Isomorphic Labs

Phase A — Understand the business

Lens 1 · Company Overview

Isomorphic Labs is a London-headquartered, Alphabet-owned AI-first drug-discovery company, spun out of Google DeepMind in 2021 to do one thing: use the protein-structure-prediction breakthrough that became AlphaFold to design medicines computationally, before the wet lab. Founded 24 Feb 2021 and publicly announced 4 Nov 2021. HQ London, with a Lausanne (Switzerland) office opened Dec 2022 and active hiring in Cambridge, Massachusetts. Headcount >350 as of the May 2026 round, with stated plans to expand across all three hubs.

What it actually is. Not a pharma company (no commercial drugs), not a SaaS company (it does not license its software). Management is explicit: "IsoDDE is not a software platform that pharma companies can license or install; we are building a unified drug design engine, not selling software". It is a two-sided R&D engine: (1) a services/collaboration business that takes pharma partners' hardest targets and designs candidate molecules in silico for upfronts + milestones + royalties; and (2) a wholly-owned internal pipeline it intends to advance through early clinical trials and then license out. The asset is the engine + the proprietary biological data + the team — not any single drug.

Key product: the Isomorphic Drug Design Engine (IsoDDE), unveiled Feb 2026 — a unified computational system built atop AlphaFold 3 that moves from target identification → candidate generation → lead optimization in silico, across small molecules, antibodies, peptides, and molecular glues.

Customers (revenue counterparties): Eli Lilly, Novartis, and (since Jan 2026) Johnson & Johnson — see Lens 2. These are the only sources of external cash today; there is no product revenue. Suppliers: compute (Google Cloud TPUs, via Alphabet), wet-lab/CRO capacity for synthesis and assays, and the public + proprietary structural-biology data that trains the models. Competitors: Recursion, Schrödinger, Insilico Medicine, Xaira, Generate Biomedicines, Relay, Iambic, plus every large pharma's internal AI group (Lens 3, 13).

Contract structure: classic biotech "biobucks" — modest upfront, partner-funded research, large back-loaded milestones, royalties on any approved drug. This is lumpy, binary, non-recurring revenue tied to scientific events, not a subscription. Most of the headline deal value is contingent and may never be paid.

Lens 2 · Supply Chain

Map the chain from inputs to medicine:

Upstream inputs →

• Compute & infrastructure: Alphabet / Google Cloud (TPUs, data-center capacity). This is a captive, in-house advantage — Isomorphic trains on Google's frontier compute as an Alphabet subsidiary, a cost/scale edge no independent AI-biotech has. (Chokepoint that becomes a strength: frontier-scale training compute is the field's scarcest input; Isomorphic effectively single-sources it from its parent.)
• Data: public structural biology (PDB, the AlphaFold database of ~200M predicted structures) + proprietary experimental data generated with partners and internally. The proprietary data flywheel is the real moat input (Lens 3).
• Talent: ex-DeepMind ML researchers + recruited medicinal chemists / structural biologists / clinical-development hires. Isomorphic has been actively poaching from rival AI startups.

→ Isomorphic (the engine) →

• IsoDDE designs/optimizes candidate molecules in silico (target ID → hit → lead).

→ Wet-lab / experimental validation →

• CROs / CDMOs and partner labs synthesize and assay the molecules. Critically, in the J&J deal the division of labor is explicit: Isomorphic does in-silico design; Janssen Biotech leads experimental validation and development. Isomorphic is upstream of the bench, not a fully integrated lab.

→ End customer (the pharma partner / eventual patient) →

• Lilly, Novartis, J&J take candidates into IND-enabling work, clinical trials, regulatory filing, manufacturing, and commercialization. Isomorphic does not (yet) run its own late-stage trials or sell drugs — it hands the molecule (and, for internal programs, plans to license after early trials).

Single-source dependencies / chokepoints:

1. Alphabet for capital + compute + brand. ~$2.6B raised, but Alphabet is the anchor owner and infrastructure supplier — concentration risk and dependence (Lens 13).
2. Pharma partners for the entire downstream. Isomorphic controls design; it does not control whether a partner advances, deprioritizes, or kills a program. A partner's pipeline reshuffle can strand a collaboration (the Exscientia/Lilly-type risk).
3. The wet-lab reality gap. The chain's weakest link is the handoff from elegant in-silico molecule to a compound that survives synthesis, ADME/tox, and human biology — where AI-designed drugs have historically died (Lens 13).

