Thesis Maintenance
Weekly review (2026-10-23, Opus 5.5 (Max)) of each thesis against recent moves + filings. These are PROPOSED updates to keep the priors honest -- review and apply the ones you agree with; nothing is auto-edited.
CEG P 0.75
Health
Stable.** The contract pillar (Pillar 2) now has real proof: three hyperscalers signed 20-year deals in one week. But Google's expanded virtual power plant is the kind of off-list signal the thesis told us to watch for. A virtual power plant pools flexible customer demand and gets paid like a power plant. That makes it a competing source of capacity in PJM, the grid region where most of CEG's plants sit. The good news and the new threat roughly cancel out.
Pillar P changes
Pillar 2 (hyperscaler PPAs): 0.70->0.82.** The pillar said "PPAs will come." Now three counterparties have signed 20-year deals, including Google's 3,590 MW in PJM. What's still open is price per MWh, not whether deals happen.
Pillar 3 (Crane restart): 0.60->0.65.** The NRC clearance noted on 10-02 is a real step forward. FERC interconnection and the actual startup date are still open, so the bump is small. (The log entry is cut off. Confirm it refers to Crane before you lock this in.)
Pillar 1 (scarcest firm power): 0.85, no change.** Virtual power plants compete for *capacity* payments. They can't supply 24/7 carbon-free power, which is what hyperscalers are signing for.
Pillar 4 (Calpine synergies): 0.55, no change.** No new evidence.
Triggers
ADD:** "Demand response and virtual power plants together clear more than ~10% of PJM capacity in the Dec 2026 auction, or push the clearing price toward $200/MW-day." That turns the Google signal into a number you can check, and it feeds the existing $200 trigger directly.
ADD (process check):** "Announced PPAs don't show up in an 8-K or the Q3 disclosure." Today there are no recent 8-Ks. Deals this big should leave a paper trail. If they don't, they may be smaller or softer than the headlines.
RETIRE:** None.
Valuation check
Stale.** The ~$368 consensus and ~27x P/E came before three 20-year contracts. Re-anchor on the disclosed PPA price per MWh across the ~147M MWh still uncontracted. That number is the whole valuation case.
Watch next
Whether CEG's Q3 filing or earnings call states PPA price terms per MWh and confirms the Crane timeline. If the price isn't disclosed, the PPA pillar gain stays unverified.
Recommended P
The tracker swung from 0.82 to 0.95 to 0.75 in three weeks. That's noise, not signal. The 10-23 cut already accounts for the virtual power plant risk. The PPA gain is real, but it isn't confirmed by a filing yet. 0.75 is the right place to hold until the Q3 disclosure.
RECOMMENDED_P: KEEP
VST P 0.76
Health
Stable. Nothing new has come in since the 10-16 review (no 8-Ks, no PPA news). The DOE loan is already priced into the 0.76, and nothing has happened since to move it either way.
Pillar P changes
Pillar 4 (Uprates add low-cost capacity): 0.55->0.65.** The DOE's planned $4.2B loan covers 433 MW of uprates. That moves uprates from "planned" to "funded at government rates." It stops short of 0.70+ because this is still a conditional commitment, not a closed loan, and no uprate megawatts have come online yet.
Pillar 3 (Cheapest multiple = margin of safety): 0.72->0.68.** The DOE money is cheap, but it is still debt on top of $15.8B. That pushes D/E toward the 3.5x trigger and strengthens the bear case. The stock has also probably re-rated on the news, so the margin of safety is thinner.
With these changes the pillars average about 0.70. The extra above that comes from the DOE event being real.
Triggers
Retire** "Revenue misses in 4 of 5 quarters." Vistra's reported revenue swings with hedge accounting (marking hedges to market value), so a revenue miss is mostly noise for this business. With 3 of 4 already counted, one noisy print could zero the thesis.
Add** "Adjusted EBITDA guide cut below the $6.8B low end." That measures what the thesis depends on.
Add** "DOE loan fails to close or is cut materially within 12 months." This isn't thesis-breaking on its own, but it should reverse the Pillar 4 bump.
Amend** the D/E trigger to "D/E above 3.5x *excluding* DOE-guaranteed nuclear debt, or any rating downgrade." Otherwise cheap federal debt could fire the trigger for the wrong reason.
Valuation check
The anchor is likely stale. The ~19x forward P/E and the $217–256 targets predate the DOE news, so re-pull the price and consensus targets before treating Vistra as "cheapest."
Watch next
The Q3 earnings date and setup, likely in early November. Watch whether EBITDA guidance holds at $6.8–7.6B. Also watch for any timeline update on Meta and Amazon PPA deliveries or on closing the DOE loan.
