AI-Bio
A real platform with a lottery ticket attached — the legacy biocatalysis P&L is shrinking and the entire equity is a bet that enzymatic siRNA manufacturing (ECO Synthesis) becomes the standard before the cash runs out in 2027; the 4x off the $0.96 March low has front-run any GMP revenue.
Research
The verdict
A real platform with a lottery ticket attached — the legacy biocatalysis P&L is shrinking and the entire equity is a bet that enzymatic siRNA manufacturing (ECO Synthesis) becomes the standard before the cash runs out in 2027; the 4x off the $0.96 March low has front-run any GMP revenue.
Codexis is a protein-engineering company that sells enzymes and enzyme-based manufacturing technology to drug makers. The engine is the CodeEvolver directed-evolution platform (in development since 2002) — machine-learning-guided design + high-throughput screening of thousands of enzyme variants to tailor an enzyme for a specific reaction. The company makes money two ways, and the mix is the whole story:
Two business areas: (1) small-molecule pharma biocatalysis — the mature, cash-generative legacy (enzymes that make small-molecule APIs cleaner/cheaper; 18 commercially-approved drugs use Codexis enzymes, plus 15 Phase 2/3 candidates as of 2025-12-31); and (2) ECO Synthesis — the growth bet, an enzymatic platform to manufacture RNAi/siRNA therapeutics. Crucially, in November 2025 management announced it is reducing emphasis on new small-molecule biocatalysis projects "due to reduced pricing opportunity of the broader enzyme market". So the company is deliberately letting its profitable legacy fade to concentrate on the unproven siRNA bet.
Customers: large pharma. Platform-technology licensees are GSK, Merck & Co., and Novartis; other recent licensees include Pfizer (Dec 2024), Aldevron (Codex HiCap RNA Polymerase), and Alphazyme. Customer concentration is severe — one customer was ~51% of FY2025 total revenue (Customer "B", i.e. Merck). Contract terms are weak from a predictability standpoint: supply agreements are finite-duration, often not "take-or-pay," and many orders are cancellable POs — management explicitly says backlog is not meaningful.
Upstream inputs (raw materials, fermentation feedstock) → third-party contract manufacturers → Codexis QC/release → pharma customer / CDMO. Codexis has limited internal manufacturing (Redwood City HQ handles only small-scale, panels/kits, and pilot/clinical quantities) and outsources large-scale enzyme production. Named CMOs (manufacturing in four locations total incl. in-house):
This is a single-source / limited-source chokepoint risk flagged in the 10-K: "we are dependent on a limited number of third-party contract manufacturers for large scale production of substantially all of our enzymes".
Downstream, the ECO Synthesis model adds a new dependency: Codexis itself does not yet make GMP-grade siRNA. It runs an ECO Synthesis Innovation Lab (non-GMP, gram-scale, built end-2024) and is partnering with three large CDMOs (feasibility stage; plans to advance ≥1 to tech-transfer in 2026) to produce GMP material. To capture more of the value chain it signed a lease in Nov 2025 for a GMP facility in Hayward, CA — construction expected fall 2026, full production by end-2027. Until then the siRNA growth story is gated on partners it doesn't control.
The defensible asset is CodeEvolver + ~1,600 active/pending patents (as of 2025-12-31; CodeEvolver-platform US patents expire 2029–2034, broader estate to ~2046) plus the Maxygen "MolecularBreeding" IP acquired 2010, and ~92 trademark registrations. The moat in pharma biocatalysis is genuine but narrow: once a Codexis enzyme is designed into a customer's approved commercial process, switching means re-validating with the FDA — high switching costs on installed assets. But the market is fragmented and price-competitive, and management is walking away from it on pricing, which tells you the moat there is not wide enough to defend margins.
For ECO Synthesis, the claimed advantages are real on paper: vs. the 40-year-old SPOS/phosphoramidite chemical standard, the enzymatic/aqueous route promises (a) much larger batch sizes (tens-to-hundred kg/run vs. single-digit kg for SPOS), (b) far less acetonitrile/toxic-solvent waste, (c) integration into existing facilities with less capex, and (d) one enzyme set re-usable across many siRNA assets (vs. bespoke per-asset enzyme projects in biocatalysis). The cross-sell logic — many existing biocatalysis pharma customers are also building RNAi pipelines — is a credible commercial-reach advantage.
Bargaining power is weak. Codexis is a ~$224M micro-cap selling to GSK/Merck/Novartis; the customer holds the power (one of them is half the revenue). Over CDMO partners it is also the smaller party. The moat is a technology lead in a nascent category, not structural pricing power.
One reportable segment since Q4 2023 (the prior "Performance Enzymes" and "Novel Biotherapeutics" segments were collapsed when biotherapeutics was discontinued). So the only meaningful disaggregation is product-type and geography.
