⚡ THIS WEEK'S SIGNAL
Three signals in eight days, one story. April 24: X-energy's IPO closes 15x oversubscribed at roughly a $12 billion implied market cap, on a company whose first commercial reactor isn't expected to generate power before the early 2030s. May 6: Riot Platforms — a Bitcoin miner under activist pressure — signs an MOU with Terrestrial Energy for up to 4 GW of pre-commercial Gen-IV reactors, with gas as bridge fuel. This morning: Constellation beats Q1 expectations, sits on 147 terawatt-hours of uncontracted existing nuclear, and declines to name a hyperscaler buyer. Three prices for nuclear scarcity, all printed in public the same week.
━━━━━━━━━━━━━━━━━━━━━━━
📰 MAIN STORY: The Existing Fleet Is Almost Spoken For. The Have-Nots Are Buying 2032.
The buyers who locked in existing nuclear capacity by 2025 are paying a fundamentally different price than the buyers who didn't. That's the story underneath the news this week. AWS sits on $18 billion for 1.9 GW of Susquehanna at roughly $68.90/MWh — power that already runs. Meta has up to 6.6 GW across Vistra, Oklo, and TerraPower. Microsoft has its 835 MW Three Mile Island restart at Crane Clean Energy Center, though that one is now flagged for potential PJM interconnection delay to 2031. Everyone else who needs gigawatt-scale firm clean baseload before 2030 is buying things that don't exist yet.
Each player is optimizing for something the others can't quite get. Constellation, sitting on 147 TWh of uncontracted output, openly told the market this morning it's waiting on PJM's co-location rulemaking before pricing the next tranche — because every month of regulatory uncertainty is a month it sells scarcity at a higher implied premium. Riot Platforms, whose hashrate business is in post-halving compression, is repositioning its existing Texas and Kentucky power infrastructure as the substrate for a nuclear-data center campus and was rewarded with a four-year stock high on the announcement. Solaris Energy Infrastructure just signed a third hyperscaler contract on April 24 — 600 MW, ten years, deployments through 2028 — expanding its behind-the-meter gas fleet 40% to 3,100 MW, which tells you where the buyers who can't wait for nuclear are putting their money in the meantime. Terrestrial Energy gets the validation event it needed to convert IMSR from concept to credible. Everyone in this picture, in their own way, gets to call this a win.
Here's the split worth sitting with. The hyperscalers who got there first are paying ~$68.90/MWh on fixed long-term PPAs against carbon-free baseload running at 90%+ capacity factors today. The hyperscalers and developers who didn't are signing MOUs for reactors with no NRC license, paired with gas as a bridge — meaning they're paying twice for the same delivery window, once in equity and again in interim emissions. Meanwhile, PJM's capacity charge has gone from $28.92/MW-day in 2024/25 to $329.17/MW-day in 2026/27, lifting region-wide capacity costs from $14.7 billion to $16.1 billion, with the residential share absorbed through bills. One set of buyers got the contracts. Another set is buying the press releases. Ratepayers get the math either way.
The numbers worth screenshotting: Constellation's 147 TWh of uncontracted nuclear; the AWS-Talen ~$68.90/MWh benchmark; X-energy at ~$12 billion market cap with no commercial revenue; Riot/Terrestrial's 4 GW MOU across 390 MW IMSR plants at Texas and Kentucky sites; Amazon's 5 GW X-energy commitment through 2039 (from a supplier Amazon's Climate Pledge Fund helped seed in Series C-1); Solaris's BTM gas fleet at 3,100 MW serving three hyperscalers, financed by $1.3 billion in fresh 6.375% senior unsecured notes; the IEA's tally of 17% data center electricity demand growth in 2025 and an SMR offtake pipeline now at 45 GW; and S&P Global's updated forecast of $1.295 trillion in U.S. utility capex from 2026 through 2030. Most of the contracted, fixed-price firm clean baseload in PJM through 2030 was signed before the AI demand inflection. Most of what's been signed since is a longer-duration bet that something becomes deliverable later.
Scenario A: Oklo achieves criticality at Idaho National Laboratory this summer (the pathway Oklo has publicly described), FERC finalizes the large-load rule cleanly enough to survive litigation despite former Chairman Mark Christie's blistering critique, and Constellation announces a named hyperscaler PPA off its uncontracted book in Q2 or Q3 — priced at a meaningful premium to the AWS-Talen $68.90/MWh benchmark. Call it $80–$95/MWh. The MOUs (X-energy, Terrestrial, Oklo, TerraPower) start to look like real capacity rather than narrative premium, and the scarcity premium gets printed as a number rather than implied.