Recommended P
RECOMMENDED_P: KEEP
TLN P 0.35
Health
Weakening.** The FERC Section 206 case against PJM (a formal FERC review of whether PJM's rules are fair, with possible refunds) is a real risk to how Susquehanna sells power, and nothing has offset it. But no new facts have come in since the Oct 9 review, and there are no new 8-Ks.
Pillar P changes
Pillar 2 (clean front-of-meter transition post-FERC): 0.60 -> 0.45.** The 206 case puts PJM's grid rules back under review, including the ones this transition depends on. The refund effective date means any change could also reach back in time.
Pillar 1 (Susquehanna + AWS PPA = durable cash): 0.70 -> 0.62.** The contract still stands. But if PJM's rules change, the economics behind about 35% of gross margin could get worse. That is a small cut, not a broken pillar.
Pillars 3 and 4: no change. There is no new evidence on Cornerstone or the AWS ramp.
Triggers
Add:** a FERC order in the PJM 206 case that changes front-of-meter or co-location terms against Susquehanna, or orders refunds that hit Talen. Set P near 0 if that happens. Right now the trigger list doesn't cover the risk that is actually live.
Add a process rule:** a docket counts as one event, so make one move per new filing or order. The tracker cut P five times on the same 206 news between Oct 2 and Oct 7, from 0.42 down to 0.22. That double-counting is why the Oct 9 review had to reset it.
No triggers to retire.
Valuation check
Stale.** The ~$498 Morgan Stanley re-rate target was set before the 206 case. Apply a regulatory-risk discount to it, or replace it with a current target.
Watch next
Watch what PJM and Talen file in the FERC 206 docket this week. The question is whether co-location and front-of-meter deals like the AWS setup are inside or outside the scope of the case. That decides whether Pillar 1 holds or Pillar 2 breaks.
Recommended P
The Oct 9 review already priced the 206 case at 0.35, and nothing new has come in since, so moving again would just repeat the same mistake.
RECOMMENDED_P: KEEP
CLF P 0.58
Health
Weakening slowly. The main bottleneck story still holds, but two October items (DOE funding transformer R&D and Vantage's solid-state transformer order) are early signs that someone is working around the bottleneck.
Pillar P changes
Pillar 2 (AI + grid drives GOES demand): 0.65 -> 0.60.** Solid-state transformers use much less GOES, the special steel only CLF makes in the US. A 1 GW order is small, but it's the first real buyer choosing a design that needs less GOES, and DOE money speeds that path. Demand for transformers is safe. The share of that demand that needs GOES is now less certain.
Pillar 1 (sole US producer): no change at 0.80.
Pillar 3 (tariffs): no change at 0.55. The September executive order helped, and the August tariff scare turned out to be a misread.
Triggers
ADD:** Solid-state transformers win a utility or substation order of 1 GW or more, or total orders pass about 5 GW. That would mean they've moved past data-center pilots into the grid, which is where most GOES goes. Treat it as a partial break: cut Pillar 2 hard rather than setting P near 0.
RETIRE:** None. All four existing triggers still apply.
Valuation check
Stale. The ~$10.25 snapshot comes from before the 10/19 earnings and two months of news. Re-mark it against the current price and the post-earnings guidance for the GOES/electrical steel business before relying on the 2–3:1 upside-to-downside ratio.
Watch next
New transformer lead-time data. If the 128–144 week lead times start shortening, Pillar 2 and the whole thesis are weakening. That's the leading signal, and it matters more than another solid-state transformer headline.
Recommended P
The 10/21 move (0.68 -> 0.61) and the 10/23 move (0.61 -> 0.58) were both reactions to the same Vantage/Heron order. That counted one piece of news twice. Undo the second cut. The Pillar 2 change above is already covered by the first.
RECOMMENDED_P: 0.61
LEU P 0.75
Health
Stable.** The demand side keeps getting better: TVA's project was approved, the six New England states are backing new nuclear, and AEP signed a 960 MW deal in Indiana with Amazon. But the pillars that make this a Centrus bet, not just a nuclear bet, have had no new evidence. Those are Centrus's own enrichment output and DOE funding.
Pillar P changes
Pillar 4 (SMR/advanced-reactor pipeline): 0.50 -> 0.58.** Three real commitments in four weeks (TVA, New England, AEP) add up. This is no longer a story; buyers are signing.
Pillar 1 (sole US HALEU enricher): 0.75 -> 0.70.** "Sole" is a fact today. Over the 2030s it is less safe, because Orano, Urenco, GLE and General Matter all plan US capacity. The thesis needs Centrus to stay the only scaled supplier, and it says nothing about timing.
Pillars 2 and 3: no change. Nothing new on the Russia ban or DOE money.
Triggers
ADD:** "Centrus cannot secure financing for the Piketon expansion, or funds it with heavy share dilution (>25%)." Being a monopoly is worthless if the company can't pay to build the capacity. This is the most likely way the thesis fails, and it isn't on the list.
