Most people met the AI-infrastructure trade in 2024, when the phrase started showing up on every earnings call. Neel Khokhani was already in it — he opened his $IREN position in 2022 and most recently disclosed 703,000 shares.

He runs Epochal Corporation, a single-family office deploying his own capital. No outside investors, no redemption calendar, no benchmark to hug. That structure is doing more work in this story than most people realise, and it's the part that's hardest to copy.

The thesis, in one sentence

"Power, land, and grid interconnection, rather than capital, are the binding constraints on growth."

That's it. Capital for data centres has been abundant for years. Energised megawatts, land with a real interconnection queue position, and the operational competence to bring both online on schedule have not been. If the bottleneck is physical rather than financial, the winners are whoever locked up the physical assets early — and the market usually doesn't re-rate that position until the constraint becomes obvious to everyone.

What he tracks on the name: data-centre footprint, pace of customer contracting, and the competitive structure of high-density compute. What he doesn't track: quarterly beats and price action.

Where the numbers come from

This is the part worth stealing.

He doesn't start with analyst models. He starts with utility interconnection queues, transmission planning documents, municipal permit filings, corporate registry records, satellite imagery of construction sites, equipment nameplates visible in site photography, and job postings.

Concrete version: reading transformer specs at a site to infer how much compute is actually going in. Cross-referencing grid operator planning data against announced capacity. Finding subsidiaries in corporate registries before they show up in a shareholder deck.

None of that requires a Bloomberg terminal. It requires being willing to read filings nobody else reads.

He fights the board, too

Being a long-term holder isn't passive here. Khokhani has publicly opposed an 18.2 million RSU grant to $IREN's co-CEOs — no performance conditions attached, meaningful dilution to everyone else — writing to the chairman directly and taking the case to proxy advisers ISS and Glass Lewis.

His framing: if you intend to hold across a full cycle, how the company treats shareholder capital is your problem, not someone else's.

The uncomfortable opinion

Khokhani's position on diversification will annoy a lot of people: adjusted for inflation, monetary debasement and tax, he argues index-level diversification frequently produces a real loss of purchasing power.

The load-bearing words are the ones that come next — "correctly-priced" and "asset-backed." It's not an argument for concentration as such. It's an argument for owning fewer things that are backed by something physical and were bought below a value you calculated yourself.

He applies a 17–20% real hurdle rate, which by design kills most ideas before they get sized.

What he built first

He came to markets from the operating side, not from research.

Aviation. Founded Soar Aviation, grew it from one aircraft to fifty-five, funded entirely from customer prepayments and operating cash — no priced round, no syndicated debt. He sold down and exited all operational and board roles; what happened to the business afterwards happened under different management.

Consumer finance. Took roughly a third of a lender, cleaned up the structure, revenue went from ~$45M to ~$82M, exited at ~$121M enterprise value.

Storage. Vachi Storage, a UAE self-storage operator, is the one he still owns. It exists to throw off predictable, capital-light cash that has almost nothing to do with the compute cycle.

That last one matters more than it sounds. Concentration in a volatile sector is only survivable if something else in the structure is boring and paying.

What actually transfers to your account

Transferable: buy the business, not the ticker. Insist on a discount to a number you worked out yourself. Keep something uncorrelated and cash-generative so you're never forced to sell the volatile position at the bottom. And do primary research — the edge is available to anyone willing to do the unglamorous reading.

Not transferable: the size. A concentrated book run on permanent capital by someone who's already exited two companies is a completely different risk profile from a concentrated book run on money you might need in five years. The method scales down. The position sizing does not.

The structural advantage he has — nobody can force him to sell — you technically have too. Nobody can redeem your brokerage account either. Most people give that away voluntarily.

Neel Khokhani publishes his positions and research openly on Substack and on X (@neel_epochal). Position disclosed: long $IREN.

This article is biographical and informational. It is not investment advice or a recommendation to buy or sell any security. Positions are subject to change without notice.