AI Power Utilities: Hold the Wires, Keep Cash for the Merchants

Daily Stock & Crypto Analysis · Wednesday, September 9, 2026 · Closing prices Sept 9, 2026 · @dailyanalysts
中文版 Chinese edition →

The call: HOLD AEP ($124.67) as the anchor; HOLD CEG ($293.91) core but add only $262–285; TRADE-BUY VST ($151.10) $140–152 — and keep 30–40% cash into PPI Sept 10 / CPI Sept 11 / FOMC Sept 15–16.

Why now: 10Y at 4.84% + Brent >$100 + 60–67% hike odds punish long-duration merchants into the macro event, while contracted AI load (AEP 69 GW, CEG Meta 1,121 MW from June 2027) is real but back-half loaded.

The disagreement: Consensus says "AI load = buy all power now"; ghost-megawatt audits (Texas freeze on 474 GW, AEP Ohio cut 30→13 GW) say half the pipeline never builds — wires get paid regardless, merchants don't.

The level that changes everything: Brent daily close <$88 kills the hike case and unleashes utilities; core PCE >0.23% or CPI >0.4% locks a Warsh hike and merchants drop another 10–15% — that's your add.

1. The macro storm matters more than the megawatts this week

Takeaway: Rates and oil, not load growth, set utility prices for the next 7 days. Dow fell 400 points for a 3rd straight day Sept 9 as Brent topped $101 and the 10Y hit 4.84%, highest since 2023. SPY closed $762.40 (-0.46%), QQQ $716.31, XLU $42.94 (-1.17%) while XLE $65.31 (+0.83%) diverged — classic inflation-week split. Schwab (Sept 2) pins it plainly: "most of the move up appears driven by a hawkish Fed and higher expected short-term rate," with nominal Q2 growth at 6.6% and July PCE +3.7% YoY. Warsh at Jackson Hole: "we must be confident underlying inflation is moving to our objective, clearly and at sufficient speed. Otherwise, we have work to do." Futures price 60–67% chance of a 25bp hike Sept 16, up from 35% a week ago. Evercore ISI's Guha says it comes to hundredths: core PCE ~0.21–0.22% = hold, ~0.23–0.24% = hike — and PPI (cons. +0.4% m/m, 5.3% y/y) hits Sept 10, CPI (cons. +0.4% headline / +0.2% core, 3.4% / 2.4% y/y) hits Sept 11, the day before this note settles. Add Iran-war oil (Brent $101.15 intraday per today's XLE note, WTI $96.30) and a $6B Treasury buyback the market called "underwhelming," plus Trump threatening trade retaliation if the Fed doesn't cut. Consequence: every long-duration asset — especially 30x P/E CEG — trades as a rates proxy until Sept 16. That's why you hold wires and reserve cash.

2. AI load is real — but the contracted part is smaller than the headline

Takeaway: About half of quoted "pipeline" is speculative; tariffs and freezes are now deleting it. Nationally the surge is undeniable: S&P Global sees US data-center demand >60 GW in 2025 → >80 GW in 2026 → 180 GW in 2030; GridStrategies sees +5.7%/yr use and +166 GW peak over 5 years; FERC sees PJM ~3.7% CAGR 2026–30; hyperscaler capex $650–700B in 2026. But Sept 1 changed the auditing: Texas — the largest hub — froze new data-center grid connections pending audit with AI requests at 474 GW, and AEP Ohio's pipeline fell more than half overnight after a $100,000 interconnection study fee (30 GW → 13 GW, per Utility Dive Mar 31 and Sept 7 coverage). SEPA counts 77 large-load tariffs across 36 states (29 approved in 2025) with upfront payments, exit fees, 80%-of-forecast minimums and 50-MW+ thresholds specifically to weed out ghost megawatts. PJM's July 14, 2026 auction for 2028/29 cleared at the $325 cap (down 2.5%) on 138,318 MW — third straight capped auction ($29 in 2024/25 → $270 → $330–333). Uncapped would be ~$530. Governors + FERC extended the collar and push a backstop auction forcing hyperscalers to fund 6–10 GW on 15-yr fixed contracts. Consequence for stock-picking: regulated wires (AEP) collect on whatever survives because hyperscalers pre-pay transmission; merchants (CEG/VST) only collect on what actually signs a PPA and clears interconnection.