Lens 3 · Competitive Advantages (moats)

1. The science lead is real and externally validated. AlphaFold 2 solved a 50-year grand-challenge (protein folding) and earned Hassabis and John Jumper the 2024 Nobel Prize in Chemistry (shared with David Baker). AlphaFold 3 (May 2024) extended this to protein-ligand, protein-protein, and nucleic-acid complexes. IsoDDE (Feb 2026) claims to "more than double the accuracy of AlphaFold 3" on the hardest generalization systems (0–20% similarity to training set), beat physics-based FEP methods on binding affinity, and outperform AF3 by 2.3× / Boltz-2 by 19.8× on high-fidelity antibody-antigen interfaces. (These are company-reported benchmarks, not independently replicated — see Lens 13.) This is the deepest structural-prediction bench in the industry.

2. Proprietary data + compute flywheel (durable, hard to copy). As an Alphabet subsidiary, Isomorphic trains on Google's frontier TPUs and accumulates proprietary experimental data from every partnered program. Each program feeds the engine; a better engine wins more programs. This is the network/data-scale moat — the asset MedCity/skeptics correctly identify as the structural concern: "if AI-designed drug discovery works at scale, the competitive dynamics … will reshape around whoever controls the most capable models and the most proprietary biological training data, which smaller biotech companies and academic programs cannot match".

3. The closed-engine strategy (deliberate moat choice). Unlike AlphaFold (open, freely available) or open-source rivals (Uni-Mol2, DiffDock, Boltz), IsoDDE is proprietary and not licensable — Isomorphic keeps the best engine behind the wall and monetizes outcomes, not seats. This maximizes value capture if the engine produces drugs, but it also means the moat is unproven until a molecule clears the clinic.

4. Brand / talent gravity. A Nobel-laureate founder, the AlphaFold halo, and Alphabet's balance sheet make Isomorphic a magnet for the scarce ML-x-biology talent the field competes for.

Bargaining power. Over partners: high and rising — three of the largest pharmas (Lilly, Novartis, J&J) came to it, and the deals are structured so Isomorphic keeps optionality and royalties. Over suppliers (compute): effectively internalized via Alphabet. The asymmetry of need: pharma needs a productivity breakthrough more than Isomorphic needs any single partner — but Isomorphic needs someone's clinical machine to validate the molecules, so it is not fully independent. Caveat: none of this moat is proven until the base rate of clinical success actually improves; today it is a moat on potential.

Lens 4 · Segments

No segment financials exist — Isomorphic is private and discloses no revenue split. segments.csv is empty; every figure here is n/a — private, not disclosed. The closest thing to a "segment" view is the structure of the business, which can be described qualitatively:

Geographic split: n/a — private, not disclosed (operations London / Lausanne / Cambridge MA; partners US + Switzerland). The honest read: there are no segment numbers to analyze; the "mix" is partnered-vs-internal program count, and both are growing while neither yet generates product revenue.

---

Phase B — Measure performance

+private swap: Lens 5 → Funding & valuation trajectory (+ pre-revenue traction). +clinical swap: Lens 5 also runs Pipeline-by-phase (the asset table is the company). +private swap: Lens 7 → Cap table & secondary marks (+ mechanism comps). Lens 6/8 carry with founder-interview / funding-event framing.

Lens 5 · Funding & valuation trajectory + Pipeline-by-phase

Funding & valuation trajectory (all ``, unaudited):

• Total raised ~$2.6B. The $2.1B Series B is reportedly the second-largest biotech financing in history, behind only Altos Labs' ~$3B.
• The syndicate is the tell. The new entrants — MGX (Abu Dhabi sovereign), Temasek (Singapore sovereign), CapitalG (Alphabet's growth fund), and the UK Sovereign AI Fund — are crossover/sovereign capital, the kind that shows up ahead of an eventual IPO or to back a strategic national-champion asset. Thrive leading both rounds signals continuity of conviction. This is not a normal biotech cap table; it is an AI-platform cap table.
• Burn signal: raising $2.6B for a >350-person pre-revenue org implies a large and growing burn (frontier compute + a clinical pipeline are both expensive). With $2.1B fresh, runway is multi-year (Lens 11), but the spend only converts to value if the clinic validates the engine.

Traction / unit economics (+private add): no ARR, no product revenue. The only "traction" metrics are ~$3B in partnered biobucks signed (mostly contingent), 3 top-tier pharma partners, and ~17 active programs. Gross-margin / unit economics: n/a — pre-revenue.

