XtalPi

Phase A — Understand the business

Lens 1 · Company Overview

XtalPi is a contract AI-drug-discovery and materials-science platform, not a drug developer. It sells the front half of the R&D pipeline — target-to-candidate discovery — as a service, and deliberately does not take assets into human trials. Founded 2015 in Shenzhen by three MIT-postdoc quantum physicists — Wen Shuhao (Chairman), Ma Jian (CEO), Lai Lipeng (Chief Innovation Officer). The pitch: use quantum-mechanics calculations to generate high-fidelity molecular datasets → train proprietary AI models → screen billions of molecules → validate the hits in in-house robotic wet-labs (the "ID4"/"Intelligent Digital Drug Discovery & Development" and "lab-in-the-loop" platform). Robotic labs run in China and the US.

Two reported segments:

1. Drug Discovery Solutions — AI+robotics discovery for pharma/biotech partners; FY2025 revenue RMB 537.9M (+418.9% YoY).
2. AI4S Intelligent Solutions (renamed twice: "intelligent automation" → "intelligent robotics" → "AI4S") — automated synthesis, R&D-lab robotics, and materials-science work (TCM, electrolyte/battery, perovskite solar); FY2025 revenue RMB 264.7M (+62.6% YoY).

Contract structure is the crux. Historically fee-for-service (per-programme, milestone-based) with no take-or-pay recurring base — revenue is lumpy and deal-driven. The FY2025 step-change is not organic services growth: it is dominated by the DoveTree collaboration (a Gregory Verdine newco), whose $51M upfront landed inside Drug Discovery Solutions. Customer breadth is real: management claims coverage of 17 of the world's top-20 pharma, with named relationships across Pfizer (10-yr deal since 2016, helped on Paxlovid), Eli Lilly (three deals), Johnson & Johnson, Merck & Co..

Plain terms: XtalPi is a quant/robotics CRO with a platform-licensing option and a nascent royalty-bearing pipeline — the bull wants you to call it "the operating system for early drug discovery"; the tape calls it a services company that just signed one enormous headline deal.

Lens 2 · Supply Chain

Map upstream → XtalPi → end customer. (No supply-chain.md on disk — web-derived.)

• Compute (upstream chokepoint). Quantum-chemistry + AI screening is GPU/HPC-hungry. XtalPi historically leaned on cloud (a longstanding Tencent Cloud / hyperscaler relationship; Tencent is also a shareholder) and its own clusters. Post-2022 US export controls on advanced GPUs into China are a structural input risk for any China-based AI-compute business — the single most important non-obvious chokepoint here.
• Robotic lab hardware / reagents. Automated synthesis stations, liquid handlers, analytical instruments, and lab consumables — a mix of Western instrument makers and Chinese suppliers. The US-based robotic labs are a hedge that lets XtalPi keep serving Western pharma even as US-China biotech decoupling advances.
• Talent. MIT-trained founders + PhD chemists/ML engineers; ~780+ staff (2023). Talent is the real input moat and the real cost base.
• XtalPi (the node). Converts compute + data + robots into candidate molecules and datasets.
• Downstream customers. Big pharma (Lilly, J&J, Merck, Pfizer), emerging biotech newcos (DoveTree), materials/industrial buyers (GCL-Poly / JinkoSolar for perovskite & battery materials ), and incubated companies it seeds and part-owns (Signet Therapeutics, Ailux).

Single-source dependencies / chokepoints: (1) advanced compute under export control — the one that could genuinely throttle the platform; (2) customer concentration on a handful of mega-deals (DoveTree alone drove the FY2025 numbers); (3) key-person dependence on the three physicist-founders. Names or it didn't happen — the chain above is specific, but note the reagent/instrument tier is the thinnest-sourced part of this lens (web only).

Lens 3 · Competitive Advantages (moats)

Where the moat is plausibly real:

• Quantum-physics-first data generation. Most AI-drug peers (Recursion = phenomics/images; Insilico = generative chemistry; Schrödinger = physics-based simulation software) train on public or assay data. XtalPi's claim is a proprietary quantum-mechanical dataset flywheel feeding its own models — if true, a data moat competitors can't cheaply replicate. Schrödinger is the closest analog (physics-based), but Schrödinger sells software licences; XtalPi couples physics to wet-lab robotic validation in one loop — the "dry-lab + wet-lab in the loop" integration is the differentiated bit.
• Robotic lab-in-the-loop at scale. Closed-loop design→make→test→analyse compresses cycle time. This is capital- and know-how-intensive; a genuine barrier vs. software-only rivals.
• Customer-relationship moat / switching costs. A 10-year Pfizer relationship and three Lilly deals create reference-customer gravity and embedded workflows.
• Platform-licensing optionality (Ailux → Lilly): shifting from per-project fees toward licensing the engine is higher-margin and stickier if it takes.

