TL;DR
The uprate trade has gone fleet-wide: Google signed a 20-year PPA with Constellation funding 890 MW of new capacity across 11 reactors, and DOE formally committed up to $4.2 billion to Vistra for 433 MW of uprates — two days apart. But the week's counter-signal is physical: a tilted fuel assembly stalled America's first full reactor restart at Palisades, the first proof that "fast megawatts" carry execution risk. Meanwhile the equity split hardened — generators ripped 15%+, developers made new lows, and uranium's record term price still can't lift Cameco.
Hyperscalers Are Now Financing the Fleet, Not Just Buying It
The corporate PPA template has evolved from buying existing output to directly funding incremental capacity. On October 6, Google and Constellation announced a 20-year agreement supporting 890 MW of uprates across 11 units in Illinois, Pennsylvania and New Jersey, backed by more than $4.3 billion of Constellation investment, with the first uprate "expected to be delivered by 2028" (Constellation press release). Reuters puts Google's total Constellation contracting at 3,590 MW, with new nuclear about a quarter of that (Reuters). Google's own framing makes the strategy explicit:
"has now enabled over 1.5 GW of new nuclear capacity for the U.S. energy system from increasing the power output of existing reactors and bringing dormant facilities back online."
— Google blog
This is the second hyperscaler–Constellation deal in a week — Amazon's Calvert Cliffs PPA closed September 30 — and Utility Dive tallies the pair at "almost 1.1 GW of nuclear expansions" (Utility Dive). Uprates win because they deliver firm megawatts in years, not the decade-plus of greenfield build — which is exactly the demand SMR developers were counting on.
What to watch: Whether the NRC's expected-uprate list starts naming specific Constellation units, since none of this power arrives without NRC approval.
DOE Money Lands on the Same Megawatts
The open thread from last cycle resolved on schedule: DOE formally announced October 5 a conditional commitment of up to $4.2 billion for uprates at Vistra's Beaver Valley, Davis-Besse and Perry, adding 433 MW and preserving 4 GW of baseload (DOE announcement). Energy Secretary Chris Wright made the logic plain:
"By getting more power out of the nuclear plants we already have, we can deliver more affordable, reliable, around the-clock energy for American families and businesses."
— ANS Nuclear Newswire
The fine print matters: the loan is conditional, and NRC uprate applications aren't due until 2029–2032 (ANS). One oddity worth tracking — Meta said in January it would support uprates at these same three plants, "the largest uprates supported by a corporate customer," yet the DOE announcement doesn't say how the two arrangements interact. VST closed the week up 15.3% at $161.48 (Vistra market view
).
What to watch: Whether the Meta and DOE arrangements get reconciled publicly, or whether Vistra is quietly double-counting the same uprates.
The First Restart Speed Bump
The restart thesis — firm megawatts in years, not decades — just met its first physical test. During fuel loading at Palisades in late August, assembly DD41 "was placed in core location H-1 but tilted over after being released by the refueling machine," left at a 30-degree angle against the core shroud (ANS). Holtec filed an emergency license amendment September 30, warning that "a delay in approval of the proposed changes would result in a delay in the resumption of activities necessary to reach the plant conditions required to restore the plant to power operation" (ANS).
The NRC has preliminarily found no significant hazards and the comment period closed October 8, so this may resolve in days-to-weeks. But no restart timeline has been republished since the incident, and the read-across touches every restart in the cohort — Crane/TMI (Microsoft), NextEra's Iowa project (Google). A single fuel assembly pausing the path to criticality is a reminder that "already built" doesn't mean "already running." Silver lining: Holtec simultaneously submitted the safety analysis for two SMR-300 units at the same site.
What to watch: NRC approval of the LAR and any republished restart date — a slip beyond early 2027 would be a negative signal for the whole restart trade.
Uranium's Record Term Price Still Can't Buy a Bid
The contract market is pinned at history while the miner trades like the theme is dying. TradeTech reports the long-term price "Remains at Historical High of $97.00 for Four Consecutive Months," while spot sits flat near $89.50 with "thin spot liquidity" (TradeTech; carboncredits.com). Cameco closed at $88.06, 34.3% below its high, RSI 40 (Cameco market view
).
The structural positives keep stacking anyway: Westinghouse began long-lead procurement for Poland's first AP1000 (first unit slated for 2036 commissioning), Cameco formally acknowledged the $120 billion US–Korea framework as separate from DOE's $17.5 billion supply-chain facility, and Canada–India nuclear talks surfaced as a further demand channel (Westinghouse; Cameco statement). The disconnect likely comes down to a simple lag: Cameco guides to 2026 realized prices of C$91–96/lb against C$93.13 in Q2 — the record term price hasn't hit the P&L yet (TIKR).
What to watch: Cameco's Q3 print on October 30 — realized prices and contracting commentary are the next hard test of whether the divergence closes or deepens.