3. AEP is the hold — 69 GW contracted, $78B capex, 4.5% yield cushions the hike

Takeaway: AEP converts AI talk into rate base even if half the pipeline evaporates. Q2 July 30: operating EPS $1.36 (miss vs ~$1.43 cons on weather/storm costs) but raised 2026 operating guidance to $6.25–$6.55 (mid $6.40, from $6.15–$6.45), lifted long-term CAGR to >9% through 2030 and contracted load to 69 GW (from 63 GW in Q1, 56 GW in Q4 2025, 28 GW in Q3 2025) — +6 GW in Q2 alone, mostly Texas, backed by letters of agreement and large-load tariffs approved in OH/IN/KY/WV. Capex $78B 2026–30 ($33B transmission, 42%), ~11% rate-base growth, +$10B line-of-sight (Piketon, Wyoming fuel cells), 13 GW gas turbines secured +10 GW evaluating, $16B lifetime cost offsets to existing customers + $1.4B federal loan/grant savings. Fundamentals Sept 9: $124.67, P/E 21.6x TTM / 18.9x forward, yield 4.48%, beta 0.47, 3-yr EPS growth 14%. At 18.9x vs sector ~18x 2027 (Gabelli Apr 7) it's fairly valued — no merchant torque, but no merchant drawdown either, and 4.5% yield vs 4.84% 10Y is the closest spread in utilities. Downstream: every 1 GW that survives = ~$6–8B investment need and ~$0.9–1B incremental revenue (Xcel math); every 1 GW deleted = zero earnings hit because tariffs force the customer to pay the study and exit fee.

4. CEG is the best asset at the wrong macro moment — wait for $262–285

Takeaway: Clinton/Meta + Calpine make CEG the only clean firm 24/7 seller at scale, but $294 into a hike is chasing. Largest US nuclear fleet (~55–60 GW incl. Calpine's ~21 GW gas/geothermal closed early 2026): Q1 revenue $11.12B +64%, EBITDA $2.78B, EPS $2.74 +28%; 2025 revenue $25.5B +8%; 2026 guide $11–12 adj EPS (light vs Street, hence -20% YTD to $293.91, 52-wk $228.63–$412.70, cap $105.9B, 30.6x TTM, beta 1.12, yield 0.91%). The load is contracted, not hoped: Meta 20-yr PPA for all 1,121 MW of Clinton IL (per June 3, 2025 Constellation release) starting June 2027 — replacing the expiring ZEC, +30 MW uprate, relicense 20 yrs, evaluating SMR on site — but, as Sept 8 coverage notes, "doesn't start paying until June 2027," nine months away. Plus CyrusOne 380 MW Freestone TX (total >1,100 MW), ~5,000 MW into PJM queue, $8.4B FCF 2026–27 → $11.5–13B 2028–29. Risks into rates: $17.6B shelf, $3.9B capex, TMI Crane restart pushed by PJM to ~2031, PJM cap politics. TIKR $499–505 by 2030 (~12–15% IRR on 10.6% growth / 18% margin / 25.2x) vs Street mean $368 (+40%) is believable only if you buy the dip — 30x into 4.84% is why it fell 4.9% on Sept 4's down day then bounced 4.9% (Sept 7 note: NRG +6.4%, CEG +4.9% while S&P -0.4%).