Where the moat is weak / contested:

• Bargaining power is with the customer. Pharma buyers are giants; XtalPi is one of several AI-DD vendors. It needs them more than they need it — the DoveTree deal's headline value masks that XtalPi only pockets ~$100M near-term cash for years of work; the $5.89B is contingent royalties/milestones it does not control.
• The Derek Lowe critique. Veteran med-chemist Derek Lowe publicly doubts the model: "the hardest part of drug discovery is human trials" — the stage XtalPi explicitly avoids. AI can improve hit-finding; it has not yet demonstrably raised clinical success rates. Until an XtalPi-originated molecule clears a Phase II/III, the moat is "we make the easy part faster," which is commoditising.
• No approved drug from the platform yet. The proof that matters is unrealised.

Verdict on moat: narrow-and-improving, resting on the data-generation flywheel + wet-lab integration, not yet validated by a clinical outcome. Ground: positioning.md/bottlenecks.md absent → web-only.

Lens 4 · Segments

segments.csv empty → all ``.

• Geography: not cleanly broken out in the sources; customers span China + US/EU pharma. US robotic labs → material US-facing revenue, the exact split undisclosed (n/a).
• Read: Strip the ~RMB 365M DoveTree upfront and FY2025 revenue is RMB 438M ($61M). So the recurring business roughly doubled 2024→25 — impressive, but the reported +201% headline is inflated by a one-time upfront. The AI4S line is the segment I'd underwrite; Drug Discovery Solutions' quality depends on whether DoveTree/Lilly recur.

---

Phase B — Measure performance

Lens 5 · Earnings Result (FY2025, reported 2026-03-03)

(Operating lens; XtalPi reports semi-annually on HKEX, no quarterly consensus — so "vs. consensus" is not cleanly available; n/a — no sell-side consensus print sourced.)

• Revenue: RMB 802.6M, +201.2% YoY. Beat the company's own pre-announcement guide of "~193% jump".
• Gross profit / margin: FY2024 GP RMB 123.4M on 46.3% GM; FY2025 GM not cleanly disclosed in the PR — n/a for the exact FY2025 %, but mix-shift toward high-margin discovery + upfront suggests GM rose.
• Profit: first-ever full-year NET PROFIT of RMB 134.6M (adjusted net profit RMB 258.2M), vs. FY2024 net loss RMB 1,514.9M. Critical caveat: the swing is flattered by (a) the DoveTree upfront, and (b) valuation gains — Globe & Mail explicitly frames it as "Revenue Surge and Valuation Gains". The FY2024 loss was itself ~58% non-cash: RMB 875.4M was a fair-value loss on convertible redeemable preferred shares (CRPS) that vanished when the CRPS converted to ordinary shares at IPO; adjusted FY2024 loss was RMB 456.8M. So the true operating trajectory is: adj. loss RMB ~457M (2024) → adj. profit RMB 258M (2025) — a real improvement, but ~$36M of "adjusted profit" against a $4.5B market cap.
• Balance sheet: cash RMB 7,068.6M (~$0.99B) at 2025-YE, plus RMB 2,536.8M (~$0.35B) raised via convertible bonds in 2026 → RMB 9.6B ($1.34B) war chest. This is the single strongest fact in the file — XtalPi is not capital-constrained; runway is effectively indefinite at current burn.
• Clients: revenue-generating clients +62% YoY; 200+ industry-specific AI models built.
• Market reaction: the stock is a momentum vehicle — 52-week range HK$5.27–15.12, i.e. it nearly tripled off the lows intraday then round-tripped; at HK$8.19 (2026-07-06) it sits mid-range. The market is trading the deals, not the P&L.
• Unusual vs. own history: the entire profit is discontinuous with 8 prior years of deepening losses — treat FY2025 as an event, not a run-rate, until FY2026 shows the recurring base holds without a mega-upfront.

Lens 6 · Earnings Calls / Management Messaging (sentiment trend)

No transcripts/ on disk; HKEX names hold results briefings, not US-style scripted calls, and clean transcripts don't scrape — so this is management-messaging analysis from results PRs + interviews ``.

• Tone arc 2024 → 2026: shifted from "pre-revenue specialist-tech, trust the platform" (IPO, mid-2024) → "revenue inflecting, path to profit" (2025 interim) → "first profitable AI4S H-share company, flywheel compounding" (FY2025, 2026-03). Confidence is rising and is now backed by cash and a marquee deal, not just narrative.
• Recurring phrases: "AI + robotics," "lab-in-the-loop," "flywheel," "de novo discovery across small molecules / biologics / ADC / molecular glue," "17 of top-20 pharma." The flywheel metaphor is doing heavy lifting.
• What they started saying: platform-licensing (Ailux→Lilly), materials/AI4S diversification (JinkoSolar tandem-solar, Jan 2026) — deliberately broadening beyond pharma-services cyclicality.
• What to watch them stop saying: if "DoveTree progressing smoothly" quietly disappears from future updates, that's the tell the flagship deal is stalling.

Lens 7 · Comps

Read: XtalPi is valued at ~7x Schrödinger's EV/Sales multiple on a smaller, lumpier, less-proven revenue base and no clinical validation, while Schrödinger and Insilico have their own clinical-stage assets. The comp table is the single clearest bear exhibit in the dossier: XtalPi's market cap is larger than Schrödinger + Certara combined, on a fraction of their recurring revenue quality. The premium is a China-AI-momentum + DoveTree-headline premium, not a fundamentals premium.

Lens 8 · Stock-Price Catalysts (moves >5%)

Short public history (IPO'd Jun 2024) — ``:

• 2024-06-13 IPO debut: +9.8% to HK$5.80 (first 18C listing; HK$989.3M / $126.7M raised; ~$2.5B open-day cap).
• 2024-08 GCL-Poly $135M+ materials contract — validated the AI4S/materials leg.
• 2025-06/08 DoveTree term-sheet → definitive $5.99B deal: shares surged (+8% to +23% intraday on the news days, closing +12.4% at HK$7.42 on one). This is the dominant catalyst in the stock's life.
• 2025-11-05 Ailux–Lilly $345M bispecific-antibody deal — platform-licensing proof point.
• 2026-03-03 FY2025 first-profit print (+201% rev) — confirmed the turn; fed the run toward the HK$15.12 high.
• 2026-H1 momentum + pullback: +115% over six months at one point, then round-tripped to HK$8.19 — classic deal-froth mean-reversion.

Pattern: the market reacts almost exclusively to big-name licensing deals (DoveTree, Lilly) and the China-biotech-boom narrative, far more than to steady-state fundamentals. That makes it a catalyst-and-sentiment stock — high beta to deal-flow headlines and to the broader "Western pharma's billion-dollar bet on Chinese biotech" trade.

---

Phase C — Judge people & books

Lens 9 · Management

• Track record: three MIT-postdoc quantum physicists — Wen Shuhao (Chairman), Ma Jian (CEO), Lai Lipeng (CIO) — who built XtalPi from a 2015 Pfizer blind-test win (100% crystal-structure prediction accuracy → 10-yr Pfizer deal) into the first Chapter-18C HKEX listing and, in 2025, the largest reported AI-DD licensing deal in history. That is a serious build. Founder-CEO Ma Jian doubles as a Zhejiang University professor — deeply technical, science-forward.
• Tenure & skin in the game: founders still Chairman/CEO/CIO 10 years in — founder-led, high alignment. Exact post-IPO insider % not cleanly sourced (insider-transactions.csv absent) — but founders + early VCs (Tencent, HongShan/ex-Sequoia China, SoftBank Vision Fund 2, Google, Susquehanna, 5Y Capital, ZhenFund, Sino Biopharm) retain large stakes. The company grants ongoing options/RSUs to the exec team — watch dilution/SBC.
• Capital allocation: raised ~$127M at IPO; then a RMB 2.54B convertible in 2026 into strength — sensible (locks in a war chest while the stock is hot). Uses cash to incubate and part-own spinouts (Signet Therapeutics, Ailux) — a venture-studio model that converts services IP into equity upside. No buybacks/dividend (correct at this stage). ROE/ROIC negative historically; FY2025 positive but deal-flattered.
• Red flags (governance): (1) Deal counterparty quality — the flagship DoveTree is a Verdine-founded newco (LLC), and Verdine is being appointed an XtalPi advisor guiding pipeline selection — an entangled, non-fully-arm's-length flavour to the biggest deal, though disclosed. (2) Related/incubated-party revenue — booking revenue from companies it seeds (Signet et al.) warrants scrutiny of intercompany terms. (3) Serial renaming of the automation segment (automation→robotics→AI4S) can obscure trend comparability. None are smoking guns; all are watch-items.
• Archetype: founder-scientist visionaries — pattern implies aggressive platform expansion and narrative-forward communication (the "flywheel"/"AI+robotics+quantum" framing). Great for optionality, requires skepticism on headline framing.