Pipeline-by-phase (+clinical swap — the asset table is the company; all ``, none independently verified):

Hard truth for the asset table: Isomorphic has disclosed ~17 programs across three therapeutic areas, but has named zero clinical candidates and zero disease targets. The "first AI-designed drug into the clinic by end-2026" is a Phase-1 entry, which tests safety/PK in healthy volunteers — it is a milestone of timeline, not of efficacy. No Isomorphic molecule has produced a human-efficacy datapoint. (Note: the widely-cited "ISM8969 FDA-cleared Jan 2026" milestone belongs to Insilico Medicine — a Parkinson's NLRP3 inhibitor — not Isomorphic; do not attribute it here.)

Lens 6 · Founder & narrative signal (calls → interviews swap)

No earnings calls exist (private). The +private substitute is the trajectory of founder commentary and disclosure posture over the last ~18 months:

• Tone: confident, deliberately opaque, increasingly under scrutiny. Hassabis has consistently framed Isomorphic as aiming to "solve disease" and compress drug timelines, but the company tells you almost nothing concrete. The most-watched shift: the clinical-start timeline slipped from "end of 2025" to "before the close of 2026," with Hassabis explicitly saying he had "misspoke" on the earlier date. A founder walking back a timeline is a yellow flag on execution discipline, even if the science is sound.
• What they keep saying: AlphaFold lineage, "drug design engine," undruggable targets, "bring treatments to millions." What they won't say: which drugs, which diseases, any clinical timeline specifics, any valuation. PitchBook's Ben Zercher: "both companies gave far less transparency than is standard for biotech companies raising large rounds".
• Net: the narrative is high-conviction and well-funded but evidence-light. In a normal biotech, investors impose "a higher evidence barrier" [Zercher, web]; Isomorphic's brand + Alphabet backing let it raise $2.1B without clearing that bar. Sentiment among believers is euphoric; among biotech veterans it is "show me a readout."

Lens 7 · Cap table & secondary marks + mechanism comps

Cap table & secondary marks (+private swap):

• Controlling owner: Alphabet Inc. Isomorphic is an Alphabet subsidiary (an "Other Bet"-style asset under the DeepMind umbrella), not an independent VC-backed startup. Exact post-Series-B Alphabet stake is n/a — not disclosed, but Alphabet was the sole/anchor funder pre-2025 and a follow-on participant in both external rounds, so it remains the majority/controlling shareholder. This is the single most important cap-table fact: unlike OpenAI (Microsoft minority) or xAI, Isomorphic is controlled by a $2T+ public parent.
• Syndicate quality: top-tier + sovereign + crossover. Thrive Capital (lead, both rounds — IPO-proximity-grade growth investor), GV + CapitalG (both Alphabet-affiliated), MGX + Temasek + UK Sovereign AI Fund (sovereigns). The presence of crossover/sovereign money is a classic pre-IPO-readiness tell — except that Alphabet's control mutes it (Lens 11). Secondary marks / mutual-fund markups: n/a — not disclosed.

Mechanism / business comps (+clinical/+private swap — comparables are AI-platform peers by business model, not P/E; multiples are `` or n/a):

Lens 8 · Catalysts / value-inflection events (price-move proxy)

No tradeable price, so the proxy is events that re-rated the private mark or the field's perception (``):

What the pattern reveals: the mark re-rates on (1) science milestones (AlphaFold versions, Nobel), (2) pharma deals (external validation that someone will pay), and (3) mega-rounds. It has not yet re-rated on clinical data, because there is none. The next — and first truly decisive — catalyst is a human-efficacy readout. Until then, every catalyst has been about capability and credibility, not proof a drug works.