5. VST is the tradable fear — $231 fair value vs $151 price, but Texas headlines own it

Takeaway: VST's fear is headline-driven (Texas freeze, ERCOT softness), its catalysts are contract-driven — that spread is the trade. 43.6 GW diversified (62% gas) + retail: Q1 revenue $5.64B +43%, EBITDA $1.48B, EPS $1.35 +315%; full-yr operating profit guide $6.8–7.6B; >$10B cash 2026–27. $151.10, 22.6x TTM, ROE 41.5%, beta 1.38, yield 0.62%, 52-wk $132.66–$219.82, cap $50.1B — down ~30% from high. What's not in guidance: Meta 20-yr ~2,600 MW PJM nuclear PPAs + AWS deal + Cogentrix +5.5 GW gas (Jan 5) at mid-teens returns. Sept 9 Seeking Alpha: $231 fair value (+60%) on guidance boost; CEO Burke bought ~7k shares ~$135 Sept 1. But Texas freeze hits VST hardest (Comanche Peak + ERCOT retail), and Q2 showed mild-weather retail pressure offset by generation. Forward PEG ~0.5x on 37% growth looks cheap until a hike reprices all high-beta merchants. One-sentence others: TLN $319.67 (loss-making TTM, beta 1.63 — higher beta CEG proxy, skip); NRG $115.37 — Sept 4's +6.4% leader, cleaner momentum than VST but less nuclear scarcity.

Flagship trade — anchor AEP, stalk merchants

CallEntryTargetInvalidationHorizonConviction
HOLD / ADD AEP (flagship)$118–125 (add <$120 on hike scare)$140 (9% CAGR + div ≈ 12% total)Weekly close <$105 (rate-case / ROE-cut contagion)6–12mHIGH (guidance raise + 69 GW collateral + 4.5% yield = 2 signals)
HOLD CEG, add dip$262–285$368 Street mean; $499–505 by 2030 baseTMI past 2031 + shelf executed <$2501–3m / 1–3yHIGH long-term; WATCH short-term
TRADE-BUY VST$140–152$210–225 (stretch $231–260)Daily close <$132.66 May low1–3mSPECULATIVE (1 signal: contract value vs headline fear)

Sizing: 60–70% in AEP + CEG core; 30–40% cash reserved for post-CPI/FOMC washout. No new full-size chase in CEG >$310 or VST >$175. No contradiction with open EQ-XLE-1 BUY $64–66 → $72 (XLE closed $65.31): XLE = oil-inflation hedge that pays if the hike thesis (Brent >$100) persists; AEP = duration anchor that wins if it breaks. They offset by design.

Bull / base / bear — what actually moves these into year-end

Takeaway: CPI/FOMC decide the next 15%, contracts decide the next 50%. Bull 25%: PPI/CPI cool (<0.3% m/m), Warsh holds Sept 16, 10Y back to ~4.3% — XLU +8–10%, CEG to $340–360, VST to $190–210, AEP to $133–138. Base 50%: mixed data, hawkish hold with hike threat intact, 10Y 4.6–4.9%, Brent $95–105 — AEP flat-to-up, CEG $270–310 chop, VST $140–170 chop; cash earns its keep. Bear 25%: hot CPI (>0.4%) + hike Sept 16 + Texas audit deletes more pipeline — XLU -5–8%, CEG to $240–260, VST to $120–132, AEP to $112–118 (add). Triggers to watch: PPI Sept 10, CPI Sept 11, FOMC Sept 15–16 with updated dots, PJM backstop auction design, Texas audit headlines, CEG Q3 and VST Cogentrix close.

What would prove me wrong

Takeaway: Three falsifiable lines — if they break, I flip. (1) If AEP weekly closes below $105 on an Ohio ROE cut or Texas SB6 reversal, the "wires always get paid" thesis is broken — exit, don't average. (2) If CEG executes the $17.6B shelf below $250 or TMI is formally pushed past 2031 with Microsoft walking, the 12–15% IRR math collapses — downgrade to WATCH. (3) If Brent daily closes below $88 (kills hike/inflation leg per XLE call) and 10Y drops through 4.30%, the "reserve cash" call was wrong — deploy cash immediately into CEG/VST on the break, don't wait for $262/$140.