The Equity Split Becomes a Chasm
The generators and the developers now live in different markets. CEG closed the week up 15.8% at $298.07 and VST up 15.3% — while OKLO hit its 52-week low at $34.70 (down 80.1% from the high) and NuScale, at $7.20, is down 86.5% and now faces a Pomerantz investigation on top of UBS's Sell rating (CEG/VST market views
; OKLO/SMR market views). The 247wallst line captures the mechanism:
"Google just signed a nuclear deal with a company that already runs reactors, yet the stocks surging hardest belong to developers whose reactors exist only on paper."
— 247wallst
The developers got their one-day sympathy bid on October 6 — Oklo and NNE up 8% each — and gave it back within days (TipRanks). The market is paying only for contracted megawatts and federal balance-sheet support; development-stage optionality prices at zero regardless of the policy backdrop. The thesis-breaker would be a binding, priced hyperscaler offtake for a first-of-a-kind unit — until then, every new deal lands on the incumbent fleet first.
What to watch: Whether any developer converts a framework (Tillman, ENTRA1/TVA) into hard, priced commitments before the capital rotation makes fundraising harder.
What surprised us
- Google's 890 MW deal isn't just a PPA — it's a two-way commercial embed. Constellation will deploy Google Cloud and Gemini Enterprise "agentic workflows" in operations and outage management. The hyperscaler-nuclear relationship is becoming a technology partnership, which raises the switching cost for the next utility deal (Constellation press release).
- A single fuel assembly can freeze a flagship restart. Palisades was the sector's proof-of-concept for fast megawatts, and one tilted assembly during loading required emergency regulatory processing. Execution risk in restarts is real and unpriced (Finding: Palisades
).
- The Meta–DOE overlap at Vistra's plants went unmentioned in the federal announcement. Either it's coordinated or it's a conflict nobody wants to narrate — worth pressing on the Q3 call (Finding: Vistra loan).
- Even the IPO channel is strained. Standard Nuclear priced at $15, raised $150 million, and promptly traded below offer — yet Cathie Wood's ARK is buying the dip in X-Energy and Oklo. Someone is catching this knife (Finding: equity selloff).
Open threads worth a vote
Nuclear Energy's Comeback — Delta Briefing
Since last time
Escalated
- The uprate trade went fleet-wide. Last cycle it was two hyperscaler deals in two weeks; now Google has signed a 20-year PPA funding 890 MW across 11 reactors (Oct 6), and the DOE loan moved from report to formal conditional commitment. The framing shifted from "buying existing output" to "financing incremental capacity."
- The equity split became a chasm. Generators ripped (CEG +15.8%, VST +15.3% on the week) while OKLO hit a fresh 52-week low and NuScale picked up a Pomerantz investigation. Same thesis, much wider spread.
- The uranium disconnect gained a causal explanation. Term price ticked $96 → $97/lb; the new briefing argues the miner's lag is a realized-price timing issue, not a demand doubt.
Demoted
- The $120B US–Korea framework was a full section; it's now a single line inside the uranium section (Cameco's formal acknowledgment). Still pending, still non-binding.
- Amazon's Calvert Cliffs PPA went from headline to one line of context for the Google deal.
Disappeared
- The NRC's Clinch River BWRX-300 construction permit — a full section last time, entirely absent now.
- The Saudi floor vote (scheduled before Dec 13) — no mention.
- NNE's $13.5M Radnostix fuel-cycle acquisition — no mention (NNE appears only in a one-day 8% rally note).
Unchanged — nothing qualifies; even the uranium section moved on numbers and mechanism.
Entirely new: the Palisades restart stumble — the first physical counter-signal to the "fast megawatts" thesis.
Hyperscalers Are Now Financing the Fleet, Not Just Buying It — Escalated
The template evolved again: from buying existing output (Amazon/Calvert Cliffs, Sept 30) to directly funding incremental capacity. On October 6, Google and Constellation announced a 20-year agreement supporting 890 MW of uprates across 11 units in Illinois, Pennsylvania and New Jersey, backed by more than $4.3 billion of Constellation investment, with the first uprate "expected to be delivered by 2028" (Constellation press release). Reuters puts Google's total Constellation contracting at 3,590 MW, with new nuclear about a quarter of that (Reuters). Google's own framing:
"has now enabled over 1.5 GW of new nuclear capacity for the U.S. energy system from increasing the power output of existing reactors and bringing dormant facilities back online."
— Google blog
Utility Dive tallies the two hyperscaler deals at "almost 1.1 GW of nuclear expansions" (Utility Dive). Uprates win because they deliver firm megawatts in years, not the decade-plus of greenfield build — which is exactly the demand SMR developers were counting on.
What to watch: Whether the NRC's expected-uprate list starts naming specific Constellation units, since none of this power arrives without NRC approval.
DOE Money Lands on the Same Megawatts — Escalated (thread closed)
Last cycle's open thread resolved on schedule: DOE formally announced October 5 a conditional commitment of up to $4.2 billion for uprates at Vistra's Beaver Valley, Davis-Besse and Perry, adding 433 MW and preserving 4 GW of baseload (DOE announcement). Energy Secretary Chris Wright:
"By getting more power out of the nuclear plants we already have, we can deliver more affordable, reliable, around the-clock energy for American families and businesses."
