Stock-Market Research Dispatch — 2026-05-29

Auto-generated by daily research routine. Not a source; not ingest-pending material. Summary of what changed in the wiki today.

Top of mind

CEG/TMI PJM interconnection delay is the most important new risk in the wiki today. The prior near-term nuclear catalyst (Crane Clean Energy Center / Three Mile Island Unit 1 restart, 835 MW, 2027 target) now faces a documented slip risk to 2031 per Constellation Energy's own Q1 2026 commentary. The 4-year delay would disrupt Microsoft's 20-year PPA and the $1.6B restart investment, and transforms CEG from a near-term revenue catalyst into a multi-year execution risk. CEG's stock is already -13% YTD. Calibration filed. The nuclear-for-AI thesis itself (structural demand for baseload) is intact — the problem is the timing of the most concrete near-term catalyst. Hyperscaler demand hasn't changed; PJM's interconnection queue is the new binding constraint.

FERC end-of-June ruling is the new near-term gate for the datacenter electrical thesis. FERC is considering streamlined large-load interconnection for hyperscaler requests (500 MW–1 GW at single points). A favorable ruling by end of June 2026 would directly unlock the ~7 GW of stalled 2026 US datacenter builds (delayed by electrical bottlenecks) and accelerate order conversions at ETN/VRT/FIX/PWR. This is more actionable than any single capex announcement and was not previously tracked in the wiki.

LIN guidance sandbagged on helium — pure upside optionality not priced into the street. LIN's Q1 2026 EPS was $4.33; FY2026 guidance is $17.60–$17.90. Critical: management explicitly stated that guidance does NOT include improvements in the helium business vs. February guidance — "any incremental volumes or price would be upside." LIN has created a structure where the consensus model has NO helium repricing embedded. APD Q3 FY2026 (late July earnings) is the first full post-strike quarter and will be the tell on whether APD's contract structure captures the repricing. Until then, LIN >> APD on thesis clarity.

Taiwan VP "no timetable" for chip tariffs (May 28) definitively clarifies Phase 2 asymmetry. Taiwan's zero-tariff MOU is operative. Phase 2 Section 232 burden falls primarily on Samsung Korea, SK Hynix Korea, and SMIC — not TSMC. Intel IFS (zero imports) and Micron (US-domiciled HBM) are the cleanest US-domicile beneficiaries. New hypothesis question created.

New wiki pages

Significant wiki updates

nuclear-baseload-for-ai-data-centers — multiple additions and one CONTRADICTION:

• ⚠ CEG/TMI PJM delay 2027→2031 risk added to Contradictions section. CEG conviction downgraded (Medium-high → Medium). Calibration filed.
• Kairos Power Hermes 2 groundbreaking: first Gen IV NRC construction permit issued. 50 MW FHR; Google PPA via TVA; target ~2030. This is the most significant SMR regulatory milestone since Oklo's DOE NSDA approval (March 2026).
• X-Energy IPO: Nasdaq XE, $1.02B raised at $23/share (above $16-$19 range), April 24, 2026. Upsized offering signals institutional demand for SMR pure-plays.
• Oklo 14 GW pipeline + Oklo-Vertiv collaboration on nuclear DC power/cooling integration — picks-and-shovels crossover with the DC electrical thesis.
• Uranium LT price: $90/lb in Q1 2026 — highest since 2008. Structural repricing underway (utilities prepaying because they've concluded spot cannot be relied upon).
• UEC (Uranium Energy Corp): analyst consensus target ~$18 (vs ~$13.65 current, +32%); added to suggested-tickers.

helium-supply-crisis-semicap — LIN/APD asymmetry and ceasefire collapse:

• LIN Q1 EPS $4.33; FY2026 guide $17.60–$17.90 explicitly excludes helium (pure upside optionality sandbagged — floor, not central case).
• APD Q2 FY2026 ended March 31 = 13 days post-strikes. The 4% EPS drag from "lower helium pricing" reflects the PRE-STRIKE environment. APD Q3 (April–June, late-July earnings) = first real post-strike tell.
• Helium spot prices: NE Asia $144.58/MCF; North America $62.79/MCF; Europe $52.51/MCF (2.3x NE Asia premium reflects LIN/APD storage buffer absorbing NA/EU pressure exactly as the moat thesis predicts).
• Iran ceasefire collapsing May 28-29: US CENTCOM struck Iranian attack drones and a ground control site at Bandar Abbas; Iran calls it a ceasefire violation; Trump has not signed the "tentative agreement" per CNN. This does not shorten the 3-5 year repair timeline — it may extend it.

