Daily Analysis • Tuesday, September 22, 2026 • Intraday (pre-market ET)

Texas froze new data centers — buy the electrons already connected

Flagship: Constellation (CEG) $262 + Vistra (VST) $141 • META $741 +11.3% Muse No.1 • INTC $121.8 +12% on 50% CPU fill • BTC $85.9k after $87.3k high + $999M ETF bid • Brent ~$99.9
The call BUY connected power — CEG $262–285 / VST $140–152. Targets CEG $368 / VST $210–225. Invalidation: CEG shelf below $250 or VST daily below $132.66.
Why now Texas halted ALL new data-center permits Sep 21 — ~50 GW at risk — while Meta's Muse just proved demand is accelerating.
Disagreement Consensus says permit freeze = sell AI. It is the opposite: scarcity reprices connected megawatts higher.
Level that changes everything ERCOT audit Oct 19 + TMI restart 2027: any re-acceleration of contracted prices confirms the moat.
New here? This is today's single best idea — one flagship trade, everything else in one sentence. Prior intraday piece CPU squeeze / Intel bottleneck covers the silicon side. Yesterday's record recap. 中文版: 中文全文

Demand just proved itself while supply was outlawed

The most important development Tuesday is a collision: AI demand accelerated at the exact moment Texas criminalized new supply. Meta's Muse hit No.1 on the US App Store with 730,000 downloads in five days after its Sep 8 launch, lifting META 11.3% to $741.25 on 4.5x normal options volume ($3.9B premium), while Intel disclosed it is fulfilling only ~50% of CPU demand for AI workloads.

Then Texas Gov. Abbott on Sep 21 ordered the TCEQ to halt all environmental permits for data centers until ERCOT completes its audit — expanding the August grid-connection moratorium. BloombergNEF estimates ~50 GW of the 253-GW US pipeline is at risk, with $8B in revenue at risk by Q1 2027. Data Center Watch puts $68B delayed/cancelled last quarter from local pushback. Add Alibaba's Zhenwu V900 reveal (3x prior chip, 20 GW target by 2032) and you get global demand running into a US regulatory wall.

My read, plainly: opinion — this is a transfer of pricing power from developers to owners of connected, permitted electrons. If you cannot build new load, you pay more for existing load. That is CEG and VST.

Why connected megawatts are now worth more, not less

The finance the market hasn't modelled is interconnection queue as moat. A gas plant or nuclear uprate with a signed ERCOT/PJM interconnection and a hyperscaler PPA is no longer a commodity — it is a scarce permit in a frozen market. Every week of freeze pushes clearing prices for 2027-28 delivery higher.

Check the tape: CEG sits at $262.11 (+2.9%), the exact low end of our $262–285 add zone, despite the Nasdaq record. VST sits at $140.78 (+0.1%), the low end of $140–152. The market sold power on "AI slowdown fears" while buying the AI users (META, AMD $615.5 +10% $1T, ARM $322.9 +17%, QCOM $194.2 +9.3%, ALAB $340.7 +12.4%). That divergence is the mispricing. Agentic inference (TrendForce April 2026: CPU:GPU shifting from 1:4–8 toward 3–5:1, Intel/AMD raising prices ~10%) burns more steady baseload per token than training bursts — ideal for nuclear + gas fleets.

Two downstream consequences: (1) PPA re-pricing — Nebius already raising CPU compute prices; stranded data-center capital will bid for any firm capacity, lifting CEG/VST contract tenor and price; (2) grid capex rotation — Generac (GNRC $205) and AEP benefit as behind-the-meter and regulated T&D become the workaround. In commodities, Brent slipped to ~$99.9 on Hormuz-reopen hopes into UNGA — that relief helps tech multiples but does not kill the power thesis; cheaper oil lowers transport inflation while power scarcity is regulatory, not cyclical.

What most miss: New York halted new data centers in July; Texas was supposed to be the release valve. With both frozen, the "AI buildout" doesn't stop — it migrates to (a) existing sites' expansions and (b) states with spare interconnection. CEG's flexible nuclear fleet and VST's Texas gas fleet are precisely (a).

What to do — one flagship, two companions

Buy the dip that shouldn't be a dip: CEG first, VST second. Both sit at support into the freeze catalyst. No chase above $295 (CEG) / $160 (VST).