Lens 10 · Forensic Red Flags

Forensic lens. All `` — no filings on disk; figures from results announcements.

• Revenue recognition (the #1 flag). FY2025's +201% and first profit are materially dependent on the DoveTree $51M upfront and, per Globe & Mail, valuation/fair-value gains. Upfronts from multi-year collaborations can be recognised in ways that front-load optics. Question: how much FY2025 revenue is ratable/deferred vs. point-in-time, and how much profit is non-cash fair-value? The PR doesn't say — a real diligence gap.
• "Bio-bucks" overstatement risk. The $5.99B DoveTree and $345M Lilly figures are contractual maxima (upfront + near-term + contingent milestones + royalties). Actual guaranteed cash is ~$100M (DoveTree) / double-digit millions (Lilly). Headlines conflate the two; a careful analyst must not capitalise the maxima. XtalPi has, to its credit, disclosed the $51M + $49M split and received the 2nd DoveTree payment with the first oncology asset advancing to IND-enabling studies — partial validation the deal is live, not vapor.
• Fair-value volatility. Pre-IPO, RMB 875.4M of the FY2024 loss was non-cash CRPS revaluation; that specific line is gone post-conversion, but the pattern shows earnings can swing hard on non-operating fair-value items (now potentially including marks on incubated-company equity stakes).
• Cash-flow vs. earnings. With a fresh convertible and a big upfront, reported profit ≠ operating cash generation. Underlying operations are likely still cash-consumptive ex-deal; the RMB 9.6B cash pile is largely financed, not earned. Operating cash flow — not disclosed in sourced PRs (n/a).
• SBC / dilution. Ongoing option/RSU grants + convertible bonds = future dilution; 4.30B shares already. Watch adjusted-vs-GAAP gap (adj. profit RMB 258M vs GAAP RMB 135M implies ~RMB 123M of add-backs, plausibly SBC + fair-value).

Regulatory findings (required sub-section):

• SEC (EDGAR): None possible — XtalPi has no CIK and is not an SEC filer; regulatory/regulatory-findings.md confirms 0 LR/AAER hits.
• Non-SEC web sweep ("XtalPi" (FTC/DOJ/FDA/consent decree/settlement/fine/penalty)): no material enforcement actions surfaced as of 2026-07-06.
• Geopolitical/regulatory overhang (material): the BIOSECURE Act became US law 2025-12-18 (in the FY2026 NDAA), restricting federal dealings with designated Chinese "biotech companies of concern." XtalPi is NOT currently named (the named CROs are WuXi AppTec/Biologics + BGI/MGI/Complete Genomics); the DoD 1260H "Chinese military companies" list is the trigger, and an OMB "companies of concern" list is due by 2026-12-18. XtalPi is a China-HQ AI-biotech serving US pharma with US robotic labs → squarely in the blast radius if the list expands. This is the dominant tail risk (see Lens 13).
• Item 3 / legal proceedings: no 10-K exists; no material litigation surfaced via web ``.
• Summary: No material accounting-fraud or enforcement findings to date (verified via: SEC EDGAR — n/a non-filer; web enforcement search; no filings). The forensic concerns are quality-of-earnings (deal-dependent revenue, non-cash gains, bio-bucks framing) and a regulatory tail (BIOSECURE), not evidence of misconduct.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection (rNPV + runway lens — +clinical overlay)

XtalPi is a hybrid: a real (small) services P&L + a pipeline/royalty optionality bucket. So project both the services engine and treat the pipeline as rNPV optionality. Fiscal year = calendar year. No forecast.ts create in the unattended watchlist loop (per skill).

