Every bitcoin mine carries a death clock. Each halving cuts the coin reward in two: the workbook runs it plainly — a 150 MW site earning $59M after tax today earns $32M after the 2028 halving at the same bitcoin price, and needs either $150k bitcoin or rigs 50% more efficient just to stand still. Capitalize that decaying stream at a 12.8% WACC and the site is worth $3.34–$3.91 per watt. That is the "alive" price.
The "dead" price — kill the mining, lease the energized land and interconnection to someone building an AI data center — starts from one fact this letter keeps returning to: the shortage is measured in years, not dollars (No. 003). A converted crypto site saves a developer ~3 years of queues and permits, worth $13.33/W on top of the asset itself, because every idle year burns GPU depreciation at four times the cost of the power. Hence the market-clearing price for hyperscaler-grade converted capacity: $15 per watt. Same transformers, same substation, same dirt — 4x the value, purely for being early.
This is no longer a thesis; it is a tape. Since December 2024, ~28 leases have converted miner megawatts into contracted data-center revenue — TeraWulf, Cipher, Applied Digital, IREN, Hut 8, Riot, Galaxy, Core Scientific — with terms that rhyme: ~15-year leases plus extensions, rents of $1.55–$2.29 per critical-IT watt per year, 85–90% operating margins, built for $9–13/W. Run each through the same machine (lease → EBITDA → 5.5x leverage → exit to a REIT buyer at 15x) and the equity created at signing clusters with eerie consistency:
Now mark the sector to that price. The ten listed miners control roughly 33 GW of 100MW+ sites — 5.1 GW operational, 6.0 under construction, 27.6 in development — against a combined enterprise value near $104B. Value the operational and under-construction megawatts at $15/W and the development pipeline at $15/W with only a coin-flip's chance of energization, and the same portfolio is indicatively worth ~$221B. The dispersion is the trade: names that already signed (IREN at $33/W on its operational fleet) have eaten much of their re-rating, while MARA — the biggest fleet, zero HPC deals signed — trades at $2.76/W, priced as a perpetual bitcoin mine.
One worked example, because the machine deserves daylight. Cipher's Barber Lake: 168 MW of critical IT, built for $1.68B ($10/W), leased at $1.79/W of revenue with an 82.5% margin — $247.5M of EBITDA, a 14.7% unlevered yield on cost. Lever it 5.5x, sell to a REIT at 15x EV/EBITDA, and $420M of equity becomes $2.35B — ~5.6x — before handing Google $288M of warrants for guaranteeing the lease. Google appears as backer or guarantor on at least five of these deals, collecting penny warrants for lending its credit rating. The hyperscaler that refused to build fast enough is being paid in miner equity for promising to pay the miners' rent — the politest vendor financing of the cycle.
| The input | The output | |
|---|---|---|
| The site | 168 MW critical IT, West Texas grid + queue position | 15-yr lease, Google-guaranteed tenant |
| The build | $1.68B construction ($10/W) | $300M/yr revenue ($1.79/W) |
| The cash flow | 82.5% operating margin | $247.5M EBITDA — 14.7% yield on cost |
| The financing | 5.5x leverage ($1.26B debt, HY at ~7%) | $420M of equity at risk |
| The exit | REIT buyer at 15x EV/EBITDA | $3.71B EV → $2.35B equity: ~5.6x (5.8x on net sale proceeds) |
| The toll | $288M of warrants to Google | Net creation: $1.42B = $8.43/W |
The $15/W clearing price is scarcity pricing, not replacement cost — new construction is $9–13/W. If the shortage closes (gas turbines arrive, chip volumes slip, efficiency jumps), the premium evaporates and the development-pipeline math — 27.6 GW valued at a 50% coin-flip — collapses toward land value. Most pipeline megawatts are announcements, not transformers. The deal economics also lean on a 15x REIT exit bid that is itself a creature of this cycle's enthusiasm, and on tenants — many of them neoclouds borrowing at 7–9.25% — honoring 15-year leases through a token-price collapse. The 1873 lesson (No. 002) applies to the landlords too: contracted revenue is only as good as the creditor behind it.
And MARA's discount may be earned, not overlooked: hosted sites, West Texas air cooling, and a balance sheet built around holding bitcoin are real conversion handicaps. Cheapness that persists for six quarters is sometimes a verdict.
Eight benchmark deals have held ~$12–13/W of equity creation for 18 months. The first deal that prints materially below tells you scarcity is closing.
The cheapest watts in the sector belong to the one major miner with no HPC deal. Either it signs (violent re-rate) or it keeps not signing (the discount is information).
Post-halving, mining margins halve — conversion supply floods in exactly as gas turbines arrive. Watch whether bitcoin under ~$70k accelerates conversions enough to crush the $15/W price.
For a decade the miners were the economy's designated sinners — men who burned rivers of power to mint coins nobody could hold, denounced at every hearing, banked by no one. What they actually built, without meaning to, was the scarcest industrial asset of the 2020s: approved, energized, high-density power. The market is now paying billions for the regulatory sin of having built infrastructure nobody sanctioned — and the miners, who wanted to be outlaws, discover that what they truly wanted all along was to be wanted: to have Microsoft sign a fifteen-year lease, to have Google guarantee their debts, to be, at last, respectable.
That is the desire under the data, and it prices the trade's end as surely as its beginning. Assets bid up by the hunger for legitimacy converge on the returns of legitimate assets. The first conversions captured the outlaw's premium; the last will earn a utility's yield, wrapped in a REIT, sold to the distant creditor from No. 002 — who will be told, correctly and uselessly, that the cash flows were contracted.