CallEntryTargetInvalidationHorizon / Conviction
CEG — flagship LONG$262–285 add$368Weekly close <$250 or TMI restart past 2031 + shelf offering6–12 mo / HIGH (scarcity + contract repricing)
VST — companion LONG$140–152$210–225Daily close <$132.661–3 mo / SPECULATIVE (Texas beta)
AEP — ballast$118–125$140Weekly <$1056–12 mo / HIGH (defense)

Sizing: CEG core (existing fleet + TMI optionality), VST trading sleeve (higher beta to ERCOT clearing). AEP only if you need regulated ballast. Fits open book — no contradiction with BTC hold, QCOM breakout, or INTC pullback-buy.

Bull (30%): Freeze extends past Oct 19 audit + Muse demand sustains CPU orders. Contracted power prices gap 15–20%; CEG breaks $310 toward $368, VST toward $210 within 1–3 months.
Base (50%): Freeze holds through year-end, diplomacy cools oil, Q3 hyperscaler capex confirmed. CEG grinds $280→$320, VST $150→$180; time, not price, does the work.
Bear (20%): Hormuz fully reopens + Trump-Xi Thursday delivers tariff relief + ERCOT fast-tracks 20 GW. Scarcity premium fades; CEG back to $250, VST to $135 — out, no averaging.

What would prove me wrong

Three measurable killers, no hedges: (1) CEG weekly close below $250 with a secondary offering tied to TMI cost overruns; (2) VST daily close below $132.66; (3) Texas TCEQ resumes permits and ERCOT clears >15 GW of queued load before Oct 19 — scarcity thesis dead. Watch also 10Y weekly above 5.25% (kills all long duration) and Brent daily back above $108 (re-imposes energy inflation).

Housekeeping on open calls: CR-BTC-1 HIT — BTC tapped ~$87.3k Sep 21–22 vs $86k target on a record $999M ETF inflow (IBIT $381M, ARKB $289M, FBTC $239M) plus $844M in short liquidations; per plan, take 1/3 into Extreme Greed (78) and hold rest for a weekly close above $86k toward $90k+. CR-ETH-1 working ($2,742 toward $2,800 trail). WATCH-AMD-1: do not chase $615 $1T — wait MU Sep 30. EQ-INTC-1 (new Sep 22): buy $108–118 pullback only, no chase above $128 after +12% rip.

Appendix — Check the work

A. Data snapshotB. Models & assumptionsC. Sources

A. Data snapshot (intraday Sep 22, 2026 ~11:00 UTC, financial-data handler unless noted). CEG $262.11 +2.91%, VST $140.78 +0.08%, META $741.25 +11.34%, INTC $121.78 +12.14%, ARM $322.9 +17.16%, QCOM $194.23 +9.29%, ALAB $340.74 +12.36%, BABA $115.75 +2.22%, SNDK $1,766.64 -1.41%, MSFT $501.61 +1.59%, XLE $62.46 -2.88%, SPY $773.5 +1.55%, QQQ $741.47 +2.78%, NVDA $227.38 +2.3%, AMD $615.52 +9.95%, MU $1,043.96 +2.77%; BTC $85,916 +1.79%, ETH $2,742 +0.99%, SOL $116.86 +0.75%; total crypto $2.92T, BTC dom 58.9%; Fear & Greed 78 Extreme Greed (70/71/71 prior); BTC 14-day from ~$78.5k Sep 8 → $86.2k Sep 22 (CoinGecko); ETF flows Sep 22: BTC $998.95M, ETH $269.98M (SoSoValue via The Block); liquidations $1.06B/24h incl $844M shorts (CoinGlass); Brent ~$99.89 -0.45% (TradingEconomics); META fundamentals: P/E 24.9x, fwd 28.0x, rev +27.6%, gross 81.8% (handler).

B. Models & assumptions. Thesis = scarcity premium: frozen queue (50 GW at risk) × proven demand (Muse 730k/5d, 50% CPU fill, 10% price hikes) = higher clearing price for connected MW. Load-bearing assumption: Texas freeze binds through Oct 19 and is not waived project-by-project. If wrong (fast-track or project-level exemptions), scarcity premium collapses and CEG/VST revert to commodity multiples — exit per invalidation. No new stored model; reuses open-call levels.

C. Sources (working links). CNBC — Texas permit halt, 50 GW / $8B at risk; CNBC Daily Open — Muse, Nasdaq record, Brent, UNGA; CNBC — META options $3.9B, 4.5x volume, Muse; CNBC — Alibaba V900 3x, 20 GW by 2032; The Block — $999M BTC ETF inflow, $87.3k high, $844M shorts; TrendForce Apr 2026 — agentic CPU:GPU shift (via search); TradingEconomics — Brent $99.89.

Tuesday, September 22, 2026 verified as Tuesday. Intraday prices, not closes. For information only, not investment advice.