RETIRE: "Uranium <$60/lb sustained." REPLACE WITH:** "SWU (the price of enrichment work) falls below ~$150 sustained." Centrus sells enrichment, not uranium, so the uranium price is a weak signal for this business.
Valuation check
Stale. The "~1.5:1 skew" has no share price or date attached, so it can't be checked. Re-anchor it to today's price, today's market cap and the latest backlog figure.
Watch next
What was in the 2026-10-19 filing that disclosed a material agreement. Centrus's Q3 results in early November will likely say more. If it is a DOE capacity contract or a firm HALEU purchase deal, it is the first supply-side evidence in a month and moves Pillar 3. If it is a financing or share deal, check how much dilution it causes.
Recommended P
The last six moves have swung between 0.70 and 0.75 on demand headlines. The demand gains are real and they add up, so the floor shouldn't keep resetting to 0.70. But 0.75 prices in supply-side progress we haven't seen. 0.72 reflects better demand with unchanged execution risk.
RECOMMENDED_P: 0.72
MP P 0.38
Health
Weakening.** No thesis-breaking trigger has fired, but the claim that MP is the *only* Western supplier is wearing down. Tanbreez got approved, Phoenix Tailings is building a $375M New York plant, and the Trump–Xi thaw makes it more likely China eases its export controls.
Pillar P changes
Pillar 1 (only integrated Western REE→magnet supplier): 0.75->0.62.** Phoenix Tailings (4,400+ t/yr of rare-earth metal) and Tanbreez are real future rivals. Neither is mine-to-magnet yet, so MP's lead holds for now, but "only" is turning into "first."
Pillar 3 (China export restrictions sustain urgency): 0.62->0.50.** Taipei says Washington reassured it after the Trump–Xi meeting, which signals a thaw, and the November 2026 expiry is close. Whether China extends, eases or reimposes controls is now roughly a coin flip.
Pillar 4 (robotics/EV/defense demand grows structurally): 0.60->0.66.** The Oct 13 signals confirmed lasting Western demand for magnets.
Pillar 2 (DOD facility de-risks demand and pricing):** stays at 0.58. There's been no facility news either way.
Triggers
Sharpen** "a second Western integrated producer reaches scale" into something you can test: a non-MP Western producer making ≥1,000 t/yr of finished NdFeB magnets (the standard rare-earth magnet), with a US government or automaker offtake. Phoenix and Tanbreez on their own don't meet this.
Add** "the DOD price floor for NdPr (the main magnet metal) is cut, renegotiated or left unfunded." That floor, not China, is now the real support under MP's pricing. Losing it breaks the thesis even if China keeps its controls.
Sharpen** "China lifts controls" to: in November 2026, China extends its suspension *and* drops the licensing regime. A plain extension isn't a break.
Valuation check
The anchor is stale. Calling MP a pure "narrative" bet undersells the DOD price floor and offtake deal, which put a contracted floor under earnings. Re-anchor value on floor-backed cash flow plus a scarcity premium, rather than scarcity alone.
Watch next
Signals on China's November 2026 export-control expiry: any MOFCOM (China's commerce ministry) statement, any US–China trade-deal text, or license-approval data. This is the biggest swing factor for Pillar 3 and for the stock.
Recommended P
The pillar moves roughly offset, as Pillar 4 rising balances Pillars 1 and 3 falling. The earlier drop to 0.06 was an overreaction, and 0.38 is about right ahead of a binary event in November.
RECOMMENDED_P: KEEP
FCX P 0.35
Health
Weakening. Grasberg is running well below normal, and two major forecasters (ICSG and BNEF) now expect a copper surplus, not a deficit. That hits pillar 1 and pillar 3 at once, and FCX feels it more than any other copper name.
Pillar P changes
Pillar 1 (structural copper deficit): 0.80->0.50. ICSG and BNEF both forecast enough refined copper. Bloomberg's finding that the rally came mostly from traders stockpiling (buying ahead of US tariffs) means the price overstated real demand. The long-term AI and grid case is still alive, but the near-term "deficit" story is now contested.
Pillar 2 (big producer wins whatever tech wins): 0.75->0.70. The logic still holds, but it only pays if copper stays tight. Small trim.
Pillar 3 (Grasberg + Americas output sustained): 0.62->0.40. The mill ran at about 67% of normal in Q3 and guidance was cut. That is close to "sustained disruption," not a one-quarter blip.
Triggers
Add:** Grasberg mill stays below 80% of normal through Q1 2027, or the full-recovery date slips past 2027. This turns the vague Grasberg trigger into one you can test.
Add:** The stockpile unwinds. US copper inventories flow back out and the COMEX–LME price gap (the premium US buyers pay over London) collapses. That would confirm the rally was positioning, not demand.
