Semis, Quantum, Rare Earth, Nuclear, Space, Robotics: Broken Thesis or Just Oversold?

95 names across Semiconductors and ten 2024-26 story/infrastructure themes — Quantum Computing, Rare Earth & Critical Minerals, Nuclear/Uranium, Space, Robotics, Defense Drones, Optical Interconnect, Power/AI Data Center Demand, Security Software, and Cloud Software — are down 25%+ (many 50-80%+) from their 52-week highs while the broader index sits near all-time highs. Split by researched thesis status: 44 Intact/Accelerating, 40 Mixed/Early-Stage, 11 Broken.**

Why the index doesn't show any of this

The index's weight sits in a handful of mega-caps with real current free cash flow — the AI-infrastructure winners from our earlier gap reports (Micron, AMD, Dell, Cisco). Most of these 77 names carry almost none of that index weight, so a 50-80% drawdown across an entire cluster of stocks can happen without moving the S&P or Nasdaq at all — Amphenol and Constellation Energy are the two real exceptions here, large enough to matter at the index level, and both are down 14-39% anyway. What you're seeing is a valuation-cycle story, not a fundamentals story: quantum, rare earth, nuclear, space, robotics, defense drones, optical interconnect, and power-for-AI were 2024-2026's "prove-the-TAM-eventually" trades — re-rated on headline momentum long before revenue caught up, many still carrying P/S multiples of 20x-900x even after the correction. 2026 flipped into the "prove it now" phase. The data below shows that for the large majority of this list, the underlying businesses are still doing exactly what the bull case required — the stock just got too expensive first.

Method

Universe: Technology/Semiconductors names ≥$1B market cap trading at ≤65% of 52-week high, plus every identifiable Quantum Computing, Rare Earth/Critical Minerals, Nuclear/Uranium, Space, and Robotics name in the database (market cap ≥$100M, current price and 52-week high on file). Classification is research-backed, not a mechanical screen — each name's most recent guidance, contract wins, revenue trend, and cash position were checked against real reporting (SEC filings, earnings calls, press), not inferred from the price chart. "Mixed/Early-Stage" mostly means genuinely pre-revenue or too newly public to make a clean call, not "data unavailable."

Optical Interconnect and Power/AI Data Center Demand were added in a follow-up pass (July 16, 2026) covering the AI-infrastructure supply chain directly — connectors, optical transceivers/interconnect, and independent power producers/utilities with signed hyperscaler PPAs — same market-cap and drawdown thresholds, same guidance/contract-verification method.

Security Software and Cloud Software were added in a third follow-up pass (July 23, 2026) — cybersecurity platform vendors and cloud infrastructure/developer-platform providers, market cap ≥$650M (one name, Rapid7, sits under the $1B floor used for the core Tech/Semis universe but is included as the clearest guided-revenue-decline case in the software cohort), down 25%+ from 52-week high, same guidance/contract-verification method. This pass deliberately excludes the legacy enterprise-workflow SaaS names (ServiceNow, Atlassian, Intuit, Oracle) already covered by an earlier "AI-will-disrupt-us" gap report — see Bottom line #7 — since that's a different correction driver than the platformization/hyperscaler-bundling pressure hitting security and cloud-infrastructure vendors.

Thesis Intact / Accelerating — the drawdown is valuation, not fundamentals (44)

Semiconductors (2)

Quantum Computing (1)

Rare Earth & Critical Minerals (2)

Nuclear / Uranium (5)

Space (6)

Robotics (5)

Defense Drones (5)

Optical Interconnect / AI Datacom (7)

Power / AI Data Center Demand (4)

Security Software (2)

Cloud Software (5)

Mixed / Early-Stage — too pre-revenue or too new to call cleanly (40)

Semiconductors (7)

Quantum Computing (7)

Rare Earth & Critical Minerals (4)

Nuclear / Uranium (3)

Space (4)

Robotics (4)

Defense Drones (2)

Optical Interconnect / AI Datacom (1)

Security Software (5)

Cloud Software (3)

Thesis Weakening / Broken — the drawdown is deserved (11)

Semiconductors (4)

Nuclear / Uranium (3)

Space (2)

Defense Drones (1)

Security Software (1)