---

Phase C — Judge people & books

Lens 9 · Management

• Demis Hassabis — Founder & CEO (also CEO of Google DeepMind). The strongest founder credential in the field: chess prodigy, founded DeepMind (2010, sold to Google 2014), led AlphaGo (beat Lee Sedol 4-1, 2016; Ke Jie 2017) and AlphaFold, 2024 Nobel laureate in Chemistry, Royal Society Fellow, Breakthrough/Lasker/Gairdner awards, Time 100 (2017, 2025), one of Time's 2025 "Architects of AI" Persons of the Year. Track record: repeatedly took a hard scientific problem from "impossible" to "solved." Caveat (skin-in-the-game / focus): he runs two organizations — his primary role is Google DeepMind. Isomorphic's day-to-day leans on the executive bench.
• Max Jaderberg — President (effective 1 Jan 2026), previously Chief AI Officer; led the AI models incl. AlphaFold 3. Note for the file: a stray search framing suggested a "departure" — that is wrong; Jaderberg was promoted to President, not lost. Strong AI-research pedigree (ex-DeepMind).
• Bench (early/known): Colin Murdoch (President, business — ex-DeepMind), Miles Congreve (CSO — veteran medicinal chemist), Sergei Yakneen (CTO), Pamela Carroll (COO/oncology). The deliberate build was C-suite first, then therapeutics + ML depth — i.e. they bolted real drug-development experience onto the AI core, which is the right instinct (the field's failures often come from ML teams that under-respect biology).
• Capital-allocation history: as a subsidiary, "capital allocation" = how it deploys Alphabet + investor money. So far: build the engine, sign partners on favorable (optionality-preserving) terms, fund an internal pipeline, raise at escalating marks. No buybacks/dividends (irrelevant). No value-destruction signals, but also no realized financial returns to judge — ROIC is unmeasurable pre-revenue.
• Founder vs. professional manager: founder-led (Hassabis), which suits a moonshot R&D company at this stage — but with the unusual wrinkle that the founder is part-time and the parent (Alphabet) holds control.
• Red flags: (1) disclosure opacity beyond biotech norms (Lens 6, 13); (2) a slipped timeline the CEO walked back ("misspoke"); (3) divided founder attention (DeepMind is the day job); (4) related-party texture — Alphabet is owner, infrastructure supplier, and co-investor simultaneously (not improper, but concentrated).

Lens 10 · Forensic Red Flags

No financial statements exist — there is no income statement, balance sheet, or cash-flow statement to forensically examine, because Isomorphic is private and unaudited. Standard forensic vectors (revenue recognition, receivables vs. revenue, SBC flattering non-GAAP, goodwill) are n/a — private, no audited financials. The forensic lens therefore re-points (per +private/+clinical) to disclosure integrity, trial-design/claim integrity, and going-concern:

• Disclosure integrity — flag. Raising $2.1B with zero disclosed drug programs, targets, or clinical timelines is below the evidence bar biotech investors normally enforce. Not fraud — but a governance/transparency posture that asks investors (and any future public-market buyer) to underwrite on faith + brand.
• Benchmark/claim integrity — watch. The IsoDDE "2× AlphaFold 3 / beats FEP / 19.8× Boltz-2" numbers are company-reported and not independently replicated. Separately, independent literature documents real AlphaFold 3 limitations the marketing glosses: ~22% residue hallucination in intrinsically-disordered regions, predictions that are static (ignore binding dynamics/conformational change), chirality errors, and atomic clashes on large proteins; one analysis even questions AF3's own benchmark choices (citing AutoDock Vina as "SOTA"). Treat headline accuracy claims as vendor benchmarks until a molecule validates them.
• Going-concern: not a concern — $2.1B fresh + Alphabet parent = multi-year runway (Lens 11). The risk is not insolvency; it is spending years and billions without a validating readout.

Regulatory findings (required sub-section):

• SEC (EDGAR LR + AAER): none possible / none found — Isomorphic has no CIK (private, not an SEC filer). regulatory/regulatory-findings.md (Step 0) confirms total_sec_findings: 0 and notes no EDGAR search is possible.
• Non-SEC (FTC/DOJ/FDA/etc.): web search for "Isomorphic Labs" (FTC OR DOJ OR FDA OR consent decree OR settlement OR fine OR penalty) surfaced no material enforcement actions, fines, or consent decrees against Isomorphic as of 2026-06-30. (The only "FDA"-adjacent hit, ISM8969's IND clearance, belongs to Insilico, not Isomorphic.)
• Item 3 (Legal Proceedings): n/a — no 10-K exists.
• Net: No material regulatory or legal findings — verified via SEC EDGAR EFTS (no CIK → none), web search (none material), as of 2026-06-30. Unaudited per public-sources caveat applies. Isomorphic does, however, carry the future regulatory dependency of any drug developer: its value ultimately routes through FDA/EMA review of molecules it (or its partners) advance.

---

Phase D — Project & stress-test

+clinical/+private swap: Lens 11 → IPO-readiness & path-to-tradeable + runway-to-catalyst (no EPS; rNPV framing for the asset).

Lens 11 · IPO-readiness & path-to-tradeable + runway/rNPV

No EPS projection is possible (pre-revenue, private). The two questions that matter are (a) when/whether this becomes tradeable, and (b) does cash reach the next value-inflection.

