Schrodinger

Phase A — Understand the business

Lens 1 · Company Overview

Schrödinger runs a dual business model that has defined the "techbio" archetype: sell the computational drug-discovery platform as software, and use that same platform to discover proprietary drugs in-house and via collaborations.

• Two reported segments — Software and Drug Discovery. In FY2025 the split was Software $199.5M (78%) / Drug Discovery $56.4M (22%) of $255.9M total revenue. Conflict flag: the KB entity page describes the model as "~60% software / ~40% pipeline" — that framing is stale; on FY2025 revenue it is closer to 78/22. The 60/40 likely referred to strategic emphasis/spend, not revenue mix.
• Software product structure: FY2025 software revenue decomposes into on-premise $101.4M + hosted $45.1M (plus professional services/other). The company is mid-transition from perpetual/on-premise toward hosted (SaaS), targeting ~75% hosted mix over a multi-year horizon.
• Customers: the world's leading pharma and biotech R&D organisations; the Schrödinger Suite (FEP+, Glide, WaterMap, LigPrep, Maestro) is the gold standard for structure-based drug design. Customer concentration is real: in FY2025 one customer accounted for 17% of total revenues (10% in FY2024); in FY2023 two customers were 26% + 11%. International = ~40% of revenue.
• Anchor collaboration — Novartis. Multi-year deal to collaborate on small-molecule discovery across specified therapeutic areas, plus an expanded enterprise software license that "substantially increased Novartis' access to our computational predictive modeling technology and enterprise informatics platform … to industry-leading scale". Economics: up to ~$2.272B in total milestones across the initial programs (none recognised as revenue yet) plus tiered royalties mid-single-digit to low-double-digit on net sales.
• Contract structure: software is annual licensing (ACV-driven, ratable for hosted, upfront-weighted for on-premise → heavy Q4 seasonality). Drug-discovery revenue is milestone-lumpy (upfront + discovery/development/commercial milestones + royalties) and therefore inherently volatile quarter to quarter.
• Hidden third leg — the equity-investment portfolio. Schrödinger co-founds biotechs that use its platform (Nimbus most famously) and holds equity. This shows up as "Equity investments" on the balance sheet ($39.8M at Q1 2026, down from $73.6M at YE2025 as positions are monetised) and as large, non-operating "change in fair value of equity investments" swings in other income.

850 full-time employees. HQ New York, NY; Delaware incorporated; CIK 0001490978; NASDAQ: SDGR.

Lens 2 · Supply Chain

For a software + drug-design company the "supply chain" is compute, IP inputs, and (for the pipeline) CDMO/CRO capacity:

• Upstream inputs:
• Foundational IP — core technology licensed from Columbia University; loss of those rights is flagged as a material risk. This is a genuine single-source dependency at the root of the platform.
• Compute — physics-based FEP+ calculations are GPU-intensive; the hosted product runs on cloud infrastructure (the cost-of-revenue step-up in software, see Lens 5, is partly cloud hosting). Hyperscaler dependence is a cost-line, not a chokepoint.
• Scientific talent — PhD-level computational chemists and ML researchers; "ability to retain key executives and attract qualified personnel" is named as a top risk.
• The company itself: software dev + an internal therapeutics group running discovery → IND-enabling → early clinical.
• Downstream / end customers:
• Software buyers — top-10 pharma + biotech + (increasingly) materials-science and electronics firms. Named anchor: Novartis (enterprise-scale license).
• Pipeline partners — Schrödinger now explicitly intends to out-license/partner mid-and-late-stage development of its clinical assets rather than fund Phase 2/3 itself. So the "end customer" for the pipeline is a pharma acquirer/partner, not a patient — Nimbus→Takeda is the template.
• For clinical execution: reliant on third-party CROs to run trials — explicitly flagged ("we rely on, and plan to continue to rely on, third parties to conduct our clinical trials"). No owned manufacturing; CDMO-dependent for drug substance.
• Chokepoints: (1) Columbia IP license at the root; (2) key-person/talent concentration; (3) a single ~17% customer; (4) for the pipeline, dependence on finding a willing partner (a demand-side chokepoint, see Lens 13).

