Frontier Tech — Fair Value Gap Screen (AI · Energy · Quantum · Space) (Frontier Gap Sector)

Executive Summary & Action Plan

Verdict

Buy the cash, avoid the story — Energy/Power is the cheapest theme, Quantum the richest. Across 24 names, only 10 trade below our IWANNAVY Fair Value. Top pick: NVIDIA (NVDA), +33% to FV and the one mega-cap that is both quality and cheap.

Price $194.83 | Market cap $4.72T | Target $260 | Upside +33.4% Screen top pick: NVIDIA (NVDA) · IWANNAVY Fair Value · price as of 2026-07-04 · Street consensus $301.62 (reference)

Abstract

This screen values 24 frontier-tech names across AI, energy, quantum, and space against a self-computed IWANNAVY Fair Value — a log-median of DCF, EV/EBITDA, and EV/FCF for cash-generative businesses, and a forward EV/Sales or backlog/milestone approach for the pre-profit majority — rather than analyst consensus. The result is a sharp bifurcation between cash and narrative. The AI-power complex (VST +29%, TLN +28%, CEG +23%) screens cheapest as a group, because out-year cash flows from AI power-purchase agreements sit largely outside current guidance. Quantum is the richest: every pure-play trades above our fair value at 100–670x sales, with the sub-$15M-revenue names (QBTS −65%, RGTI −64%, QUBT −50%) carrying the largest downside. AI is split — NVDA (+33%, low PEG) is the standout while AMD, PLTR, MRVL, and ALAB are priced for flawless custom-silicon ramps. Space is a barbell: backlog-rich LUNR (+68%) and RDW (+15%) are cheap while narrative names RKLB (−32%) and ASTS (−30%) are not. Net: value concentrates in cash-generative power and NVDA; story premiums are widest in quantum and second-tier AI.

Widest Gaps — where FV and price diverge most

Sorted by upside to our fair value: the 6 cheapest and 6 richest of all 24 names.

---

AI / AI-Semiconductors

Value concentrates in NVDA; the rest is priced for perfection. NVDA's 28x EV/EBITDA is undemanding against 85% growth (PEG well under 1). AMD, PLTR, MRVL, and ALAB already discount flawless custom-silicon and commercial ramps. MU is the cyclical trap — a low 15.9x peak EV/EBITDA masks near-peak FY27 earnings, so a through-cycle multiple leaves our FV below the tape.

---

Energy / Power — the AI-power nexus

The cheapest theme. Cash-generative IPPs (VST, CEG, TLN) trade ~23–29% below our forward-FCF/EBITDA fair value because out-year cash flows from AI PPAs and the Cogentrix/Calpine/Cornerstone deals sit outside current guidance. GEV is roughly at fair value (backlog capitalized). Pre-commercial SMRs (OKLO, SMR) sit near our probability-weighted FV — consensus prices near-full pipeline conversion we haircut for 2030–2032+ deployment and dilution.

---

Quantum — the richest theme

Every pure-play is pre-profit with trailing P/S of 98–670x. On disciplined forward EV/Sales — even generous to a path-to-margin story — all five screen above our fair value. IONQ is the least-overvalued (leader, ~$470M RPO, real system sales). The sub-$15M-revenue names (RGTI, QBTS, QUBT, ARQQ) carry the largest downside: trailing multiples are unsupported by recognized revenue and consensus prices option value we do not underwrite.

---

Space — a barbell

Bifurcated. Narrative names RKLB and ASTS trade at 68–145x forward sales, above fair value even crediting large TAM optionality. Backlog-rich LUNR (3–5x forward sales) is the cheapest name in the whole screen (+68%); RDW is also cheap. KTOS carries a real profitable defense franchise; PL is a high-margin EO-data compounder priced richly.

---

Fair Value Notes

Method: cash-generative names = log-median of DCF / EV/EBITDA / EV/FCF; pre-profit names = forward EV/Sales or backlog/milestone, flagged higher-uncertainty. Consensus is reference only.

• NVDA +33% — FY27 rev ~$391B, 76% DC gross margin; 23x fwd EBITDA vs 85% growth. Rubin ramp is the swing.
• VST +29% · TLN +28% · CEG +23% — IPP out-year FCF from AI PPAs not in guidance; TLN at ~9% FCF yield with buybacks.
• LUNR +68% · RDW +15% — backlog-to-revenue conversion at 3–5x forward sales; the cheap corner of space.
• AMD −31% · ALAB −34% · PLTR −23% — priced for flawless custom-silicon / commercial ramps; no margin of safety.
• MU −13% — HBM tailwind real, but FY27 EPS near-peak; through-cycle multiple caps FV below price (post-7/2 crash aftermath).
• QBTS −65% · RGTI −64% · QUBT −50% — trailing 470–670x P/S unsupported by recognized revenue; option-value only.

What to Watch Now

• Energy: Vistra Cogentrix close + any Meta/AWS PPA disclosure; CEG Calpine synergies; SMR year-end TVA decision. Summer power prices swing IPP FCF.
• AI: NVDA Q2 FY27 (Rubin contribution); MU HBM4 pricing / FY27 EPS; hard evidence of AMD MI400/Helios revenue conversion.
• Space: RKLB Neutron first-flight attempt (2026); LUNR/RDW backlog-to-revenue prints; ASTS satellite cadence (45–60 sats by end-2026).
• Quantum: Q2 prints (IONQ raised FY guide to $260–270M); watch for dilutive secondaries given persistent cash burn.

What to Watch Next — Structural

• Power: data-center load growth vs firm-capacity additions; SMR deployment timelines (2030–2032+) decide whether OKLO/SMR optionality ever converts to cash.
• AI: merchant-GPU vs custom-ASIC share split; whether HBM scarcity persists "well beyond 2027" (MU's through-cycle multiple).
• Quantum: logical-qubit / error-correction milestones and RPO-to-revenue conversion — the only path to a defensible step-up in multiples.
• Space: transition from lumpy hardware/launch to recurring margin-positive services; defense-budget flow ($156B reconciliation) into KTOS/RDW.

---

Special Notes — Scope & Methodology

Why SPCX (SpaceX) is not in the Space peer set. SPCX is deliberately scoped out of the pure-play screen: (1) it already carries dedicated IWANNAVY coverage (standalone tearsheet, 2026-06-18, fair value $145), and (2) at a ~$1.23T market cap with a Starlink-plus-launch model and a dual-class structure, it is a mega-cap outlier that would distort an EV/Sales screen built for $2–58B pure-plays. For completeness: on our existing fair value of $145 against the current $162, SPCX screens ~−10% (overvalued) — so including it would reinforce, not change, the theme conclusion that space narrative names are richly priced.

How the MU fair value is built. Our Micron fair value of $850 (bear $520 / bull $1,350) is the log-median of DCF, EV/EBITDA, and EV/FCF — but the decisive input is the earnings base we capitalize. FY27 consensus EPS sits near a cyclical peak of ~$112; the Street's $1,486 target applies ~13x to that peak number. We instead apply a through-cycle multiple to normalized earnings: $850 is ~7.6x peak EPS but ~12–15x a normalized mid-cycle EPS of ~$55–70. That is why our base sits −13% below the tape while consensus sits +52% above it — the gap is almost entirely a peak-versus-normalized methodology choice, not a different read on the current quarter.

What the MU fair value does NOT include. Critically, our base fair value does not extrapolate the memory/HBM demand surge. We normalize memory cyclicality by design — "this time is structural" (HBM TAM $35B→$100B by 2028, shortage persisting well beyond 2027) is the consensus bet we decline to underwrite in the base case. If that structural-scarcity thesis is right, the base is too conservative and the correct anchor is our bull case of $1,350 (+38%). Read the MU line on the chart as: overvalued only if memory stays cyclical; materially undervalued if HBM scarcity re-rates memory to a structural-growth multiple. The −13% base and +38% bull bracket exactly that debate.

---

Catalysts & Risks

---

References

• Constellation–Calpine / nuclear PPA coverage — Reuters Energy · IonQ / quantum sector updates — The Quantum Insider · Rocket Lab Neutron program — Rocket Lab IR
• Coverage inheritance: NVDA fair value re-derived this screen (prior tearsheet 2026-07-01); memory names cross-checked vs 2026-06-30 sector work.

Note: all prices are 2026-07-04 authoritative market data. Fair values are self-computed (IWANNAVY Fair Value); pre-profit names (most Quantum/Space, OKLO/SMR) use forward EV/Sales or backlog/milestone methods and carry materially higher uncertainty than the DCF-based estimates. This is a valuation-gap screen, not individual buy/sell recommendations — position sizing and diligence per name required.