IPO-readiness (grounding from private-watch.json): stage late, ipo_readiness: 3 (late-stage) on the 1–5 scale, lead investors Alphabet + Thrive, catalyst = "AlphaFold-derived drug design; pharma deals (Lilly, Novartis)". This dossier argues that entry is now stale and should arguably be downgraded on tradeability even as the business strengthens — here's the tension:

• Arguments for higher readiness: $2.1B Series B at ~$15–20B, sovereign + crossover money (MGX/Temasek/UK fund), Thrive leading — exactly the syndicate that precedes an IPO.
• The decisive counter: Alphabet controls it. A controlled subsidiary of a $2T+ parent does not need to IPO for capital and may never spin out — Alphabet can fund it indefinitely (cf. Waymo, Verily, which have stayed private for years). So the "path to tradeable" is genuinely uncertain and likely long; this is not a name a public-markets investor can buy soon. The realistic tradeable events are (1) an eventual Alphabet-blessed IPO (years away, contingent on clinical validation), or (2) secondary access via the private syndicate. Net readiness: ~3, capped by parent control — high financing-readiness, low independence-readiness.

Milestones that would unlock an S-1 / re-rate: first-in-human Phase 1 start (end-2026, timeline-only); first human-efficacy readout (the real one, likely 2028+); a partnered program hitting a clinical milestone (cash + validation); and, structurally, an Alphabet decision to let it stand alone.

Runway-to-catalyst (the question that actually matters): with ~$2.1B fresh + Alphabet backstop, runway comfortably reaches the end-2026 Phase-1 start and the multi-year wait for first efficacy data. Runway is not the risk; the readout is.

rNPV sketch (illustrative, ``, do not trust as a mark): a single Phase-1-entry oncology asset with ~$1–2B unrisked peak sales × ~10% PoS-to-approval × discount ≈ low-hundreds-of-$M rNPV per program; ~17 programs of varying maturity + ~$3B in contingent biobucks builds a multi-billion risk-adjusted asset base — but the ~$15–20B mark prices in engine-level success across the portfolio (a platform re-rate), not the sum of today's de-risked assets.

(No forecast.ts create in --watchlist mode. The forecast to log on a future conviction pass is binary: "Isomorphic's first wholly-owned candidate enters Phase 1 by 2026-12-31, p≈0.6" — a timeline bet, explicitly not an efficacy bet.)

Lens 12 · Bull vs Bear

Bull case. Isomorphic owns the best computational drug-design engine in existence, validated by a Nobel Prize and three of the largest pharmas paying real money for access. If IsoDDE genuinely lifts the base rate of clinical success — even modestly, from ~10% toward, say, 15–20% at Phase 1 — the value is enormous, because drug R&D's core problem is attrition, and Isomorphic attacks it at the design stage where errors are cheapest to fix. It has Alphabet's compute and balance sheet (no funding risk), a proprietary-data flywheel that compounds with every program, a closed-engine strategy that captures the upside, and ~17 shots on goal across oncology/immunology/cardiovascular plus three partner pipelines. The crossover/sovereign syndicate at ~$15–20B is betting this becomes a generational platform. The asymmetric bull: this is the company most likely to produce the first unambiguously AI-designed approved drug, and the first such readout would re-rate it (and the field) violently.

Bear case (permanent-impairment risks). (1) The engine may not change the base rate. AI-designed drugs have a brutal track record — Recursion's first trial showed no reportable efficacy, Exscientia discontinued its lead, and a decade of the field is "failure after failure." Elegant in-silico binding does not predict tox, off-target effects, or human PK; biology is where these die. If Isomorphic's molecules fail in humans at the normal rate, the entire ~$15–20B premium — which is priced on the engine working — evaporates. (2) It's priced for success it hasn't shown. ~$15–20B is 6–10× Recursion's whole market cap and ~4× Relay's, with less clinical validation than either; that is a platform multiple on a pre-clinical asset. (3) Alphabet control = no clean exit and divided founder attention — a public-markets investor can't buy it, the founder is part-time, and the parent could deprioritize it. Pre-mortem (18 months out, thesis broke): the first wholly-owned candidate either slipped again past 2026 or entered Phase 1 and read out unremarkably on safety, while a partnered program got deprioritized — the market realized "AI designed it" ≠ "the drug works," and the private mark got cut on the next round as biotech-style skepticism finally applied to it. Are multiples too high? Against clinical-stage biotech logic, plainly yes. Against frontier-AI-lab logic, defensible. Contrarian view of what the market refuses to see: the bull narrative quietly conflates "AlphaFold predicts structure brilliantly" with "therefore Isomorphic will design drugs that work in people" — but structure prediction and clinical efficacy are different problems, and no amount of structural accuracy has yet been shown to move the clinical base rate. The market is paying for the second claim while only the first is proven.