— ANS Nuclear Newswire
New fine print since the Reuters report: the loan is conditional, and NRC uprate applications aren't due until 2029–2032 (ANS). One oddity worth tracking — Meta said in January it would support uprates at these same three plants, "the largest uprates supported by a corporate customer," yet the DOE announcement doesn't say how the two arrangements interact. VST closed the week up 15.3% at $161.48 (Vistra market view).
What to watch: Whether the Meta and DOE arrangements get reconciled publicly, or whether Vistra is quietly double-counting the same uprates.
The First Restart Speed Bump — New
The restart thesis just met its first physical test. During fuel loading at Palisades in late August, assembly DD41 "was placed in core location H-1 but tilted over after being released by the refueling machine," left at a 30-degree angle against the core shroud (ANS). Holtec filed an emergency license amendment September 30, warning that "a delay in approval of the proposed changes would result in a delay in the resumption of activities necessary to reach the plant conditions required to restore the plant to power operation" (ANS).
The NRC has preliminarily found no significant hazards and the comment period closed October 8, so this may resolve in days-to-weeks. But no restart timeline has been republished since the incident, and the read-across touches every restart in the cohort — Crane/TMI (Microsoft), NextEra's Iowa project (Google). "Already built" doesn't mean "already running." Silver lining: Holtec simultaneously submitted the safety analysis for two SMR-300 units at the same site.
What to watch: NRC approval of the LAR and any republished restart date — a slip beyond early 2027 would be a negative signal for the whole restart trade.
Uranium's Record Term Price Still Can't Buy a Bid — Escalated
Same divergence as last cycle, sharper numbers and a new explanation. TradeTech reports the long-term price "Remains at Historical High of $97.00 for Four Consecutive Months," while spot sits flat near $89.50 with "thin spot liquidity" (TradeTech; carboncredits.com). Cameco closed at $88.06, 34.3% below its high, RSI 40 (Cameco market view).
The structural positives keep stacking: Westinghouse began long-lead procurement for Poland's first AP1000 (first unit slated for 2036 commissioning), Cameco formally acknowledged the $120 billion US–Korea framework as separate from DOE's $17.5 billion supply-chain facility, and Canada–India nuclear talks surfaced as a further demand channel (Westinghouse; Cameco statement). The new mechanism for the disconnect: a realized-price lag — Cameco guides to 2026 realized prices of C$91–96/lb against C$93.13 in Q2, so the record term price hasn't hit the P&L yet (TIKR).
What to watch: Cameco's Q3 print on October 30 — realized prices and contracting commentary are the next hard test.
The Equity Split Becomes a Chasm — Escalated
Last cycle's "split" is now a chasm. CEG closed the week up 15.8% at $298.07 and VST up 15.3% — while OKLO hit its 52-week low at $34.70 (down 80.1% from the high) and NuScale, at $7.20, is down 86.5% and now faces a Pomerantz investigation on top of UBS's Sell rating (CEG/VST market views; OKLO/SMR market views):
"Google just signed a nuclear deal with a company that already runs reactors, yet the stocks surging hardest belong to developers whose reactors exist only on paper."
— 247wallst
The developers got their one-day sympathy bid on October 6 — Oklo and NNE up 8% each — and gave it back within days (TipRanks). The market pays only for contracted megawatts and federal balance-sheet support; development-stage optionality prices at zero. The thesis-breaker would be a binding, priced hyperscaler offtake for a first-of-a-kind unit — until then, every new deal lands on the incumbent fleet first.
What to watch: Whether any developer converts a framework (Tillman, ENTRA1/TVA) into hard, priced commitments before the capital rotation makes fundraising harder.
What surprised us
- [NEW] Google's 890 MW deal isn't just a PPA — it's a two-way commercial embed. Constellation will deploy Google Cloud and Gemini Enterprise "agentic workflows" in operations and outage management. The hyperscaler-nuclear relationship is becoming a technology partnership, raising switching costs for the next utility deal (Constellation press release).
- [NEW] A single fuel assembly can freeze a flagship restart. Palisades was the sector's proof-of-concept for fast megawatts, and one tilted assembly required emergency regulatory processing. Execution risk in restarts is real and unpriced (Finding: Palisades).
- [NEW] The Meta–DOE overlap at Vistra's plants went unmentioned in the federal announcement. Either it's coordinated or it's a conflict nobody wants to narrate — worth pressing on the Q3 call (Finding: Vistra loan).
- [NEW] Even the IPO channel is strained. Standard Nuclear priced at $15, raised $150 million, and promptly traded below offer — yet Cathie Wood's ARK is buying the dip in X-Energy and Oklo. Someone is catching this knife (Finding: equity selloff).
Open threads
Closed:
DOE formally announces $4.2B Vistra uprate loan — resolved October 5; covered in the body above.
Dropped from the list (still live, still unresolved):
- US–Korea framework definitive agreements / named AP1000 sites — no longer a tracked thread; the framework got only a passing mention this cycle.
- Q3 earnings calls (SO, CEG, VST, CCJ) — narrowed to just the Cameco October 30 print as a watch item.
New: