What to Own if the Tech Boom Falters | Steven Feldman

The host asks Steven Feldman whether signs of trouble are emerging in the AI-driven equity rally. Feldman points to two developments: AI adoption may be rising while the price companies can charge for it is falling, and the industry’s capital structure is becoming strained. He contrasts Microsoft, which can probably service its borrowing, with CoreWeave, a less established company whose bonds fell to roughly 70 cents on the dollar. Bondholders have limited upside and are signaling that repayment is less certain, making CoreWeave a “canary” for malinvestment in a capital-intensive sector.

Feldman discusses an $18 billion data center project involving Blue Owl and Oracle. Oracle needed a permit for a pipeline to supply energy, but the permit was delayed; Blue Owl had supplied $3 billion and lenders another $15 billion. Oracle then issued a force majeure notice saying it had a reason not to pay rent without technically being in default. Feldman says force majeure has occurred in real estate before, but rarely on an $18 billion project. He is not declaring the AI trade over, but says previously discussed leverage and execution risks are now becoming news.

Asked why these incidents might extend beyond isolated projects, Feldman cites Aon’s CEO warning that potential losses could be larger than insurers are underwriting, while noting that even AIG’s history shows insurers can misjudge crises. He describes a fragile, highly leveraged, interconnected system in which Nvidia acts simultaneously as bank, joint-venture partner, investor, and customer to companies building AI infrastructure. Nvidia may now matter to the economy almost as much as JPMorgan. Feldman recommends a more research-driven approach: take profits from winners such as Nvidia, Hewlett Packard, and Dell rather than assuming they can rise another fourfold.

The host asks whether commodities depend on continued AI data-center construction. Feldman distinguishes mathematical demand from investor emotion: even an underperforming data center still needs electricity, so the infrastructure trade can survive weaker returns. Utility stocks inflated by AI enthusiasm could revert to normal utility valuations, but electricity transmission was already undersupplied before AI. Copper was previously supported by electric vehicles and broader electrification, then received an additional boost from data centers; a calmer AI buildout would reduce attention without eliminating the underlying shortage. Feldman says investors should separate scarce inputs from expensive momentum trades.

Feldman argues that inflation favors scarce assets, but not scarce assets dependent on rolling large amounts of debt. A New York building may become more valuable as replacement costs rise, yet rising interest rates can reduce its net operating income and levered equity value. He explains why gold outperformed mining companies for years: gold required no refinancing, while miners needed capital and faced higher debt costs. The same principle applies to real assets generally—owning the commodity can be preferable to owning a highly indebted operator exposed to it.

The host connects the leverage warning to farmland, noting that a farmer can own valuable land while still suffering if input costs and financing costs rise. Feldman says crop prices have recently helped, but an owner whose borrowing cost rises from 3% to 7% can see levered equity deteriorate. His farmland asset manager, Goldcrest, deliberately operates without leverage and treats farmland as a “get rich slow scheme.” There is limited new land, and he frames it as civilization’s infrastructure—without land there is no food or humanity. It has had four down years in 50, each around 3%, and is relatively uncorrelated.

Feldman views Berkshire Hathaway’s roughly $370 billion in cash and the succession of Greg Abel as meaningful signals. Warren Buffett built around franchises such as Apple, Coca-Cola, American Express, and insurance; Abel ran Berkshire’s real-assets operations and understands railroads, pipelines, and power plants. Feldman expects Abel to “grind it out” while waiting for an obvious opportunity. Unlike Apollo or Blackstone, which may be exposed to the AI trade and less willing to buy during a selloff, Berkshire could deploy perhaps $50 billion. He imagines it rescuing a company such as Oracle after a broad decline.

Feldman says current inflation reflects energy shocks from the Ukraine and Iran wars, diesel-heavy supply chains, and a wealth effect that has let companies keep raising prices. He cites diesel at $8.85 per gallon locally, steakhouse steaks exceeding $100, luxury goods such as a $9,000 Chanel skirt, and hotel rooms costing $2,000 per night. He expects China’s manufacturing glut to export deflation and believes energy pressure could eventually ease, but says $2 trillion annual deficits, roughly $40 trillion of debt, and higher interest costs create “fiscal dominance.” No political faction appears willing to raise taxes or cut spending.

Feldman says the two most inevitable portfolio themes are monetary protection and electricity. Gold functions like insurance against geopolitical and monetary shocks, while electrification supports copper, uranium, pipelines, data centers, batteries, and domestic mineral supply chains. He rejects automatically buying every dip: a weak company can become even cheaper if its cash flow never returns, whereas a market crisis may offer bargains in great franchises. He favors diversification, including S&P 500 or Dow index funds that continually remove failing companies and add stronger ones; historically, fewer than 100 companies generated about 90% of stock-market returns. He also sees a renewed role for 5%–10% cash and bonds.