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

Dismantling the bull case as a skeptic:

• Where the money breaks: Isomorphic has no product revenue and no licensable software — 100% of upside routes through (a) contingent pharma milestones it doesn't control and (b) an internal pipeline that hasn't cleared the clinic. If partners deprioritize programs (routine in pharma) and internal assets fail in humans (the base case statistically), there is no recurring revenue floor — unlike Schrödinger, which at least has a real $255M software business.
• Revenue concentration: three partners (Lilly, Novartis, J&J) + one owner (Alphabet). Lose a partner's commitment or Alphabet's patience and the picture changes fast.
• Why the moat may be weaker than bulls think: the scientific moat (structure prediction) is partly commoditizing — open-source AF3-class models (Boltz, Chai, HelixFold) and every pharma's internal AI group are closing the gap on the published benchmarks. Isomorphic's edge is the proprietary data + compute + the unproven claim that its closed engine is meaningfully better at the part that matters (designing molecules that work in humans) — and that last part has zero clinical evidence behind it.
• Most dangerous competitor bulls underestimate: not another startup — it's large pharma's own internal AI + the open-source community. If a $50B-R&D-budget pharma's in-house AI gets "good enough," Isomorphic's design-services value compresses. And open models erode the "only-we-have-it" framing.
• Worst governance/incentive issues: opacity (raised $2.1B disclosing nothing), divided founder (Hassabis's day job is DeepMind), related-party concentration (Alphabet as owner+supplier+co-investor), and a slipped, walked-back timeline.
• What must hold for ~$15–20B: that IsoDDE materially raises clinical success rates (unproven), that multiple programs reach the clinic and read out positively, that partners stay engaged, and that the platform-multiple framing survives contact with the first real readout. If growth/validation disappoints by 20–30% — i.e. timelines slip again and the first readouts are unremarkable — the private mark is highly vulnerable to a down-round, because there's no revenue to backstop it.
• Single scenario that permanently impairs: the first few Isomorphic-designed molecules to reach humans fail at the normal rate, publicly demonstrating that AI-design did not change the base rate. That would re-rate not just Isomorphic but the entire AI-drug-discovery thesis — and there is no P&L to cushion the fall. Plausibility: uncomfortably high, given ~90% Phase-1→approval attrition and the field's poor track record.

Lens 14 · Management Questions (ordered by information value)

1. What is the clinical success rate of IsoDDE-designed molecules so far — for every molecule the engine has designed that reached IND-enabling tox or beyond, how many advanced vs. failed, and how does that compare to your historical-industry base rate? (The whole thesis lives or dies here.)
2. When does your first wholly-owned candidate read out human-efficacy data (not just enter Phase 1) — and what specific endpoint will tell you the engine worked?
3. You've named no drugs, targets, or timelines while raising $2.6B. What is the disclosure roadmap — at what milestone do you start reporting like a biotech?
4. Independently replicate the IsoDDE benchmarks: will you publish head-to-head, third-party-audited comparisons vs. AF3/Boltz on prospective (not retrospective) targets?
5. Of your ~17 programs, how many are internal vs. partnered, and what is the phase distribution today?
6. On the partnered deals (Lilly/Novartis/J&J), how many milestones have actually been earned and paid vs. signed "biobucks"?
7. What is your current annual burn, and at what point do milestone inflows offset it?
8. Hassabis runs DeepMind full-time. Who makes Isomorphic's day-to-day scientific and capital decisions, and how is your attention split?
9. What is Alphabet's ownership stake and governance role post-Series B, and what is the explicit path (and timeline) to becoming an independent/tradeable company?
10. Where does IsoDDE still fail — disordered proteins, conformational dynamics, novel chemotypes — and how do those gaps affect the molecules you advance?
11. How do you defend against open-source structure models and pharma's internal AI commoditizing your benchmark edge?
12. What happens to a program if a partner deprioritizes it — do you retain rights, and have any been returned?
13. For internal assets, what is the license-out strategy and at what stage — and what economics do you expect to keep?
14. How much of your edge is the model vs. the proprietary data + Alphabet compute — and which is more durable?
15. What single result in the next 24 months would most change your own confidence in the platform?

---