By product type (FY, $000s):
| Line | FY2023 | FY2024 | FY2025 | Trend |
|---|---|---|---|---|
| Product revenue | 42,906 | 36,786 | 26,028 | decelerating / shrinking (−14%, then −29%) |
| R&D revenue | 27,237 | 22,559 | 44,359 | lumpy; +97% in FY2025 on Merck |
| Total | 70,143 | 59,345 | 70,387 | flat over 3 yrs despite the Merck spike |
By geography (FY2025, $000s): Americas $43,537 (US $43.5M), EMEA $9,292 (Switzerland $4.2M, Ireland $0.7M), APAC $17,558 (China $9.3M, India $5.7M, Singapore $0.4M). The 3-year shift is dramatic: APAC collapsed from $33.5M (FY2023) to $17.6M (FY2025) and the Americas tripled from $13.7M to $43.5M — the business has re-centered on US pharma (the Merck deal) as Asian volumes fell. The takeaway: the headline "flat-ish $70M" masks a shrinking recurring product base propped up by a one-off US license.
No transcripts in the research layer (transcripts=0); sentiment from web call coverage. The arc across the last several quarters is a deliberate narrative pivot from "diversified enzyme company" to "the siRNA manufacturing platform company." Management's recurring talking points: an expected "inflection in 2026" as existing customers move into larger ECO Synthesis contracts; a stated $2B TAM by 2030 for the platform; a commercial pipeline of "50+ opportunities across ~40 companies" and three CDMO partners; reaffirmed $72–76M guide. What they've stopped emphasizing: new small-molecule biocatalysis wins (explicitly de-prioritized) and anything about the old biotherapeutics pipeline (divested). Tone is "balancing breakthroughs and risks" — the bull framing is credible-but-promissory (data and contracts are "expected in 2H 2026"), which is exactly the kind of forward-leaning language that gets punished if catalysts slip.
| Company | Ticker | Mkt cap | TTM rev | EV/Sales | Note |
|---|---|---|---|---|---|
| Codexis | CDXS | ~$224M | ~$70M (FY25) | ~2.7x (FY26E) | enzymatic siRNA mfg + biocatalysis |
| Maravai LifeSciences | MRVI | ~$1.24B | ~$205M | ~6.7x | RNA/mRNA reagents (TriLink); oligo scale-up |
| Ginkgo Bioworks | DNA | ~$581M | ~$180M (TTM); ~$19.5M Q1'26 cont. ops | ~3.2x P/S | synthetic-bio platform; revenue halved, divested biosecurity |
P/E, EV/EBIT, dividend yield, 5-yr ROE for all three: n/a — all are loss-making or in transition; not sourced as clean multiples. Read: CDXS trades at a discount to MRVI on EV/Sales but that gap is justified — MRVI has higher-quality recurring reagent revenue, while CDXS's "growth" line is a single lumpy license. The comp set tells you the market is pricing CDXS as an option, not a going concern with durable revenue.
Pattern: the market reacts to (1) license/milestone-driven revenue surprises and (2) ECO Synthesis validation/partnership news — and almost not at all to the slow bleed in the legacy product line. This is a binary-narrative stock: it trades on whether the siRNA platform is being adopted, not on quarterly product run-rate.
Accounting quality is, somewhat surprisingly, clean — the risks here are commercial, not forensic.
Regulatory findings (required sub-section):
regulatory/regulatory-findings.md (EDGAR EFTS, LR + AAER, period 2021-06-20 → 2026-06-20) returned 0 findings.Company is loss-making, so the projection is revenue / cash-runway / loss-narrowing, not EPS-positive. Build from the FY2026 guide and the disclosed cost trajectory:
Inputs labeled: product line assumed roughly flat (legacy de-emphasized, offset by gross-margin mix); R&D-revenue durability is the key uncertainty (Merck dependence); opex held near the post-RIF ~$90M/yr run-rate then rising with GMP; dilution ongoing via Cantor ATM ($26.4M remaining). Not logging a Brier forecast (watchlist-loop rule: no forecast.ts create). If one were logged, the scoreable binary is not EPS but: "CDXS announces a GMP-grade ECO Synthesis siRNA supply/tech-transfer agreement with a CDMO or pharma by 2027-06-30" — the catalyst the entire thesis hinges on.
Bull case. Enzymatic siRNA manufacturing is a real, large, fast-growing wave — oligo synthesis $8.9B (2024) → $24.7B by 2030 at ~18.6% CAGR, oligo CDMO $2.55B → $18.37B by 2034 at ~21.8% CAGR, siRNA = 64% of RNAi. Codexis has a patented technical lead (CodeEvolver, ~1,600 patents), demonstrated end-to-end enzymatic synthesis of an approved siRNA, three CDMO partnerships advancing toward tech-transfer in 2026, a $2B-by-2030 platform TAM, a credible CDMO-capture move (Hayward GMP), and — the validation tell — Alnylam committing $250M to add enzymatic ligation capacity, proving the category. Product gross margin hit 71%, burn has been cut by more than half, and a Sierra-Oncology-exit-pedigree chairman is steering. At ~$224M / ~2.7x EV/Sales the platform optionality is cheaply priced if even one siRNA contract scales — and a strategic acquirer (a CDMO or large-cap RNAi player) is a plausible exit.