Sharpen the "new supply" trigger:** fire it when two or more major forecasters show a 2027 surplus. Two already show one for the near term, so this is close to firing.
Retire:** none.
Valuation check
Stale. The ~2:1 upside-to-downside math assumed a deficit and full Grasberg output, and the $12,075/t copper price needs re-marking. Rerun the scenarios with Grasberg at about 70% and a flat-to-surplus copper market.
Watch next
FCX's Q3 earnings call, which falls around now. Focus on the Grasberg ramp timeline and 2027 copper volume guidance. No 8-K has reached the tracker yet, so check whether it's already out.
Recommended P
Slightly up from 0.35. Two of the six recent cuts (10-05 and 10-14) said outright that they reused evidence already counted, so about 0.14 of the drop counts the same news twice. No trigger has fully fired, and copper is still well above $10,000/t. Grasberg is close to tripping its trigger, though, so I'm only taking back about a third of the double-count, not all of it.
RECOMMENDED_P: 0.40
SCCO P 0.54
Health
Stable.** The surplus forecasts are real bad news for the shortage pillar. But the tracker cut P five times on what is really one fact: forecasters (ICSG, the industry's main forecasting body, then Bloomberg and BNEF repeating the same call) now expect a near-term copper surplus. That pushed P from 0.73 to 0.54 by counting the same story over and over. The cost-leader pillar and the Peru/Mexico pillar haven't changed, and copper at about $12,075/t is far above the $10,000 break line.
Pillar P changes
Pillar 1, structural deficit: 0.80->0.62.** Three forecasters agree on a near-term surplus, so this is a real hit. It isn't fatal: the thesis rests on the long mine cycle and AI-grid demand, and a 1–2 year surplus doesn't kill that.
Pillar 2, lowest-cost producer: no change. A surplus actually tests this pillar, and nothing so far says SCCO's costs have moved.
Pillar 3, Peru/Mexico: no change. There's no new policy or output news.
Triggers
ADD:** "Major forecasters push the surplus into 2027–28 AND LME inventories rise two quarters in a row." That's the line where a near-term surplus becomes a broken deficit thesis. Today's surplus call is only the first half.
ADD to the process (not a thesis trigger):** treat repeat coverage of the same forecast as one event. Five cuts in four days on one story is the main reason P is too low.
Valuation check
Getting stale.** The premium over FCX (Freeport-McMoRan, the main peer) was partly paid for the shortage story. Re-check it against the Oct 21 results, which I haven't seen. If SCCO still trades at a big premium while the shortage case is weaker, the stock has less room for error.
Watch next
LME and Shanghai copper inventories this week.** If stocks are actually building, the surplus forecast is turning into fact, and Pillar 1 should go lower. If stocks stay flat or fall, the forecasts are running ahead of the physical market. Also read the Oct 21 earnings for C1 cash cost (the cost to mine a pound of copper) and output guidance. That checks Pillars 2 and 3, and no one has done it yet.
Recommended P
A fair fix is to take the 0.73 prior and cut it once for the Pillar 1 hit, about 9–11 points. That lands near 0.62–0.64. I went with 0.62 because the earnings haven't been read yet.
RECOMMENDED_P: 0.62
Note: web search was blocked, so this review uses only the thesis and conviction log you gave me. I haven't checked the Oct 21 earnings numbers or current inventory data.
ETN P 0.89
Health
Stable.** Demand still looks solid, but all of the recent good news comes from peers like ABB and Schneider, not from Eaton itself. The jump to 0.89 is borrowed confidence.
Pillar P changes
Pillar 2 (backlog + pricing power): 0.75->0.78. ABB's US orders up 82% and strong Schneider demand point to a tight market for power equipment. That supports pricing, but Eaton's own book-to-bill (new orders divided by shipments) hasn't confirmed it yet.
Pillars 1 and 3: no change. On Pillar 3 the two signals cancel out. EV charging load grew only 8% (weak), while utilities rebuilding the grid looks strong.
Triggers
Add:** "Eaton's data-center orders grow well below peers (ABB, Schneider, Vertiv) for two quarters in a row." The thesis now leans on peer read-through, so the real risk is Eaton losing share while the overall market stays strong. No current trigger would catch that.
Retire:** "Durable AI demand-flattening (inference efficiency)." You can't measure it directly, and if it ever matters it will show up as the capex-cut trigger. Fold it into that one.
Valuation check
The anchor is stale because it has no number in it. "Premium multiple" can't tell you when the stock is overpriced. Set a forward P/E range (price divided by expected earnings) so a move to the top of that range triggers a review.
Watch next
Eaton's Q3 earnings (expected late October or early November).** Look at Electrical Americas organic orders, backlog, and data-center bookings. This is the first Eaton-specific test of the 0.89. Also read the Oct 16 "other material event" filing before the call. It hasn't been reviewed yet and could matter.
