand: “extremely robust.”

TSMC physically manufactures the chips that power the entire AI ecosystem — Nvidia’s Blackwell and Vera Rubin, Apple’s silicon, Broadcom’s custom XPUs, AMD’s accelerators. When TSMC raises guidance and maxes out capex, it’s telling you something about every one of those customers simultaneously.

Companies don’t spend $56 billion on manufacturing equipment for demand they expect to slow.

The read-through for earnings season: if TSMC’s chip shipments are growing 30%-plus, then every company that designs chips for TSMC to manufacture — and every company that uses those chips for AI workloads — is in a fundamentally strong position.

Nvidia’s May earnings will confirm demand. Broadcom’s next quarter will confirm custom silicon acceleration. Micron’s HBM numbers will confirm memory strength. The ecosystem is on fire. TSMC told you so with actual revenue, not management promises.

Credo Technology (CRDO), whose acquisition of DustPhotonics expands its portfolio of optical interconnect solutions, could be the future of data center connectivity. Credo is, after all, one of my favorite AI infrastructure stocks. I wouldn’t be surprised to see the stock at $250 by the end of the year.

But I want to add an important short-term caveat. The Nasdaq is working on a 12-day win streak — the first since 2009.

I ran the historical precedents: January 1992 produced a 9% 12-month forward return but a 10% correction within five months. February 1986 delivered 18% over 12 months but was flat seven months later. November 1980 and August 1979 both saw sizable pullbacks within months before resuming higher.

History tells us stocks should head higher over the next 12 months. But along the way, we will likely get a sizable correction within months. TSMC reported a perfect quarter and the stock dropped 3%, because the rally had already priced the good news in.

The right thing to do here is to trim very little, stay long and strong, and recognize that the next four to six weeks belong to patience and selectivity… even as the 12-month thesis has never been stronger.

The Bottom Line

This episode of Being Exponential is one of those conversations where every segment feeds into the next. The Iran wind-down reshapes sector rotation. The energy question reshapes AI infrastructure economics. Prediction markets reshape how investors process geopolitical risk. Populist backlash reshapes the narrative. And the TSMC quarter confirms that underneath all of it, the fundamental demand for AI compute hasn’t even begun to slow.

If you want to understand how war, energy, orbital compute, fintech innovation, and macro sentiment are colliding to shape the next phase of markets, this episode connects the dots.

Watch the full episode of Being Exponential here.