Vault Bio

• Sector frontier is real: generative models (RFdiffusion, ProteinMPNN, ESM-2) plus wet-lab-in-the-loop have demonstrably designed novel functional and thermostable enzymes; de novo backbone design with directed-evolution refinement is a validated 2022–2024 workflow, so the core approach is grounded, not hand-wavy.
• Business model is industrial biocatalysis, NOT therapeutics — so the quant model's 46/100 feasibility, 36/100 regulatory drag and 8–12yr binary-trial framing are largely miscalibrated; specialty-chemical enzymes face no FDA clinical trials, timelines are 12–36mo to pilot, de-risking the single biggest scored penalty.
• Traction is the strongest signal for Series A: 3 paid discovery contracts with chemical majors, one enzyme already at pilot-scale, and $3.4M committed milestones — evidence of real technical delivery and customer pull rather than a pure platform story.
• Genuine value creation requires enzymes that hit industrial process specs (>60–80°C thermostability, organic-solvent tolerance, kg-scale titer, cost < incumbent petrochemical catalyst); design-and-validate is proven at bench, but manufacturability at ton-scale via fermentation is the unproven leap.

• Scale-up/manufacturing risk: designing a thermostable enzyme in silico is far easier than achieving cost-competitive fermentation titers and process integration at industrial volume — most enzyme startups die in the pilot-to-commercial gap, not at design; 95% capital intensity score reflects this.
• Value-capture asymmetry: chemical majors fund discovery contracts to internalize IP and know-how; risk that Vault is a fee-for-service CRO with milestone crumbs rather than owning royalties/recurring revenue — patent moat means little if customers own the resulting strain.
• Competition from incumbents (Novozymes/Novonesis, Codexis, Ginkgo, Arzeda) with decades of biocatalysis IP and installed fermentation capacity; a generative-design edge may compress as foundation-model tooling (open-source ESM/RFdiffusion) commoditizes the design step.

• Category mismatch inflates the model: the $1.77T TAM and 8-12yr trial/regulatory drag reflect therapeutics, but Vault sells industrial enzymes to chemical majors — no clinical trials, no FDA approval path, revenue in quarters not decades. True SAM is the ~$6-7B industrial enzyme market (specialty-chem catalysis subset likely $1-2B SOM), so market-size score (74) is directionally right but for the
• Traction is early but real for Series A: 3 paid discovery contracts + 1 pilot-scale enzyme + $3.4M committed milestones validate demand. However $3.4M is committed, not recognized recurring revenue — need to see conversion from discovery-fee to production royalty/volume economics, which is where enzyme platforms actually monetize.
• Unit economics claim (~85% mature gross margin) is plausible for licensed/royalty enzyme IP but the 95% capital intensity flag is the tension: wet-lab-in-the-loop + pilot/scale-up fermentation is capex-heavy. $22M Series A is thin if they must fund scale-up manufacturing vs. licensing to partners' plants — clarify capital-light licensing vs. capital-heavy production model.
• Moat rests on IP patents (80) but generative-protein design is a crowded frontier (Cradle, Basecamp, Arzeda, Codexis, Ginkgo). Defensibility likely comes from proprietary wet-lab data flywheel + design-test cycle speed, not patents alone — need cycle-time and hit-rate metrics to confirm.

• Data gaps: no CAC/LTV, no contract-to-royalty conversion rate, no gross margin realized (only 'mature' projection), no cash runway or burn disclosed — cannot verify unit economics or timeline to self-sustaining revenue on $22M.
• Enterprise-sales concentration: 3 chemical majors = high customer concentration; discovery contracts are often exploratory and may not convert to volume production, and incumbents (Novozymes/Novonesis, DuPont) can in-license or build competing enzymes.
• Kill metric — technical hit rate: if designed enzymes fail thermostability/activity at pilot-to-commercial scale-up (the classic in-silico-to-fermentation gap), the platform's core value prop collapses regardless of market size.

• Sector-classification mismatch inflates the model: this is industrial biocatalysis, not drug development. It avoids the $1.77T therapeutics TAM but also the binary trial risk — realistic served market is closer to the ~$6-8B industrial enzymes segment (Novozymes/DSM-Firmenich duopoly, ~10% CAGR), not $1.77T.
• Demand is derived and price-elastic: chemical majors adopt bio-catalysts only where they beat petrochemical routes on delivered cost. Enzyme swaps must clear a customer hurdle of both capex retrofit and per-kg economics vs. incumbent thermal catalysts — adoption is gated by feedstock/energy spreads, not just performance.
• Moat is dual: sequence IP patents plus a proprietary wet-lab-in-the-loop dataset that compounds with each contract. This is the durable rent source — the design-test data flywheel is harder to replicate than the generative model itself, which is increasingly commoditized (AlphaFold, ESM open-weights).
• Traction quality is high for Series A: 3 paid discovery contracts + 1 pilot-scale enzyme + $3.4M committed milestones shows genuine willingness-to-pay from majors. But contract R&D revenue is not equity-value-creating unless it converts to royalty/production economics with recurring per-kg margins.

• Value-capture leakage: chemical majors have far more bargaining power than a Series-A vendor. Discovery contracts can become work-for-hire where the customer captures the production economics, leaving Vault as a low-multiple services shop rather than a royalty-bearing platform.
• Capital intensity 95% + wet-lab loop means each design cycle is expensive and slow; $22M funds a limited number of enzyme campaigns before needing scale-up capex or partner-funded manufacturing. Runway-to-recurring-revenue gap is the central execution risk.
• Macro sensitivity: the entire substitution thesis depends on petrochemical/energy price spreads. A sustained drop in oil/gas prices erodes the cost advantage of bio-catalysts, deferring customer adoption regardless of technical success.

• Regulatory posture is materially better than the 36/100 therapeutics-coded score implies: industrial specialty-chemical enzymes are governed by EPA TSCA (biotech MCAN filings under 40 CFR 725, ~90-day review) and OSHA/worker-exposure rules, NOT FDA IND/BLA trials — the 8–12yr binary trial framing does not apply to this use case and should be re-scored upward.
• Moat is genuine but composition-of-matter claims on engineered enzyme sequences are narrower than they appear: enzyme designs are easily engineered-around by competitors optimizing different residues, so protect via trade-secret (fermentation/process know-how) plus patents; confirm freedom-to-operate against Codexis, Novozymes/Novonesis, Ginkgo, and academic directed-evolution IP.
• Data/privacy exposure is low (no PHI/patient data despite the organoid-screen sector tag) — key contractual issue is IP ownership in the 3 discovery contracts: verify Vault retains platform IP and background models while customers get only field-limited licenses to delivered enzymes; grant-backs and improvements clauses are the deal-critical terms.
• Investor protections for a $22M Series A: standard 1x non-participating liquidation pref, milestone-based tranching tied to the $3.4M committed payments and pilot-to-commercial conversion, IP assignment/inventor-assignment audit, and reps on FTO and open-source model-weight/licensing hygiene given generative-AI training-data provenance risk.

• Customer-concentration + IP leakage: 3 chemical majors control the revenue and could insource or reverse-engineer enzymes after discovery contracts; weak field-of-use carve-outs or broad improvement grant-backs would strip platform value.
• Generative-model provenance: training data or third-party model dependencies (open weights, restrictive licenses) could impair ownership of designed sequences or create infringement/derivative-work claims — diligence the AI stack's licensing chain.
• Capital intensity (95%) means scale-up to commercial fermentation is capex-heavy and outside the platform's asset-light thesis; a down-round or dilutive process-plant financing is plausible before enzymes reach margin-positive commercial volume.