Recommended P
The P has swung between 0.82 and 0.92 in three weeks without any new Eaton data. Hold it below 0.89 until Eaton's own orders confirm what the peers are showing.
RECOMMENDED_P: 0.85
PWR P 0.84
Health
Stable. The sector news is still good (grid failures, the permitting fight, new-project data from ISO-NE, New England's grid operator), but nothing new has come from Quanta itself. The last four moves swung P from 0.94 to 0.78 and back to 0.84 on headlines, not company data.
Pillar P changes
Pillar 1 (transmission is the binding step): 0.80->0.83. Real grid failures, regulators focusing on grid upgrades, and ISO-NE's new-project data all point the same way: the wires are what's slowing new load.
Pillar 2 (backlog visibility): no change at 0.75. There is no new backlog number to update on.
Pillar 3 (labor moat): no change at 0.65. There is no new evidence either way.
Triggers
ADD: "Local opposition to data centers turns into federal or state siting limits (moratoria, restrictions on serving large loads)." It moved P down on 10/09. It's a real path to weaker demand that the current triggers don't cover, because it slows new load without any drop in AI demand.
No triggers to retire.
Valuation check
Stale. The anchor has no numbers. Add a forward P/E or EV/EBITDA range and a target book-to-bill (new orders vs. work completed) so "priced for execution" can actually be tested.
Watch next
Quanta's Q3 earnings, likely late October or early November (last year's Q3 came out October 30). The key numbers are total backlog, the electric-segment book-to-bill, and electric margins. These are the first hard data in weeks, and they test pillars 2 and 3 directly.
Recommended P
The pillars average out to about 0.74, and the current 0.84 sits well above that because of sector headlines. I'd trim it a bit until earnings either confirm the rise or reverse it.
RECOMMENDED_P: 0.80
GEV P 0.62
Health
Stable.** Every leading signal still points up: proposed US gas capacity is up 44%, and PJM (the grid operator for much of the eastern US) is studying a ~10 GW data-center load. None of the four break triggers has fired. The one negative, a broad AI-stock selloff, hits the multiple, not the business.
Pillar P changes
Pillar 1 (gas turbines are the near-term firm-power answer): 0.72->0.75.** The 44% jump in proposed gas capacity and the 10 GW PJM study are direct demand evidence. The NRC permit for TVA's small nuclear reactor (the BWRX-300) doesn't threaten this. Those reactors won't produce power until the 2030s, and GE Vernova co-owns the BWRX-300 design through its venture with Hitachi. The tracker scored that permit as a negative, and it shouldn't have.
Pillar 4 (the premium multiple is justified): 0.50->0.45.** Investors are selling AI stocks across the board. A premium-priced AI-power name is first in line when the market is willing to pay less for growth. The business hasn't changed, but the risk to the multiple has gone up.
Triggers
ADD:** "A top-4 hyperscaler (Microsoft, Amazon, Google, Meta) cuts data-center capex guidance." This is a measurable, early version of the vague "AI demand-flattening" trigger.
RETIRE:** None. All four still apply.
Valuation check
Stale.** "Premium, growth priced in" has no number behind it. Re-anchor to a forward P/E or EV/EBITDA (company value vs. operating profit) range, and update the $163B backlog figure after the Q3 report.
Watch next
The Q3 earnings report** (GEV usually reports in the last week of October). Check whether orders and backlog grew from Q2, whether turbine slot reservations went up, and whether Power segment margins met guidance. One report tests pillars 1–3 and the margin-miss trigger.
Recommended P
The two pillar changes roughly cancel out, and the pillar average (~0.66) sits close to the current 0.62. The recent 0.37 low was noise, not signal. There's no reason to move before earnings.
RECOMMENDED_P: KEEP
BE P 0.24
Health
Stable. The business evidence got better: the AEP/Google 702 MW deal is real proof that buyers beyond Oracle want this. The last two cuts in P came from a stock drop and Gulf war news, and neither one hurts the thesis itself.
Pillar P changes
Pillar 1 (on-site fuel cells solve the grid bottleneck for data centers): 0.62->0.70. A utility (AEP) is spending about $4B and Google signed a 20-year contract to buy the power. That is the strongest outside proof so far that this approach works at scale.
Pillar 2 (Oracle/Brookfield orders turn into revenue): no change at 0.58. We have no new data on order conversion yet. The strong quarter fits the thesis but doesn't prove it.
Pillar 3 (cost curve keeps improving): no change at 0.55.
Triggers
Retire** "Premium compresses on AI demand-flattening." It tracks the share price, which this framework counts as a lagging indicator. Last week's 3-month low after a strong quarter shows the problem: it set off a cut in conviction with no change in the business. Move it to the valuation notes.