Services/operating base (bottom-up, ``):

• FY2025 recurring revenue ex-DoveTree-upfront ≈ RMB 438M (~$61M) [est]. AI4S growing ~40–60%, discovery (ex one-offs) ~30–50%.
• FY2026 base: RMB ~1.05–1.2B revenue only if DoveTree/Lilly deliver additional near-term payments ($49M DoveTree near-term + Lilly milestones) [est]; RMB ~650–750M on recurring alone [est]. Wide cone — deal timing dominates.
• FY2027–28 base: RMB ~1.4–2.0B if platform-licensing (Ailux model) compounds and 2–3 more mega-deals land; margins improving on mix + scale [est].
• EPS: with 4.30B shares and adj. profit that is deal-timing-dependent, a clean 3-year EPS path is n/a — not credibly sourceable; GAAP profit could oscillate around breakeven ex-deals. This is why I refuse a point EPS forecast here.

Pipeline optionality (rNPV, ``, illustrative):

• XtalPi carries no wholly-owned clinical asset it controls; upside is royalties/milestones on partner/incubated programs (DoveTree oncology asset → IND-enabling; Signet's SIGX1094 → Phase II Q3'26 ). A single DoveTree program reaching market at, say, $1B peak sales × mid-single-digit royalty × ~10% PoS × discount → low-hundreds-of-$M rNPV per program; the $5.89B royalty maximum is a portfolio maximum at 100% success, not an expected value.
• Runway-to-catalyst (the question that matters): ~$1.34B cash vs. modest operating burn → runway is effectively unconstrained through multiple value-inflection catalysts (Signet Phase II readout Q3'26; DoveTree IND filings; more licensing deals). XtalPi will not be forced to raise into weakness — a genuine strength.

Base call (not logged as a tracked forecast in this unattended run): recurring revenue roughly doubles FY24→FY26 organically; reported profitability remains deal-dependent and lumpy; no controlled clinical asset de-risks before 2027. Falsifiable trigger to log later: "2228.HK FY2026 revenue ex-one-time-upfronts ≥ RMB 900M."

Lens 12 · Bull vs Bear

Bull case. XtalPi is the category-leading, best-capitalised AI+robotics+quantum drug-discovery platform, now cash-flow-validated by big pharma: 17 of top-20 pharma as customers, three Lilly deals, a $5.99B DoveTree flagship with real cash flowing and assets advancing, and a $1.3B war chest to fund a decade of platform expansion and equity incubation. The AI4S/materials leg (perovskite, battery, TCM — JinkoSolar, GCL-Poly) diversifies beyond pharma cyclicality into China's energy-transition supply chain. The pivot to platform-licensing (Ailux→Lilly) converts a lumpy services model into recurring, high-margin, royalty-bearing revenue. If one incubated/partnered molecule clears clinical proof-of-concept, the market re-rates the entire platform from "CRO" to "royalty compounder," and the current ~31x EV/Sales looks cheap in hindsight. China-biotech is in a structural bull market as Western pharma out-licenses aggressively from Chinese labs.

Bear case. Three things can permanently impair the thesis: (1) Valuation. At ~$4.5B cap / ~31x EV/Sales / ~57x on recurring revenue, XtalPi is priced as a proven royalty platform while it is still a sub-$120M-revenue services company with zero clinical validation — a 7x premium to Schrödinger, which has better revenue quality and its own pipeline. Multiple compression alone is a 50%+ downside risk with no operational failure required. (2) The proof never comes. Per Derek Lowe, AI hasn't raised clinical success rates; XtalPi avoids trials, so its value hinges on partners succeeding at the hard part — outside its control. If DoveTree's assets fail in the clinic (base-rate: most do), the $5.89B evaporates and the "flywheel" is just a faster hit-finder in a crowded field. (3) BIOSECURE / decoupling. A China-HQ biotech dependent on US pharma revenue and export-controlled compute could be cut off if the OMB "companies of concern" list (due Dec 2026) expands, or if US pharma pre-emptively de-risks its China exposure. Pre-mortem (18 months out): the stock is down 60% because (a) FY2026 lapped the DoveTree upfront and revenue fell YoY, exposing the lack of a recurring base; (b) a BIOSECURE designation or Lilly/pharma pullback spooked the China-biotech trade; (c) the "first profit" reversed once the one-time gains rolled off. Contrarian view the market is refusing to see: the FY2025 "profit" is a deal-and-mark artifact, not a durable earnings base — and the stock is being valued as though the $5.99B is bankable.