Lens 3 · Competitive Advantages (moats)

The moat is real and unusually durable for a computational tool — but it is narrowing in scope, not depth.

• Physics, not just ML — the core differentiator. FEP+ combines quantum/molecular mechanics with ML to compute binding free energy to ~1 kcal/mol, approaching experimental accuracy across broad chemical space. Crucially, the open question the bears raise — "can open-weight protein models replace this?" — is answered against them on the precise axis that matters: peer-reviewed work shows AlphaFold2-predicted structures are often not accurate enough as inputs for FEP-grade free-energy calculations. AlphaFold predicts structure; FEP+ predicts affinity ranking of candidate molecules — complementary, not substitutable. This is the crux of the bull case for platform durability.
• Regulatory validation as a moat — FEP+ predictions are accepted by the FDA as supporting evidence in IND/NDA filings when validated. A regulator-blessed computational method is extraordinarily sticky.
• Switching costs / workflow lock-in — the Suite (Maestro front-end + FEP+/Glide/WaterMap) is embedded in medicinal-chemistry workflows at most top-10 pharma; multi-year enterprise agreements (Novartis "industry-leading scale") raise the exit cost.
• Data + process moat — two decades of physics-based methods development, validated against proprietary experimental datasets; hard to replicate by a startup retraining an ML model.
• Bargaining power: asymmetric. Over small biotech customers Schrödinger has strong power (the Suite is a near-necessity). Over Novartis-scale customers, less so — a 17%-of-revenue customer has leverage on renewal pricing.
• Where the moat is thinner: (1) ML-first players (Recursion/Exscientia, Isomorphic, Insilico) are racing on hit generation and phenotypic discovery where physics is less dominant; (2) the platform's value is in precision late-stage optimisation, a smaller TAM slice than "all of AI drug discovery"; (3) materials-science extension is real but early/small.

Lens 4 · Segments

segments.csv is empty — built from the filings' statements of operations.

By segment (FY2025 vs FY2024 vs FY2023):

• Software is the steady compounder: $159M→$180M→$199.5M, low-double-digit growth. But growth has decelerated from the 38% of 2020 (the original de-rating trigger, Lens 8) to ~11%. Within software, hosted is the fast-grower — Q1 2026 hosted was 34% of software revenue vs 24% a year earlier — but it depresses reported revenue and margin because it's recognised ratably (see Lens 5).
• Drug discovery revenue is lumpy, not a trend — it halved in 2024 ($57.5M→$27.2M) then doubled in 2025 ($56.4M), driven by milestone timing and accelerated recognition of deferred collaboration revenue, not a product ramp.
• By geography: international ~40% of revenue FY2025. The filing gives concentration but not a full segment-by-geo operating-income bridge; that granularity is n/a — not disclosed at the segment-profit level.
• Segment profitability: the company reports consolidated operating loss only; it does not publish clean segment operating income, but cost-of-revenue is split — FY2025 software COGS implies software gross margin ~49% blended on-prem+hosted vs drug-discovery running at a gross loss (drug-discovery COGS $62.3M > drug-discovery revenue $56.4M). The pipeline is a cost centre that occasionally prints a milestone.

---

Phase B — Measure performance

Lens 5 · Earnings Result (latest print — Q1 2026, reported 2026-05-05)

• Revenue beat headline, but the mix spooked the market. Total $58.6M beat consensus, but it was carried by lumpy drug-discovery (+124%) while software fell 21% YoY — the line investors actually pay for. Shares fell ~6% after-hours on the print.
• The software optics problem is structural, not a stumble. Software gross margin fell to 69% from 80%, explicitly because of the planned acceleration to hosted licensing — ratable recognition pushes revenue and margin out in time. The leading indicator, ACV, grew +12% to $28.4M — i.e. underlying demand is fine; the reported software line is being whipsawed by the accounting transition.
• Guidance reaffirmed: FY2026 ACV $218–228M (+10–15%), drug-discovery revenue $55–65M, opex below 2025 levels, multi-year path to ~75% hosted mix, modest remaining clinical R&D as trials conclude.
• Balance-sheet flags (mostly positive):
• Cash + marketable securities ≈ $398.9M at Q1 2026 (cash $260.3M + restricted $7.5M + securities $138.7M).
• Zero financial debt — no borrowings; liabilities are deferred revenue + leases.
• Deferred revenue $162.1M ($103.1M current + $59.0M long-term) — a real backlog cushion.
• Q1 operating cash flow −$14.8M (seasonal — Q4 is the collection quarter); the big AR swing (−$55.8M source) is the prior-year Q4 software billings being collected.
• Unusual vs its own history: the −21% software decline is the standout — but it is a comp/accounting artifact of the hosted shift and a tough prior-year quarter, not lost customers (ACV +12% proves it). The risk is that the market keeps reading the reported line literally.

Lens 6 · Earnings Calls (sentiment trend)

transcripts/ is empty in the research layer; sentiment built from web transcripts/summaries across the last ~3 calls.

• Q3 2025 (reported Nov 2025): beat forecasts, stock dipped. Tone: platform-validation confident, but the market focused on the loss. Strategic-priorities update foreshadowed the pipeline pivot.
• Q4 2025 / FY guidance (early 2026): "Provides Update on Progress … Outlines 2026 Strategic Priorities" — the explicit reframing toward software-led growth + partnering the pipeline.
• Q1 2026 (May 2026): CEO Ramy Farid "off to a strong start"; the narrative pivoted hard to ACV (the metric management wants you to watch) and away from reported revenue, plus framing the hosted transition as "prioritising long-term ACV expansion and customer lifetime value over near-term margin optics".
• Sentiment shift over time: from "look at our pipeline AND platform" (2024) → "the platform is the business; the pipeline gets partnered" (2025–26). New recurring phrases: "ACV," "hosted transition," "customer lifetime value," "strategic partnership," "operating expenses below prior year." Things they stopped saying: ambitions to independently advance a broad clinical pipeline; the multi-program internal-pharma story has been quietly retired. The tone is disciplined-defensive: cost-out, focus, prove-the-software — a management team that has heard the market and is repositioning rather than doubling down.

Lens 7 · Comps

Operating + clinical hybrid — this name straddles two comp sets. Multiples are `` with source/date or marked not-sourced. Never fabricated.

Software/platform comp set:

Techbio / AI-drug-discovery comp set (by model, not P/E — all loss-making):

• Read: SDGR at ~2.5x EV/sales is being valued as a slow-growth software company, not as "AI drug discovery." Its own P/S has compressed from a 5-yr average of ~17.7x to ~3.6x (and EV/sales 2.5x) — the de-rate is essentially complete. It trades roughly in line with profitable Certara (2.3x) despite faster software growth and an optionality-rich pipeline + equity portfolio that Certara lacks.
• P/E: n/a — loss-making. 5-yr average ROE: n/a — negative equity returns (cumulative losses); not a meaningful metric for this name. Dividend yield: 0% — none.
• Caveat: market-cap figures diverge across sources ($932M stockanalysis vs $1.18B Robinhood/Morningstar) because of the recent price move and share-count timing; EV/sales math is the more stable anchor.