Sharpen** the gas trigger to: "US Henry Hub (the main US natural gas price) above $6/MMBtu for a full quarter." Bloom's plants run on US pipeline gas, not Gulf gas shipped by tanker. Iran's attacks on Gulf shipping only matter if they push up US gas prices.
Add** "AEP Wyoming project delayed more than 12 months or its financing pulled." This is now the main proof point outside Oracle, so it carries the same concentration risk Oracle does.
Valuation check
Stale. The 116–230x earnings range is too wide to be useful and comes from before the Q3 results. Re-anchor it on forward earnings after the quarter.
Watch next
Whether Henry Hub gas prices rise with the Gulf conflict. If US gas stays flat, the 10/20 cut was an overreaction and should be reversed.
Recommended P
The running P (0.24) is now below where it was before the AEP/Google news (0.27), even though that news was the best evidence in this period. I recommend 0.34. That reverses most of the price-driven 10/23 cut and keeps a smaller discount for gas risk until Henry Hub shows whether that risk is real.
RECOMMENDED_P: 0.34
FSLR P 0.45
Health
Stable.** No trigger has fired. The newest evidence helps the thesis: solar power-purchase prices rose 5% in Q3 because of tariffs, which supports First Solar's module prices. Gas winning more data-center load is a real headwind, but it hurts slowly. Part of the drop from 0.67 to 0.45 came from noise, such as the 09-27 cut that was driven by gold.
Pillar P changes
Pillar 1 (cheap solar fills the AI power gap): 0.68->0.60.** Utilities keep adding gas plans for data-center load and cutting wind plans. The EIA's 82.7 GW renewables pipeline shows solar still gets built, but gas is taking a bigger share of new AI load than the prior assumed.
Pillar 3 (domestic manufacturing + 45X credits): 0.58->0.63.** The 45X manufacturing credit survived the 2025 budget law. Tariffs are now visibly raising US solar prices (the LevelTen data), so the protection is working in practice, not just on paper.
Pillar 2: no change (0.75), until the valuation anchor is re-checked (see below).
Triggers
RETIRE and SPLIT** "IRA / 45X repealed or cut." The demand-side credits (the 48E/45Y project credits) were already phased down in 2025, so that half has partly happened and is priced in. Replace it with two triggers:
45X manufacturing credit cut or capped** (thesis-breaking).
Foreign-entity (FEOC) rules or tariffs loosened** so cheap Asian modules get back into the US (thesis-breaking, because it removes the price shield).
ADD: bookings below shipments for two quarters in a row.** This is the earliest sign that buyers are walking away as the project credits phase out. It would fire before a 30% price crash shows up.
Valuation check
Probably stale.** The ~14x forward P/E anchor was set at an older price and older earnings estimates. Recompute it after the earnings report next week before trusting Pillar 2's 0.75.
Watch next
First Solar's Q3 earnings (late October).** Watch net bookings vs. shipments, the average selling price on new contracts, and any cut to 2027 guidance tied to the project-credit phase-out. This is the first hard read on whether tariffs are offsetting the loss of those credits.
Recommended P
I'm nudging P up slightly from 0.45, not resetting it toward the 0.67 prior. The pricing evidence improved, no trigger fired, and some of the earlier cuts were macro noise. But the shift to gas for data-center power is real, and earnings next week could move this either way.
RECOMMENDED_P: 0.48
ISRG P 0.62
Health
Stable, but this review is blind.** ISRG reported on 2026-10-20, and I couldn't read the results: my web search was blocked, and the filing contents weren't included. So nothing here reflects that quarter. Everything since July has come from the same summer hit to procedure growth, plus robotics news about other companies.
Pillar P changes
Pillar 2 (procedure growth + recurring revenue): 0.70->0.60.** Your notes say the July hit was never repaired. The pillar's number should show that, instead of hiding it in the running P.
Pillar 3 (da Vinci 5 upgrade cycle): 0.65->0.58.** Three REVIEW downgrades in nine weeks named the same unfixed issue. That is a pattern, not noise.
Pillar 1 (the only proven, profitable robotics model): no change at 0.85.** If anything, Figure scrapping its fleet backs this pillar up.
These changes bring the pillar numbers into line with the 0.62 running P. Right now the pillars still add up to the old 0.73.
Triggers
ADD:** "Full-year procedure-growth guide cut below 12%." This is an early warning that fires before the 10% break. A guidance cut is a leading signal; the 10% line only trips after the damage is done.
ADD (process rule, not a trigger):** Humanoid news (Optimus, Figure) can't move ISRG's P. On 09-25 and 10-03 the P jumped on other companies' news and then fell back within a week. Use B4 news to judge the humanoid theses against ISRG, not the reverse.