Lens 13 · Devil's Advocate (short-seller)

Dismantling the bull case.

• Revenue concentration is extreme. FY2025's optics rest on one deal (DoveTree ~$51M upfront). Strip it → ~$61M recurring, growing but ordinary. What happens FY2026 when that upfront doesn't repeat? Very plausibly a YoY revenue decline — devastating for a momentum name priced for hypergrowth.
• The flagship deal's counterparty is a newco, not Merck. DoveTree is a Verdine LLC; the $5.89B is contingent royalties on drugs that don't exist yet, and Verdine is being made an XtalPi advisor — an entangled arrangement that a short would frame as "headline manufacturing." The real cash is ~$100M for years of work.
• The moat is weaker than bulls think. XtalPi does the commoditising half (hit-finding); Schrödinger, Recursion (post-Exscientia merger), Insilico (which has a Phase-II asset — ahead of XtalPi), and NVIDIA-backed platforms all crowd the same space. Barriers to "AI screens molecules" are falling, not rising.
• Most dangerous competitor bulls underestimate: Insilico Medicine (2546.HK) — same China-AI-biotech trade, but further along clinically (ISM001-055 Phase IIa positive in IPF). If Insilico gets a drug approved first, it — not XtalPi — becomes the category's proof point.
• Worst capital-allocation / governance optics: revenue from self-incubated companies + an advisor-conflicted flagship deal + serial segment renaming + fair-value-gain-flattered earnings = a quality-of-earnings profile a short can attack even without alleging fraud.
• What must hold for today's price: (a) DoveTree/Lilly milestones keep landing on schedule; (b) recurring services keep ~doubling; (c) no BIOSECURE designation; (d) a clinical proof-point emerges within ~2 years. If growth disappoints 20–30% (i.e. FY2026 flat-to-down as the upfront laps), the ~31x EV/Sales likely halves toward the peer 10–15x, implying ~40–55% downside.
• Single scenario that permanently impairs: a BIOSECURE "company of concern" designation (or a US-pharma pre-emptive pullback) severs XtalPi from Western pharma revenue and advanced compute simultaneously — plausibility low-but-rising given the Dec-2026 OMB list and the explicit "China biotech" targeting of the Act.

Lens 14 · Management Questions (ordered by information value)

1. Of FY2025 revenue, how much was point-in-time upfront (DoveTree/Lilly) vs. ratable/recurring, and what is FY2026 revenue excluding all one-time upfronts? (The whole thesis hinges on the recurring base.)
2. How much of the FY2025 net profit was non-cash fair-value gains (incl. marks on incubated-company equity), vs. cash operating profit?
3. Under DoveTree, what are the guaranteed cash payments across the next 24 months, independent of milestones — and what are the actual per-milestone amounts behind the "$5.89B"?
4. What is operating cash flow ex-financing, and when does the core services business become self-funding without upfronts or raises?
5. Has XtalPi received any indication it could be added to a DoD 1260H / BIOSECURE "company of concern" list, and what is the contingency if US pharma must offboard it?
6. What share of revenue and compute depends on US-facing customers and export-controlled GPUs, and how resilient is the platform to tighter controls?
7. Governance of the DoveTree relationship: given Prof. Verdine is both counterparty-founder and XtalPi advisor, how are conflicts and pricing handled?
8. For incubated companies (Signet, Ailux, others): what are the intercompany terms, and how much reported revenue is intragroup/related?
9. Has any XtalPi-originated molecule reached the clinic under XtalPi's own name, and what is the earliest human proof-of-concept the platform can point to?
10. What is the realistic conversion rate from "discovery collaboration" to milestone-triggering development stages, based on the portfolio so far?
11. On platform-licensing (Ailux→Lilly): is this a repeatable, scalable revenue line, and what does the licensing pipeline look like?
12. What is the expected share-count dilution (options/RSUs + the 2026 convertible) over three years?
13. How defensible is the quantum-dataset moat as rivals scale their own physics/AI stacks — where specifically is XtalPi's data non-replicable?
14. What is the AI4S/materials segment's addressable market and margin profile, and can it become the stable ballast to lumpy pharma deals?
15. What are the capital-allocation priorities for the ~$1.3B cash — organic build, M&A, more equity incubation, or buffer — and what ROIC hurdle governs it?

---