Scenario B: Oklo's INL timeline slips, FERC's large-load rule gets enjoined by a state regulator coalition, Crane formally moves to 2031, and the next eighteen months of hyperscaler capacity decisions get routed through BTM gas — which is precisely where Solaris, VoltaGrid, and BloombergNEF's 4.9 GW of co-located storage-plus-gas already say the market is going. The 4 GW Riot/Terrestrial MOU becomes an emblem rather than a project. The existing fleet remains the only nuclear that delivers.
One thing to watch: whether Oklo achieves criticality at INL before Constellation's Q2 earnings call in early August. Both events compress timing assumptions across every nuclear procurement decision currently being modeled.
━━━━━━━━━━━━━━━━━━━━━━━
PRESENTED BY [SPONSOR]
[Sponsor copy — one-line value proposition, two sentences of context, CTA.]
━━━━━━━━━━━━━━━━━━━━━━━
⚡ QUICK HITS
Riot Platforms + Terrestrial Energy MOU at up to 4 GW: The Bitcoin miner and the Gen-IV IMSR developer signed a Memorandum of Understanding on May 6 to co-develop nuclear-powered data centers at Riot's Texas and Kentucky sites, with natural gas as bridge fuel. RIOT hit a four-year high and IMSR jumped 7–15% on the announcement — the bet is that an existing hashrate operator's land, transmission, and power-engineering footprint is the right substrate for a nuclear-anchored campus. (BusinessWire)
Solaris Energy Signs Third Hyperscaler at 600 MW, 10-Year Term: Solaris's April 24 contract scales BTM gas deployments through 2028 and expands its fleet 40% to 3,100 MW serving three hyperscalers, financed by a $1.3B note issuance. This is the canonical IPP template for BTM-gas-for-hyperscalers — contracted, long-duration cash flows that don't depend on nuclear delivering anytime soon, and the model every infrastructure fund underwriting AI power is now benchmarking against. (BusinessWire)
Clearway Lands New 15-Year Hyperscaler PPA at Mesquite Sky: Clearway's Q1 report disclosed a long-term PPA with an investment-grade hyperscaler replacing a prior commodity contract at the West Texas wind project, at more favorable pricing terms. Worth flagging because renewables-plus-storage is still winning hyperscale PPAs in ERCOT — a useful counter-signal to the assumption that everyone is racing only into nuclear. (GlobeNewswire)
Constellation Q1 Beats, Hyperscaler Customer Still Unnamed: Constellation reported $11.12B in Q1 revenue against $8.7B consensus, EPS of $2.74 against $2.61, and reaffirmed full-year guidance of $11–$12. The call did not name a new PPA off the 147 TWh uncontracted book — the deal-pipeline silence is the catalyst, and the next two earnings windows are the data. (Constellation IR)
S&P Global: $1.295 Trillion U.S. Utility Capex Forecast Through 2030: S&P's April update marks the largest sustained utility capex projection in the sector's modern history, driven explicitly by data center load growth. Eversource's $1.5B in junior subordinated notes oversubscribed 5x last month is the canary — capital markets are funding the buildout faster than regulators are finalizing the rules for who pays for it. (S&P Global)
━━━━━━━━━━━━━━━━━━━━━━━
🔧 TOOL / RESOURCE OF THE WEEK
Third Way's Advanced Nuclear Map: A continuously updated tracker of every announced advanced reactor project in North America, classified by stage — concept, design, NRC licensing, construction, operation. For separating the deals where capacity is actually getting built from the deals where capacity is currently a press release (a distinction the Riot/Terrestrial MOU, the X-energy IPO, and the Oklo Pike County campus all sit on different sides of), it's the cleanest free reference. → thirdway.org/graphic/advanced-nuclear-map
━━━━━━━━━━━━━━━━━━━━━━━
💬 CLOSING THOUGHT
Cautious optimism here looks like this: the procurement mechanisms — fixed PPAs against existing fleet, equity in advanced reactor developers, MOUs for Gen-IV capacity, BTM gas as bridge — have all become legible enough that capital is flowing into each layer. What hasn't been priced is which of those layers delivers on the schedule the AI buildout assumes. Riot's 4 GW MOU is real if it becomes a construction permit. It's an emblem if it doesn't.
If you've evaluated a Gen-IV nuclear MOU in your territory in the last six months — your own or a competitor's — what's the internal threshold that converts it from optionality into a line item in your capex plan? An NRC filing? A DOE loan guarantee? A specific equity round? I'd actually like to know what trigger you use. Reply if you can — I'll synthesize the answers anonymously next week.
— The Watt Report