What to do

Takeaway: Hold your electrons, hoard your dollars for 7 days. Keep AEP (DRIP on), keep CEG core, keep VST starter; place stink bids (CEG $270, VST $142, AEP $118) good-through Sept 17 FOMC; do not chase Sept 4-style green days (+5% on down tape = short-cover, not signal). Revisit after CPI Sept 11 morning: hot print = cancel bids lower; cool print = lift VST to half size. Opinions are mine; facts above are sourced below.

Appendix

Contents: A. Data snapshot · B. Model & load-bearing assumption · C. Sources

A. Data snapshot (Sept 9, 2026 closes unless noted)

TickerPriceDay52-wkP/E TTM / fwdYield / beta
CEG$293.91-1.72%$228.63–$412.7030.6x / 45.7x*0.91% / 1.12
VST$151.10-0.41%$132.66–$219.8222.6x / 53.1x*0.62% / 1.38
AEP$124.67-0.60%$105.70–$140.5821.6x / 18.9x4.48% / 0.47
TLN / NRG$319.67 / $115.37-1.87% / -3.57%TLN $288–$451TLN neg EPS
XLU / XLE / SPY / QQQ$42.94 / $65.31 / $762.40 / $716.31-1.17% / +0.83% / -0.46% / -0.29%
BTC / Fear-Greed$78,284 (-0.29%)66 Greed

*Forward P/E via handler looks stale vs Street (CEG ~22x FY26, VST ~18x per TIKR/Fool) — use Street multiples for decisions, handler for levels. Macro: 10Y 4.84% Sept 9; 30Y ~20-yr high; Sept hike odds 60–67% (CME); Brent >$100–101; CPI July 3.4% y/y (+0.1% m/m); PCE July +3.7% y/y; Q2 nominal growth 6.6%. Dates verified: Sept 9 = Wednesday, FOMC vote Sept 16, PPI Sept 10, CPI Sept 11.

B. Model & load-bearing assumption

CEG $499–505 by 2030 assumes ~10.6% revenue CAGR (Calpine + Meta/CyrusOne), 18% op margin, 25.2x exit P/E = ~12–15% IRR from ~$262–294. Load-bearing input: contracted nuclear floor price (Meta Clinton + PJM capacity). If PJM backstop forces sub-$150/MW-day 15-yr builds or PTC is cut, exit multiple compresses to ~18x and target falls to ~$350 (–30%) — the "hold core, add dip" sizing. VST $231 assumes Cogentrix closes at mid-teens returns + Meta 2.6 GW enters 2027 guidance at >$6.8B EBITDA; if Texas audit strands ERCOT retail or leverage stays elevated, fair value collapses to ~$160 (TIKR bear). AEP $140 assumes >9% CAGR on $78B capex at 9.5% ROE; if allowed ROE cut to ~9.5%→9.0% or load deleted >20 GW, CAGR falls to 6% and target to $125 (hold, no add).

C. Sources (working URLs only)

Schwab — Why Treasury Yields Are Rising (Sept 2, 2026)
George Smith Partners — September FOMC: To Hike or Not (Sept 3, 2026)
CNBC — Fed call could hang on hundredths (Sept 8, 2026)
Constellation — Meta 20-yr Clinton 1,121 MW PPA (June 3, 2025)
Utility Dive — Large-load tariffs proliferate (Mar 31, 2026)
AEP — Q2 2026 earnings, guide raised, 69 GW (July 30, 2026)
PJM — 2028/29 auction at $325 cap, 138,318 MW (July 14, 2026)
TradingEconomics — 10Y 4.84% Sept 9, 2026
Prices via financial-data handler Sept 9, 2026 closes; news via news_feed/market_news Sept 9. Prior CEG/VST/AEP deep context: TIKR CEG-vs-VST June 16, 2026; AEP Q1 PR May 5, 2026; Gabelli Utilities Outlook Apr 7, 2026.