hbm-supply-bottleneck — HBM4 dynamics clarified:

• Samsung cleared Nvidia HBM4 qualification at both 10Gb/s and 11Gb/s (SK Hynix still optimizing 11Gb/s — Samsung has a temporary edge). Samsung targeting 28% of Nvidia HBM4 for Rubin; capacity ramping ~170k → 250k wafers/month by end-2026.
• SK Hynix trimming HBM4 volumes 20-30% on 1c DRAM yield challenges — a surprise for the market leader. Remains ~50% global bit output but share declining.
• Micron began HBM4 volume shipments one quarter ahead of schedule (Q1 2026 vs Q2 guidance) — but positioned for mid-tier Rubin CPX (inference), not flagship Vera Rubin training. 2026 HBM 100% committed.
• 16-layer HBM4 race is the next leg: supply constraint self-renews with each generational step as stacking complexity reduces yields.

cowos-packaging-capacity-crunch — Vera Rubin demand pressure and Nvidia $150B Taiwan commitment:

• Nvidia $150B/year Taiwan pledge (Computex, May 27-28); Huang met TSMC Chairman C.C. Wei specifically to discuss Vera Rubin CoWoS production commitments.
• Vera Rubin's six-die architecture demands more CoWoS interposer area per unit than GB300 — TSMC must simultaneously ramp for GB300 tail AND Vera Rubin front-end ramp on an already-oversubscribed line.
• CoWoS customer allocation at 130k WPM target: Nvidia 60% (~800-850k wafers/yr); Broadcom ~20% (~240k/yr) (Google TPUs, Meta custom, OpenAI); AMD ~11% (~105k/yr). CoWoS-L/S fully booked through 2027.
• GTC Taipei June 1, 2026 = next milestone for Vera Rubin commentary.

section-232-phase-2-us-fab-premium — Phase 2 asymmetry clarified:

• Taiwan VP May 28: "no timetable" for chip tariffs on Taiwan. Commerce/USTR implementing Taiwan relief with 3 MOU preferential tracks.
• Phase 2 tariff burden falls primarily on South Korean suppliers (Samsung, SK Hynix) and SMIC — not TSMC (protected by Taiwan MOU) and not US-domiciled fabs (Intel IFS, Micron).
• Cross-link added to new samsung-skhynix-section-232-phase2-asymmetric-exposure hypothesis.

datacenter-construction-electrical-picks-shovels — FERC ruling and Meta capex update:

• FERC large-load interconnection ruling expected by end of June 2026 — this is the single most actionable near-term regulatory catalyst for the DC electrical thesis. A favorable ruling unlocks the stalled 7 GW of delayed 2026 builds.
• Meta capex raised to $125B-$145B (May 27 shareholder meeting). Total hyperscaler 2026 capex: $725B → projected >$1T in 2027. At 60%+ flowing to power/construction, the addressable market for VRT/ETN/FIX/PWR is ~$435B+ in 2026 alone.

Watchlist

Undervalued candidates (today)

• LIN (helium-supply-crisis-semicap): FY2026 guide explicitly excludes helium as upside. Street models have zero helium repricing embedded. With a 3-5 year repair timeline confirmed by QatarEnergy, Linde's ~6-month global storage buffer controls when the shortage bites customers — the moat is durable across the entire repair horizon. The question is whether LIN's multiple already reflects the JP Morgan upgrade ($455→$525 PT). Until LIN explicitly confirms contract escalators in earnings, there's structural upside optionality.
• MU (hbm-supply-bottleneck + samsung-skhynix-section-232-phase2-asymmetric-exposure): Micron is at the intersection of two independent structural tailwinds: the HBM supply bottleneck (only US-domiciled HBM producer, 2026 100% committed) AND potential Phase 2 tariff premium over Korean-origin HBM (Samsung/SK Hynix Korea). If Phase 2 triggers post-July 1, Micron gains pricing power by default. Both tailwinds are imperfectly priced independently; together, they may be substantially unpriced.