Lens 8 · Stock-Price Catalysts (what actually moves SDGR)

• The defining drawdown — early 2021: after 38% software growth in 2020, management guided 10–18% software growth for 2021 → stock dropped 27% in a week, ~50% from earlier-2021, ~70% from the high. The single most important pattern: this stock lives and dies on the software growth rate, specifically the forward guide. Decelerating software guidance is the kill-shot.
• 2023 — Nimbus/Takeda windfall: $111.3M (then total ~$147.3M) cash distribution from the sale of Nimbus's TYK2 inhibitor to Takeda. Proof the equity-portfolio optionality is real money, and a positive catalyst class the market under-weights.
• Recurring >5% movers: (1) quarterly software/ACV guidance revisions (biggest); (2) clinical readouts (SGR-1505 ASH data; SGR-3515 data expected 1H/Q2 2026); (3) collaboration announcements (Novartis expansion); (4) equity-portfolio monetisations (Nimbus-type events); (5) sector-wide techbio de-rating beta (RXRX P/S −46% in 7 months dragged the whole cohort).
• Q1 2026: revenue beat but software-mix/loss miss → −6%. Consistent with the pattern — the tape rewards software quality, not total-revenue beats driven by lumpy milestones.

---

Phase C — Judge people & books

Lens 9 · Management

• CEO Ramy Farid — founder-archetype operator-scientist. President since 2008, CEO since 2017, director since 2012. PhD Caltech (chemistry); authored the induced-fit docking algorithm (most-cited J. Med. Chem. article of 2006). This is a builder of the actual science, not a hired operator — a genuine asset for a physics-platform company.
• Skin in the game — the soft spot. Farid owns 330,824 shares (~$4M) as of Mar 2026, with 0 buys / 3 sells over 5 years. For a "founder," that is modest direct ownership and a sell-only insider record — not the aligned, buying-the-dip founder you'd want at a 70%-off stock. Base salary $730k; 2024 cash bonus paid at 115%; "compensation actually paid" $0.67M in 2024 (equity-linked, so depressed by the stock). Ownership guideline = 3x base, which he meets — but 3x base is a low bar.
• CFO turnover — watch item. Geoffrey Porges departed (mutually) May 2025, formalised Sept 2025; replaced by Richie Jain, internal SVP of strategic finance + corporate/business development. Signal: elevating a BD/strategic-finance person to CFO is squarely consistent with the pivot to partnering the pipeline and monetising the equity portfolio — the CFO's job is now deal-making, not just reporting. A reasonable, even shrewd, fit for the new strategy — but C-suite turnover during a strategic pivot is execution risk.
• Capital-allocation history — mixed-to-good:
• Good: co-founding Nimbus and harvesting $147M; cutting opex (R&D $201.8M→$173.1M FY24→FY25; total opex $341.4M→$309.5M); no debt, ~$399M cash; the discipline to stop funding a money-losing internal pipeline.
• Bad/unproven: years of heavy R&D spend on a clinical pipeline now being given away to partners — i.e. shareholders funded the discovery, partners may capture much of the upside; cumulative operating losses with negative ROIC throughout.
• Red flags: sell-only insider record; equity-portfolio fair-value swings flatter "other income" and muddy the underlying loss; related-party amounts appear in revenue/expense lines (small). No promotional/hype behaviour — management is, if anything, under-promoting (leaning on ACV rather than the AI narrative).

Lens 10 · Forensic Red Flags

• Revenue recognition is the audit focus, and it's handled cleanly. KPMG (auditor since 2010) issued a clean opinion + effective ICFR, and flagged exactly one Critical Audit Matter: identification of performance obligations in material software arrangements. This is the right thing to scrutinise (multi-element software + collaboration deals) and there's no qualification — a positive.
• Where cash flow diverges from earnings — and why it matters: FY2025 printed +$13.9M operating cash flow vs a $(103.3M) net loss. Do not be fooled — this is not a profitability inflection. The +$13.9M is almost entirely a $152.7M accounts-receivable collection of a huge 2024 software deal (the offsetting FY2024 lines: AR build −$169.7M, deferred-revenue build +$155.5M). Strip working capital and the business still burns cash at roughly the operating-loss rate. The Q1 2026 −$14.8M operating outflow confirms the underlying burn.
• Non-GAAP / SBC: stock-based comp is $43.0M FY2025 (down from $49.9M FY24) and $9.1M in Q1 2026 — material (~17% of FY25 revenue) but trending down with headcount discipline, and not used to manufacture a flattering non-GAAP profit (the company still reports GAAP losses). Lower risk than typical for the cohort.
• Equity-investment fair-value volatility: "change in fair value of equity investments" swung +$48.2M (FY2025), +$5.7M (FY2024), +$53.5M (FY2023) and −$13.5M in Q1 2026. This injects large non-operating noise into reported results both ways — analytically, strip it out; it is real value but lumpy and non-recurring.
• Goodwill/intangibles: negligible ($4.8M goodwill) — no impairment overhang.
• Receivables vs revenue: AR is highly seasonal (Q4-heavy billing) — the swings are timing, not channel-stuffing; deferred revenue backlog ($162M) supports the recognised numbers.

Regulatory findings (required):

• SEC: No Litigation Releases and no AAERs naming Schrödinger in the search period (2021-06-18 → 2026-06-18), verified via SEC EDGAR EFTS (LR + AAER).
• 10-K Item 3 (Legal Proceedings): the company's risk disclosures discuss litigation/IP risk generically; no material pending legal proceeding is highlighted as outcome-determinative in the filing's risk summary. (No specific material case surfaced.)
• Non-SEC enforcement (web): no material FTC/DOJ/FDA/CFPB enforcement actions, consent decrees, fines or penalties against Schrödinger surfaced in search. The only FDA-related disclosure is inbound risk — the FDA/HHS ~20,000-person reduction-in-force (and ~3,500 FDA cuts) as a risk to review timelines for SDGR's product candidates, not an action against the company.
• Conclusion: No material regulatory or legal findings — verified via SEC EDGAR EFTS (LR, AAER), web search, and 10-K Item 3 as of 2026-06-18. Clean.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection (rNPV + EPS path — the hybrid view)

This is a software-EPS story with a pipeline-optionality call layered on. I build the software/operating EPS path bottom-up and treat the pipeline as rNPV optionality, not base-case EPS.

Operating model — bottom-up from guidance (FY ends Dec; project FY2026E / FY2027E / FY2028E):

• Revenue: software grows with ACV (+10–15% guide → use ~12%); drug-discovery held at guide midpoint $60M FY2026 then flat-to-modestly-up (milestone-dependent, conservatively flat). Note: reported FY2026 revenue may optically dip toward the ~$234M consensus because the hosted shift defers software revenue even as ACV grows — a key reason the Street trimmed numbers.
• FY2026E total ≈ $245M; FY2027E ≈ $275M; FY2028E ≈ $310M.
• Gross margin: software GM troughs ~69–72% during the transition, recovers toward high-70s as the hosted base compounds; drug-discovery stays gross-loss-to-breakeven.
• Opex: guided below FY2025's $309.5M; assume R&D continues down as clinical trials conclude → total opex ~$295M FY26, ~$300M FY27 (re-investment), ~$310M FY28.
• Operating loss path: FY2026E ≈ $(95M) narrowing to ≈ $(70M) FY2028E. EPS (GAAP, ex equity-FV noise): FY2026E ≈ $(1.10), FY2027E ≈ $(0.95), FY2028E ≈ $(0.75) on ~75M shares. The company does not reach GAAP operating breakeven within the 3-year window on the software business alone — breakeven needs either software reacceleration above ~15% sustained or a step-change milestone/partner payment.
• Pipeline rNPV (optionality, not in EPS): lead assets SGR-1505 (MALT1; 100% ORR in Waldenström, Fast Track + Orphan) and SGR-3515 (Wee1/Myt1; 65% disease-control). Both being partnered, so SDGR captures upfront + milestones + royalty, not full asset value. Rough optionality: n/a — peak-sales and deal terms not sourced; conceptually each partnered oncology asset could be worth $50–150M+ in upfront/near-term milestones on a Nimbus-style outcome. The ~$2.272B Novartis milestone ceiling is gross/unrisked and mostly years out.
• Runway: ~$399M cash, no debt against ~$80–100M/yr underlying burn → 4+ years of runway, comfortably past the SGR-3515 readout and multiple partnering windows. Runway is not the binding constraint — unusual and favourable for a clinical-stage-adjacent name.