Bottom line

1. Roughly 88% of this list (84 of 95 names) is not a "broken thesis" situation — 44 are Intact/Accelerating (guidance raised, new contracts, revenue still compounding) and 40 are Mixed/Early-Stage, where the correction is a valuation reset on stocks that were never going to be judged on 2026 financials anyway. Only 11 show real fundamental deterioration.
2. The genuinely broken names cluster by a specific failure mode, not a theme: regulatory delay (Oklo — reactor still not NRC-approved, worst cash burn in company history), chronic underperformance (Virgin Galactic, Momentus), real demand misses (Navitas, Applied semis substrate, Skyworks, Photronics — all already flagged in the earlier gap report), or competitive share loss (Rapid7 — guided to a 2-3% revenue decline as Tenable/CrowdStrike/Palo Alto take vulnerability-management share). Notice these span every theme; none of these story themes as a category is broken.
3. Space and Nuclear are the most bifurcated themes — Rocket Lab, AST SpaceMobile, Planet Labs, Redwire, Kratos, Intuitive Machines, Firefly are all raising guidance and winning contracts (Space's Intact bucket is its largest), while Oklo and NuScale sit at the bottom on real regulatory/execution risk in the same sector. Same pattern in Nuclear: Cameco and Centrus just landed multi-billion-dollar contracts while Oklo/NuScale burn cash with no revenue. Don't trade "nuclear" or "space" as one basket — the dispersion inside each theme is bigger than the dispersion between themes.
4. Rare Earth is the cleanest theme — every name with enough data to judge is either Intact (MP, USA Rare Earth) or genuinely early-stage-but-real (NioCorp's financing is progressing, not stalling). No broken names identified.
5. Defense Drones is the theme where fresh reporting most directly overturned this database's own fundamentals field — RCAT's on-file revenue_growth_yoy reads -59% (likely a stale/mismatched comp), but its actual Q1 2026 revenue was +849% YoY on a new US Army contract that displaced Skydio; reclassified from Mixed to Intact on that basis. Ondas is the standout: FY26 guide raised to $390M+ and a $982M Army loitering-munitions prime contract via its Mistral acquisition. AgEagle (UAVS) is the one real washout — revenue down 62% YoY despite a new counter-drone JV announcement.
6. A useful screening rule from this data: "did guidance get raised or a major contract signed in the last 90 days despite the stock being down 50%+" separates almost the entire Intact bucket from the Mixed bucket cleanly — MP, Cameco, Centrus, Rocket Lab, Planet Labs, Redwire, Kratos, Intuitive Machines, IonQ, Symbotic, and Serve Robotics all hit that bar directly in this research pass.
7. This is a distinct pattern from the "AI-will-disrupt-us" correction hitting traditional enterprise software (ServiceNow, Accenture, Intuit, Atlassian, Oracle — all down 50-80% off highs on a different fear, covered in the earlier Tech gap report) — UiPath (Robotics/agentic-AI) is the one name in this list showing early signs of that same pattern despite genuinely improving fundamentals, worth watching as a bridge between the two corrections.
8. Optical Interconnect and Power/AI Data Center Demand are the two cleanest themes in the entire list — 11 of 12 names Intact/Accelerating (only IPG Photonics Mixed), zero Broken in either theme, and nearly every name carries a guidance raise or a signed multi-year contract from the last two quarters (Ciena's FY26 guide up to $6.3B, Amphenol's record Q1, Constellation's Microsoft/CyrusOne PPAs, Vistra's Meta/AWS nuclear deals). Yet CEG (-39.1%) and CIEN (-39.0%) are taking drawdowns as large as pre-revenue quantum names. That's the clearest evidence yet in this dataset that the correction is being applied almost mechanically by "how AI-adjacent does this look," with little regard for whether the underlying business actually proved anything this quarter.
9. Security Software and Cloud Software add a third, distinct correction narrative to this dataset — not the "AI will replace my SaaS app" fear hitting legacy enterprise workflow tools (point 7) and not the pure momentum unwind hitting quantum/space, but platformization: bigger security suites (CrowdStrike, Palo Alto Networks, Microsoft Entra/Defender) are bundling out point-solution vendors, and hyperscalers (AWS/Azure/GCP) are squeezing independent cloud-infrastructure vendors at the edge. Cloud Software is the cleanest theme in the dataset after Optical Interconnect/Power — 5 of 8 names Intact, zero Broken, guidance raised on 5 of 8 most recent prints (MongoDB, DigitalOcean, Dynatrace, Elastic, Fastly). Security Software has this dataset's clearest software-side Broken case: Rapid7's guided revenue decline as competitors consolidate its market — the platformization thesis made concrete. Palo Alto's $25B CyberArk acquisition (closed February 11, 2026) is the buy-side mirror of the same trend.