New hypotheses opened (not yet graded)

• nextera-dominion-merger-ai-power-consolidation — Priority: medium. Gate: FERC + state PUC approvals, 12-18 months. Near-term catalyst: FERC end-of-June large-load ruling (relevant to merged entity's competitive positioning). The combination of Dominion's Northern Virginia grid (highest AI DC density globally) + NextEra's nuclear restart and renewable expertise is structurally compelling — the question is regulatory risk in a $67B utility merger.
• samsung-skhynix-section-232-phase2-asymmetric-exposure — Priority: medium. Gate: July 1, 2026 Commerce report + any US-South Korea semiconductor MOU announcement. This hypothesis resolves as a binary: either South Korea gets Taiwan-equivalent treatment (falsified) or it doesn't (confirmed, MU/INTC beneficiaries).
• meta-cloud-business-fourth-hyperscaler-disruption — Priority: low-medium. Gate: any Meta cloud pilot or Q2 2026 earnings commentary on compute monetization. The $125B-$145B capex overshoot thesis depends on overbuild occurring; if Meta's internal AI demand absorbs all capacity, the hypothesis fails silently. Watch Meta's actual vs. projected capacity utilization.

Calibration logged

1. CEG/TMI 2027 restart target — PJM interconnection delay to 2031. Prior: 2027 online date was the primary near-term nuclear-for-AI catalyst (medium-high confidence). Updated: 2031 delay risk flagged in CEG Q1 2026 commentary (low-medium confidence on timing now). Pattern: grid-interconnection-risk-underestimated — same class of error as the helium-repair timeline: assumed regulatory timelines proportionate to urgency, not gated by queue position in an overwhelmed interconnection process.

Open questions for human review

1. FERC end-of-June ruling timing. This is now the most actionable near-term gate for VRT/ETN/FIX/PWR. Should this be a formal watch item with its own question page? A favorable ruling would likely catalyze the deferred ~7 GW of builds converting to active orders — potentially visible in Q3 2026 backlogs.
2. CEG: hold, size down, or exit? Stock already -13% YTD. PJM delay risk to 2031 is now documented. The Calpine integration narrative (gas acquirer with nuclear assets) has already disrupted the pure-play premium. The nuclear-for-AI structural thesis doesn't require CEG specifically — BWXT, CCJ, and the broader reactor-restart universe are not PJM-dependent. Worth reviewing position sizing if CEG was sized as a near-term catalyst play.
3. LIN position sizing. The street model has zero helium repricing embedded in FY2026 guidance (LIN confirmed). With NE Asia spot at $144/MCF vs. NA $62/MCF and LIN's storage buffer controlling when NA/EU customers face the shortage, the repricing cycle is real and multi-year. What does LIN's next contract roll cycle look like? When does the repricing actually hit their income statement?
4. APD Q3 FY2026 earnings (late July). Mark calendar. This is the primary binary for the APD leg of the helium thesis. Q2 ended March 31 (13 days post-strikes, pre-strike pricing). Q3 reflects the full post-strike environment. If APD Q3 shows helium pricing improvement, the thesis confirms for APD specifically. If not, the LIN >> APD asymmetry may be structural, not timing.
5. Podcast feeds (14th consecutive 403 failure). Network policy block persists. No action until run from local environment.
6. GTC Taipei June 1. Jensen Huang Vera Rubin announcements — CoWoS-specific commentary and shipment timeline will either confirm or adjust the six-die CoWoS demand pressure thesis.

What I looked at

• Step 1 thesis refresh topics: Nuclear/AI DC PPA update; helium LIN/APD valuation + Iran ceasefire; HBM4 supply dynamics; CoWoS/Nvidia $150B Taiwan; Section 232 Taiwan relief breaking news; macro bucket 7 exec capex May 27-29
• Step 2 macro buckets: Bucket 7 (exec capex statements — Meta $145B, FERC gate, $725B→$1T); Bucket 1 (semis/AI infra — covered via HBM4/CoWoS refresh); Bucket 4 (China/geopolitics — covered via Taiwan Section 232 relief and helium/Iran)
• Podcast episodes transcribed: 0 (all RSS feeds HTTP 403; 14th consecutive failure)
• Sources promoted: 6 (2026-05-29 autoresearch batch); sources ingested: 6
• Valuation snapshots refreshed: 0 (WebFetch 403 — 14th consecutive)
• New entities created: 1 (constellation-energy/CEG)
• New question pages created: 3 (nextera-dominion merger, samsung-skhynix-232, meta-cloud-business)
• Conviction changes: CEG downgraded (Medium-high → Medium) on PJM delay risk; no upgrades today
• Calibration logged: 1 (CEG/TMI PJM interconnection delay)