RETIRE:** none.
Valuation check
Still right as a quality anchor. But "premium quality" isn't a number. Add a forward P/E band or EV/sales ceiling so you can tell if the premium has gotten too high.
Watch next
The 10-20 results, mainly the full-year procedure-growth guide. Raised, held, or cut? Is it above or below 12%? Also check whether the da Vinci 5 placement mix fixed the issue your reviews keep flagging.
Recommended P
I'm holding at 0.62 because I haven't seen the one piece of evidence that matters, the 10-20 earnings. Don't treat this as a fresh read. Rerun this review with the results in hand. A guide held at 13% or higher with clean DV5 placements would justify moving toward 0.70. A guide cut below 12% would mean moving down to about 0.45.
RECOMMENDED_P: KEEP
VRT P 0.36
Health
Stable. None of the four triggers has fired. The cuts from 0.68 to 0.36 came mostly from opinion and credit-market mood (Gartner, CoreWeave bonds, Heatmap). The one piece of hard data, Schneider's record Q3, points the other way.
Pillar P changes
Pillar 2 (orders track hyperscaler capex): 0.70->0.65. CoreWeave's falling bonds show that neoclouds (newer GPU-rental clouds) may struggle to fund their builds. They are part of VRT's order base, and they are more likely to cancel or delay orders than hyperscalers.
Pillar 3 (premium valuation justified): 0.55->0.45. Gartner calling AI data-center spending "unsustainable" is exactly the story that shrinks a premium multiple, even before any capex is cut.
Pillar 1 (pure-play incl. liquid cooling): no change at 0.80. Nothing in the evidence touches VRT's position.
Triggers
ADD: "Customer financing stress turns into order cancellations or push-outs." Example: a neocloud or data-center developer defaults or delays a project, and VRT discloses a backlog hit. Why: the current triggers only watch hyperscaler capex. The credit stress shows up first at smaller, debt-funded buyers, and nothing in the list catches it.
RETIRE: none.
Valuation check
Stale. "Premium multiple" has no number attached. Set a hard anchor, such as forward P/E or EV/EBITDA versus organic order growth, and define what counts as "too rich," so trigger 4 can actually be checked.
Watch next
VRT's Q3 earnings, due around now (no 8-K has been filed yet). Watch organic orders and book-to-bill (new orders divided by shipments), and whether management says anything about cancellations or delays from neocloud or debt-funded customers. One good orders print would reverse most of the narrative-driven cuts. Weaker orders would count as the first quarter toward trigger 2.
Recommended P
The running P of 0.36 punishes the thesis for headlines while the hard data (no triggers fired, capex steady, a record quarter at its closest rival) holds up. I'd move it partway back up, but not to the prior, until VRT's own orders confirm. This is based only on the evidence in the tracker; I haven't checked live prices or VRT's results.
RECOMMENDED_P: 0.44
WULF P 0.44
Health
Weakening, but only a little.** Nothing at TeraWulf itself has broken. The real hit is to financing: the market for data-center debt is getting tighter, and WULF pays for its buildout with borrowed money.
Pillar P changes
Pillar 3 (financed without crippling dilution): 0.45->0.38. The CoreWeave and data-center bond selloff makes new debt costlier and harder to place. When debt gets expensive, equity becomes the fallback, and selling new stock is exactly the dilution this pillar is about.
The other three pillars don't move. Demand for power and AI hosting looks intact, since Cipher says power scarcity is holding leasing demand up. There's no news on WULF's contracts, ramp schedule, or zero-carbon pitch.
Triggers
Add:** "A hosting tenant (or its credit backer) shows financial distress or misses payments." The CoreWeave stress shows that neoclouds (newer, heavily borrowed GPU-cloud companies) are the weak link. If WULF's tenant wobbles, the contract is only as good as whoever stands behind it. Check whether WULF's leases have a big-tech guarantee (Google backstopped Fluidstack's earlier leases) and write that into the thesis.
Add:** "WULF's next debt deal is pulled, downsized, or prices at distressed yields (roughly 12%+)." Today's triggers only cover equity raises, but debt is the main funding channel. A failed bond deal is the earliest sign that Pillar 3 is breaking.
Retire:** none.
Valuation check
The anchor is partly stale. The "contracted data-center multiple" it targets has itself come down with the CoreWeave selloff, so the upside gap between a miner multiple and a data-center multiple is smaller than when the thesis was written.
Watch next
The trading prices of WULF's own bonds and convertible notes.** Falling prices mean lenders are pricing in more risk. That's the most direct real-time read on Pillar 3, and it beats sector headlines.
Recommended P
The tracker has been reacting to sector news, not company news. It swung from 0.55 up to 0.70 and back down to 0.44, and nothing WULF-specific happened along the way. The 10-16 and 10-18 down moves both counted the same CoreWeave bond story, so 0.44 overstates the damage. Starting from the 0.55 prior, the only evidence-backed change is a moderate cut for financing risk. That lands just under 0.50.