Brier forecast: not logged (--watchlist unattended; forecast.ts create is skipped per skill rules). Candidate for a later manual log: "SDGR FY2026 ACV ≥ $223M (guide midpoint), p≈0.6, resolves 2027-02" and "SGR-3515 Phase 1 initial data reported by 2026-09-30, p≈0.8."

Lens 12 · Bull vs Bear

Bull case. Schrödinger is the one techbio with a proven, regulator-validated, physics-based moat that ML-first rivals cannot replicate on the precision-affinity axis. Software is a real SaaS compounder (ACV +10–15%) finally being valued like software (2.5x EV/sales) after a 3-year, ~80% de-rate — most of the disappointment is priced in. The hosted transition that's depressing reported revenue is increasing customer lifetime value and recurring quality; when the optics clear, the multiple re-rates toward profitable-SaaS levels. On top sits free optionality the market ignores: a $399M net-cash balance sheet, a pipeline being de-risked off-balance-sheet via partners (Nimbus→Takeda proved the model pays $100M+), and ~$2.27B of gross Novartis milestone potential. Earnings surprise lever: any single partnering deal or milestone is a step-function to the P&L the Street isn't modelling.

Bear case. Three things could permanently impair the equity: (1) Software growth structurally settles in the low teens — at ~11% it already trades like a no-growth software company, and physics-based optimisation may be a narrower, slower-growing TAM slice than "AI drug discovery," with ML rivals expanding the parts of the workflow that grow fastest. (2) The pipeline pivot is a tacit admission the in-house drug bet failed — shareholders funded a decade of R&D and now the upside is shared with partners on terms SDGR doesn't control; if no attractive partner materialises, those assets are stranded sunk cost. (3) Customer concentration + the hosted-recognition drag mean reported revenue can decline (consensus models ~$234M FY2026, down YoY) even as the business is healthy — and this market sells SDGR on reported-software weakness every single time (2021 precedent: −70%). Pre-mortem (18 months out, thesis broken): software ACV guidance for FY2027 comes in at ~8%, SGR-3515 data is unremarkable and no partner bites, the techbio cohort stays out of favour, and equity-FV losses turn the optionality narrative into a liability — the stock is sub-$10 and the "platform re-rate" never came because growth, not multiple, was the problem. Are multiples too high? No — at 2.5x EV/sales they are arguably too low if software reaccelerates; the risk is the fundamentals, not the multiple.

Contrarian view (what the market refuses to see): the market is treating SDGR as a failed drug-discovery company (pipeline shrinking, losses) when it has quietly become a cheap, moated, net-cash compute-for-chemistry software company with a free portfolio of biotech call options. The hosted transition is being read as deterioration when it is the build of a higher-quality recurring base. The asymmetry: limited downside (net cash is ~⅓ of market cap; software alone arguably worth the EV) against a multi-bagger if either (a) software reaccelerates or (b) one Nimbus-style monetisation lands.

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

Dismantling the bull case:

• The growth story is already broken and bulls are in denial. Software grew 38% in 2020 and is ~11% now; ACV "+10–15%" is not a re-rate trigger, it's a mature-software number. The de-rate from 17.7x to 2.5x P/S wasn't an over-reaction — it was the market correctly repricing a decelerating asset. There is no catalyst to reaccelerate a workflow tool that's already in every top-10 pharma (penetration ceiling).
• The pipeline pivot is the tell. A company that believed in its drugs would raise capital and run Phase 2 — it has $399M and chose not to. "Seeking strategic partnerships" for both clinical assets is code for "we can't justify the spend and we're hoping someone takes them off our hands." If the data were strong, partners would already be signed. The most dangerous read: the in-house drug thesis — the entire reason for the premium multiple in 2020-21 — has quietly failed, and what's left is a slow software company.
• Most dangerous competitor bulls underestimate: not a single rival but the ML-first stack collapsing the workflow — Isomorphic Labs (DeepMind), Recursion+Exscientia, and open models commoditising the generative steps, squeezing physics-based tools into an ever-narrower "final optimisation" niche. Even if FEP+ stays best-in-class for affinity ranking, the value capture shrinks if the rest of the pipeline is automated by cheaper tools.
• Capital-allocation/incentives: founder-CEO owns only ~$4M of stock and has only ever sold — if the platform optionality is so compelling, why isn't he buying at a 70%-off price? CFO swapped to a deal-maker mid-pivot. Equity-FV gains have been used to soften the reported loss narrative for years.
• What must hold for today's ~$1.1B EV: sustained ≥10% software growth, and margin recovery post-hosted-transition, and at least one pipeline monetisation, and no further techbio de-rate. That's a lot of "ands."
• Down 20–30% growth scenario: if software growth slips to mid-single-digits and drug-discovery milestones dry up, FY2027 revenue could be flat-to-down ~$230M with no path to breakeven — at a profitable-software 2x that's a sub-$10 stock; the net-cash floor ($399M ≈ $5/share) is the only thing under it.
• Single permanent-impairment scenario (and plausibility): a next-gen ML method achieves FEP-grade affinity prediction without the compute cost, collapsing the core differentiator. Plausibility: low-to-moderate over 3 years (the physics gap is real today per bioRxiv), but rising — and it's the one thing that turns the moat to dust.

Lens 14 · Management Questions (ordered by information value)

1. Software grew 38% in 2020 and ~11% in FY2025 — what is the realistic steady-state organic software growth rate once the hosted transition is fully lapped, and what specifically reaccelerates it above 15%?
2. You have $399M and no debt, yet you're partnering rather than developing SGR-1505 and SGR-3515 yourselves — what return threshold made not funding Phase 2 the better capital-allocation decision?
3. For each lead asset, what deal structure are you targeting (upfront vs milestones vs royalty), and what would you walk away from — i.e. what's the floor that makes partnering better than shelving?
4. How much of FY2025's $199.5M software revenue is expansion within existing top-20 customers vs new logos, and what is net revenue retention?
5. The hosted transition cut software gross margin to 69% — at ~75% hosted mix in steady state, what is the normalised software gross margin and operating margin?
6. One customer was 17% of FY2025 revenue — what is the renewal timing and concentration trajectory, and how do you de-risk a Novartis-scale renewal?
7. Where exactly does FEP+/physics retain a durable edge as ML structure/affinity models improve, and which parts of the discovery workflow do you concede to ML-first tools?
8. The equity-investment portfolio has driven huge non-operating swings and a $147M Nimbus harvest — what's the current portfolio's mark, pipeline of potential monetisations, and how should investors value it?
9. What is the underlying (ex-working-capital, ex-equity-FV) annual cash burn, and at what revenue level does the software business alone reach operating breakeven?
10. Why has the founder-CEO's direct ownership stayed modest with only sell-side transactions — what would change insider buying behaviour?
11. The CFO transition elevated a BD/strategic-finance leader — does this signal a fundamental shift toward a licensing/royalty business model over proprietary drug development?
12. What is the materials-science software opportunity (battery, electronics, aerospace) as a share of ACV today, and its growth rate vs the life-sciences core?
13. How exposed is your hosted economics to hyperscaler GPU/compute pricing, and what's the gross-margin sensitivity?
14. What are the realistic timelines and probability-of-success you assign internally to SGR-3515's path beyond Phase 1, given you're not funding it alone?
15. If the techbio multiple compression persists, would you consider returning capital (buyback at a ~30% discount to cash-rich intrinsic value) given the stock trades near ~3x the net-cash-adjusted software value?

---