raw scansnapshot — prices as of scan date, not live69 rows · screens, not recommendationsported fromresearch/classic/scans/2026-04-26-ai-infrastructure.md
The AI infrastructure buildout is in a confirmed supercycle: LITE +165% 3M, VRT +75.6% 3M, FIX +53% 3M, CRDO +87.7% 30D, CLS +35.7% 30D — the physical layer of AI (cooling, power, networking, optical) is exploding while nuclear and solar names diverge sharply (SMR -52% 3M, OKLO -33% 3M vs NEE +12.2% 3M).
Quick Snapshot
Signal
Reading
Overall
🟠 PHYSICAL AI SUPERCYCLE CONFIRMED — Optical interconnects, power/cooling, and data center infrastructure are in synchronized breakout. Nuclear speculative names collapsing.
SMR (-20% 30D, RSI 37, -83% from ATH), FCX (-13.8% 7D, RSI 44), WOLF (+32% 30D but speculative-down trend), FSLR (flat, strong-down)
Key insight
The optical supercycle (LITE, CRDO, GLW, ANET, NOK) is the single most powerful theme in this scan. CRDO up +87.7% in one month. This is real AI demand flowing through fiber and coherent optics.
SMR (RSI 36.5, -83% from ATH, -51.7% 3M) — NuScale SMR thesis in freefall. Small modular reactor commercialization timeline pushed out too far. Avoid.
Oversold — Approaching Entry (RSI 35-45)
FCX (RSI 44.1, -13.8% 7D) — Freeport copper pulling back sharply on tariff fears. Copper is essential for all AI buildout (power wiring, cooling, chip packaging).
AWK (RSI 45.9) — rate-regulated utility, not infrastructure-levered.
RSI 51, death cross, -27% from ATH — cybersec thesis intact, wait for RSI 40.
Cross-Theme Analysis
The 3 sub-themes driving this scan:
Optical interconnect supercycle is real.LITE +165% 3M, GLW +85.6% 3M, CRDO +87.7% 30D, ANET +23% 3M. AI data centers need exponentially more fiber bandwidth between GPUs. Every major hyperscaler is in a bandwidth arms race. LITE at RSI 58 is the best entry in the scan — strong thesis, healthy valuation.
Power/cooling/electrical is the picks-and-shovels story.VRT +75.6% 3M, ETN +28% 3M, FIX +53.2% 3M, CAT +30.9% 3M. These companies build the physical infrastructure that makes AI possible. Unlike chips, they aren't competing with each other — they're all capacity-constrained by the same buildout. The backlog visibility is 2-4 years.
Nuclear speculative names have been destroyed.SMR -51.7% 3M (-83% from ATH), OKLO -33.3% 3M (-67% from ATH). The market bid up SMR/OKLO on the AI power narrative but is now repricing for timeline reality — small modular reactors are 5-10+ years from meaningful commercial deployment. NEE and CEG (existing nuclear capacity) are the safer expressions.
What Changed vs 2026-04-10
Previous scan (2026-04-10): Optical supercycle emerging. VRT and ETN extended but not extreme. Nuclear names (CEG, VST, OKLO) mixed. CRDO at much lower levels.
Today (2026-04-26): The optical and power/cooling themes went parabolic.CRDO up +87.7% in a single month. LITE is up +165% in 3 months. GLW +85.6% 3M. The physical AI infrastructure is no longer "emerging" — it's in confirmed breakout. Nuclear names have diverged sharply: utility-scale nuclear (CEG, NEE) treading water while pure-play SMR names (SMR, OKLO) have been crushed -33% to -52% in the quarter. The AI power thesis is real; the SMR execution timeline is not. CLS (Celestica) emerging as a key AI server manufacturer — +35.7% 30D to RSI 74, near ATH.