RECOMMENDED_P: 0.48
IREN P 0.44
Health
Stable, leaning weaker.** Nothing new has come out of IREN itself. The last two markdowns both come from one cause: lenders getting nervous about CoreWeave, a peer. That is a real risk to IREN's funding, but it should be counted once, not twice.
Pillar P changes
Pillar 3 (growth self-funds without heavy dilution): 0.50->0.42. If lenders are pulling back from GPU-cloud companies, IREN's next raise will cost more or lean on equity, which means more dilution. This is the one pillar the CoreWeave news actually affects.
Pillar 1 (owned power pipeline): no move. The 900 MW Wagga Wagga campus adds pipeline. But pipeline is not energized power. It only counts once a schedule and funding are attached.
Pillars 2 and 4: no move. No new data on how busy the GPUs are or on lease signings.
Triggers
ADD:** "GPU-cloud credit window closes." That means CoreWeave-type debt goes into distress, or IREN raises money at punitive terms (convertible coupon above ~8%, or equity issued well below the recent trading range). Why: the current dilution trigger only fires after the damage is done. This one fires first, and it's what the last two moves were really about.
RETIRE:** none.
Valuation check
The anchor is stale. The sum-of-parts value needs to subtract debt and convertibles, and it needs to value the GPU cloud on signed contract revenue minus financing cost, not on a peer multiple. CoreWeave's multiple, the obvious comparison, is falling.
Watch next
Whether IREN raises money before its November earnings, and on what terms. Cheap terms would mean the CoreWeave worry doesn't apply to IREN. Expensive terms or heavy share issuance would confirm the Pillar 3 cut.
Recommended P
The move from 0.47 to 0.44 counted the same CoreWeave lender worry a second time. Undoing about two-thirds of it, while keeping the real Pillar 3 cut, puts conviction at 0.46.
RECOMMENDED_P: 0.46
KEEL P 0.10
Health
Weakening.** No new company news since the Oct 2 review, but the three sector hits from September still stand: Pennsylvania permitting, state large-load tariffs that screen out speculative data-center demand, and the rate shock. Each one hurts a company that has only a pipeline and no tenant, and none of them has reversed.
Pillar P changes
The pillar odds still sit at their starting values (0.45–0.50) while the running P is 0.10. They no longer describe the thesis and need to come down:
Pillar 1 (2.2 GW converts to leased capacity): 0.50->0.20. Large-load tariffs and the PA permitting hit go straight at turning the pipeline into leases, and none of the 2.2 GW has a tenant.
Pillar 2 (signs a credible anchor lease): 0.45->0.20. No lease, no letter of intent, and no 8-K. Tenants will choose sites that already have firm power before they choose Keel's.
Pillar 3 ($533M funds the buildout without heavy dilution): 0.50->0.25. Q1 lost about $145M on about $37M of revenue. At that pace, plus buildout costs, the money runs out in a few quarters unless a lease brings in project financing.
Pillar 4 (spread across PA/WA/QC lowers risk): 0.50->0.35. Pennsylvania has already gone against Keel, but Québec's cheap hydro power and Washington still give it some cover. This is the strongest pillar left.
Triggers
Add:** Pennsylvania adopts a large-load tariff that requires collateral or proof of a committed tenant before granting interconnection, the right to connect to the grid. That cuts off a big part of the pipeline without a single permit being formally denied.
Add:** Liquidity (cash + bitcoin) falls below about $350M. That replaces the vague "liquidity falls sharply" with a number we can check each quarter.
Modify:** Put a start date on "a full year passes with no anchor lease." The clock should run from the rebrand or pivot announcement. Without a date this trigger can never fire.
Retire:** "Key pipeline sites stall in permitting" as a general trigger. Part of it already happened in PA. Replace it with the two specific triggers above.
Valuation check
The anchor is stale. Pricing the pipeline at a per-MW data-center value assumes it gets leased. Until a lease is signed, value Keel at its liquidity minus cash burn to date, and treat the pipeline as a small option on top.
Watch next
The Q3 report and whether the date has been announced. Q3 liquidity against the $533M figure is the number that tells us whether Pillar 3 holds or the dilution trigger is close. Any anchor-lease or LOI headline before then outranks everything else.
Recommended P
RECOMMENDED_P: KEEP
There's no new Keel-specific evidence since Oct 2, so 0.10 is still right. The swings between 0.05 and 0.12 in September came from counting the same sector-wide news, like the rate shock and the state tariff moves, as fresh Keel-specific damage more than once. Hold at 0.10 until Q3 numbers or an actual lease or financing event gives a real reason to move.