Riot Platforms signed a 20-year lease with Anthropic for 191 megawatts of IT capacity at its Rockdale, Texas campus. The contract runs through June 2048 and is worth roughly $9.1 billion. Two five-year extension options would take it to about $16.1 billion.
The stock jumped roughly 25% when the agreement was first reported, then spent the next session giving most of it back, closing at $20.24, up 4.33%.
The contract
| Term | Value |
|---|---|
| IT capacity leased | 191 MW |
| Initial term | 20 years, through June 2048 |
| Contract value | ~$9.1B |
| With both extensions | up to $16.1B |
| 96 MW online by | December 2027 |
| Full buildout by | June 2028 |
| Q2 2026 revenue | $174.2M (beat) |
Riot's chief executive Jason Les put total signed capacity at 241 megawatts and roughly $9.8 billion of long-term contracted revenue, accumulated in six months. The other counterparty is AMD.
What actually changed
Nothing about the buildings changed. What changed is the shape of the revenue.
A bitcoin miner earns a floating, unhedgeable payout: hash price times uptime, set by a market it does not influence, halved on a schedule it does not control. A data center landlord earns a fixed rent from a named tenant on a contract with a maturity date. Same concrete, same substations, same transmission interconnect — completely different asset class on the income statement.
That is the whole re-rating. Bernstein moved its target to $35, Citi to $32, both calling the deal transformational, and both are really saying the same thing: the discount rate applied to Riot's cash flow should fall, because the cash flow stopped being a bet on bitcoin.
The part the headline skips
Read the delivery schedule again. Ninety-six megawatts by December 2027, the full 191 by June 2028. This contract produces almost nothing for two years, and in the meantime Riot has to build it.
Turning a mining hall into hyperscaler-grade capacity is not a retrofit. It is redundant power paths, cooling density that mining never needed, network, security, commissioning — and it is paid for before the first rent cheque arrives. The financing of that gap is the risk in this trade, and it is a risk the $9.1 billion headline actively hides. A signed lease makes borrowing cheaper; it does not make it unnecessary.
The second thing to watch is what a 20-year contract means in a technology cycle that reprices every 18 months. Power is the durable part. The specific compute in the building will be replaced three or four times before 2048, and the tenant, not Riot, carries most of that. That asymmetry is exactly why power owners are winning this cycle.
The fade is the tell
A 25% pop that closes at plus four is not a rejection of the deal. It is the market doing the arithmetic above in real time.
The first buyers priced $9.1 billion. The closing price prices $9.1 billion arriving from 2028, minus the cost of getting there, times the probability of getting there on schedule. That is a much smaller number, and the gap between the two is the entire distance between a headline and a valuation.
How I read it
The interesting number is not $9.1 billion. It is 241 megawatts signed in six months against a market that cannot get power connected in under four years.
I treat Riot now as an energy landlord with a bitcoin option attached, not the reverse. The mining business becomes the thing that pays the bills while the leases are built, and the leases become the thing that sets the valuation.
What would break the thesis: a capital raise on bad terms in 2027, or a delivery slip that pushes rent commencement past the contracted date. Both are execution risks, both are visible quarters before they hit the stock, and both are more likely than a closing price up 4% suggests the market is pricing.
What would confirm it: a third lease. Two counterparties in six months is a pattern. Three would make it a business.