Correction 2026-05-31: this cadence-seeded event was superseded by the official NVIDIA IR date captured as 2026-05-20-nvda-earnings. Keep this file as lineage for pre-print watchlists; do not treat 2026-05-28 as the active earnings date.

NVIDIA's fiscal Q1 FY27 (covering Feb 1 → May 3, 2026) reports after-market on or around 2026-05-28 based on prior-year cadence. First earnings print since the AI Power Crisis perspective set crystallized (NuttyCLD series Mar 30 – May 1, Crusoe operator-side capture 2026-05-02). Both ai-power-bottleneck and ai-power-delivery carry direct read-through; this world-event is the first cross-perspective affects: test in the event-sourced system.

What this event tests for each perspective

ai-power-bottleneck (generation-side)

• Hyperscaler capex commentary. NVDA's customer language re: data-center power constraints. If Jensen surfaces explicit power-bottleneck framing on the call, the entire generation-side cohort (BE / CRWV / CORZ / IREN / APLD / GEV) gets a third-party validator beyond Aschenbrenner and Crusoe.
• Onsite gas-turbine + GE Vernova read-through. Crusoe's 2026-05-02 GEV bull is a single-operator read; if NVDA hints at hyperscaler customers shifting to onsite gas-turbine generation as the timeline-binding constraint, GEV bull thesis hardens to multi-operator consensus.
• Compute-depreciation language. Listen for NVDA's framing on H100 / Blackwell residual value. Lochmiller's read on 2026-05-02: H100 spot pricing has risen, suggesting 8-10 year depreciation is more accurate than the 5-6 year industry standard. NVDA commentary (or a sell-side analyst question) probing residual-value assumptions would be a calibration data point.

ai-power-delivery (delivery-side)

• MPWR Vera Rubin socket share. Sell-side projects ~70% socket share with ASPs ~60% higher than Blackwell. NVDA usually doesn't disclose vendor-specific power-delivery share, but Q1 FY27 is the window where Vera Rubin shipments first show in the customer mix — any framing on power-delivery cost per XPU or design-win vendor diversity is high-signal.
• TSMC A16 + Super Power Rail timing. TSMC pushed A16 SPR ramp to 2027 (was 2H 2026) per April 2026 NA Tech Symposium. NVDA's commentary on Feynman (2028 — likely first AI accelerator on A16+SPR) is the volume-validation signal for the BSPDN delivery-side thesis.
• Co-Packaged Optics (CPO) progression. Quantum-X / Spectrum-X Photonics generations. CPO at NVDA's roadmap pace pulls COHR / LITE forward. (Cross-cuts into optical-supercycle perspective when that one converts.)
• Power-solution dollar content per rack. Morgan Stanley estimates Kyber rack power-solution value rises >10x vs GB200. NVDA disclosure — even indirect — on rack BOM trends validates or breaks the multi-bagger framing on the delivery-side cohort.

Why this is a world-event and not an input

Per event-schema: an input is something a human/agent produced and we captured (the earnings call transcript, sell-side notes, Twitter reactions). The earnings release itself — the underlying real-world occurrence — is the world-event. When the print lands, derived inputs (the transcript, the AlphaSense summary, the sell-side notes, the @x_userhandle reaction thread) get their own input-type events with source.parent_event pointing back to this event_id.

This is the first event in research/world-events/; the pattern it establishes:

1. World-event seeded ahead of the date with status: scheduled and forward-looking watch_for-style content in the body.
2. On the day, status flips to captured (or superseded if the actual earnings date moved); body gets the actual print metrics.
3. Derived input events (transcript, sell-side notes) reference this event via source.parent_event.
4. thesis-shift events on ai-power-bottleneck and ai-power-delivery log.md files reference this event via source.parent_event if the print moves the thesis.

The graph: one world-event → N input events derived from it → M thesis-shift events that consume them. Each event-id is greppable; the cross-references make the multi-perspective fan-out queryable rather than embedded in prose.

Watch-for shortlist (pre-print)

Compiled from the affected perspectives' current watch_for fields, narrowed to what NVDA print can plausibly inform:

1. Power-infrastructure language in customer-mix commentary — generation-side validator
2. Vendor diversity / dual-source language on power delivery — MPWR socket-share concentration risk
3. Vera Rubin shipment commentary — first window where the Rubin power-delivery cohort gets revenue mix
4. A16 / Feynman timeline — TSMC SPR ramp validation
5. CPO product roadmap mentions — pulls COHR / LITE (cross-thesis with optical-supercycle)
6. Hyperscaler capex outlook — Microsoft / Meta / Google / Oracle / xAI commentary embedded in customer commentary
7. Compute residual-value framing — depreciation calibration

Cross-references

• Perspectives:
  • ai-power-bottleneck — generation-side; this event is in its watch_for
  • ai-power-delivery — delivery-side; first earnings since perspective spark
• Recent related events (informally — to be linked properly once those events are converted):
  • 2026-05-02 — Crusoe / Lochmiller transcript (operator-side validator) — inputs/podcasts/2026-05-02-crusoe-chase-lochmiller-ai-stack-economics.md
  • 2026-05-01 — NuttyCLD AI Power Crisis 6-part series final part — inputs/articles/2026-05-01-nuttycld-the-next-bloom-energy.md
• Schema reference: event-schema
• Master architectural brief: event-sourced-research-os