Bear case (permanent-impairment risks). (1) The recurring business is shrinking — product revenue −29% in FY2025 and management is deliberately de-emphasizing the only profitable line; strip out the $34M Merck license and FY2025 revenue was ~$36M and falling. (2) Customer concentration is existential — one customer = 51% of revenue; lose or fail to renew Merck and the P&L caves. (3) The siRNA bet may be won by the customers themselves — Alnylam adding in-house enzymatic ligation is as much a threat as a validation; big RNAi players and CDMOs (Agilent et al.) can in-source, and incremental SPOS improvements (ligation, liquid-phase, solvent recycling) may close the gap before ECO Synthesis reaches GMP scale. (4) Liquidity cliff: ~$65M cash, ~$13M/qtr burn, a GMP build starting fall 2026, and a $17.6M debt principal due 2027 — the math forces dilution (ATM already running) or a raise into a sub-$3 stock. Pre-mortem (18 months out, thesis broke): the CDMO partnerships stayed in "feasibility" without a signed GMP supply deal, FY2027 revenue missed because Merck's license didn't repeat and no new siRNA contract scaled, the Hayward build ate cash, and the company did a dilutive raise or sold itself cheap — the $0.96 low gets revisited. Are multiples too high? ~2.7x EV/Sales is not demanding if the platform inflects, but it's too high if you value only the durable ~$36M shrinking base (that's <2x on declining revenue with no profits). Contrarian view the market is missing: the bullish "doubling revenue" headline is a Merck-funded mirage over a structurally shrinking enzyme business — the real question isn't growth, it's whether ECO Synthesis reaches paid GMP scale before the 2027 cash/debt pinch forces a bad financing.
What structurally breaks this? Revenue concentration + a one-off masquerading as a trend. A short would argue: the entire FY2025/Q1'26 "turnaround" is the recognition of a single Merck license — non-recurring, balance-sheet-released deferred revenue — laid over a product line that fell 29% and that management is actively abandoning on price. The moat in biocatalysis is so weak the company is exiting it; the moat in siRNA is an unproven-at-GMP-scale technology lead in a category the deep-pocketed incumbents (Alnylam, Agilent, large CDMOs) can fund and in-source faster than a $224M micro-cap with $65M of cash. Most dangerous competitor bulls underestimate: Alnylam itself — by building enzymatic ligation in-house ($250M Norton expansion) the single most important siRNA customer is signaling it may not need Codexis. Worst capital-allocation reality: $615M accumulated deficit, a dead biotherapeutics pipeline divested for milestones-that-may-never-pay, and an open ATM that will dilute holders at depressed prices. Assumptions that must hold for today's ~$2.48: that ≥1 CDMO partnership converts to a paid GMP supply contract in 2026–2027, that Merck-type license revenue recurs, and that the company funds the Hayward build without a punitive raise. If FY2027 revenue disappoints 20–30% (siRNA contracts slip, Merck doesn't repeat), the EV/Sales re-rates against a sub-$40M declining base and, combined with the $17.6M 2027 principal, forces emergency dilution — a plausible path back toward $1 and renewed delisting risk. Single scenario that permanently impairs: the lead CDMO tech-transfer fails technically or commercially and no GMP siRNA revenue materializes by 2028 — at which point the platform thesis is dead and the company is worth only its shrinking enzyme cash flows minus debt. Plausibility: moderate, not remote — the technology works at gram scale; the open question is whether it works and sells at GMP/commercial scale and economics.
A fortress-margin vertical-SaaS monopoly trading at a growth-stock funeral price (~20x forward EPS, near 52-wk lows) because the market is pricing a Salesforce-Agentforce CRM war that threatens the contested ~40% (Commercial) while ignoring the defensible, faster-growing ~60% (R&D/Quality); BULLISH at $153 on a 1–3Y view, but the CRM-migration-to-2030 is a real, watchable execution overhang — not a phantom.
A real, fast-growing oncology-data + diagnostics franchise wrapped in an "AI" narrative it can't yet monetize — own the genomics flywheel, but the round-trip-flavored deals, 30-vote founder, and a CEO famous for cashing out cap the multiple until cash flow turns.
A profitable biosimulation pure-play that the public market broke (Pro-ficiency wrote off ~$72M of a $100M deal, growth fell to single digits) — and is now exiting via a $18.50/sh all-cash Altaris take-private. The fundamental thesis is moot; the only live question is deal-closes-vs-breaks, and this closes (board-unanimous, 16% founder block locked, Q4-2026 close). Merger-arb, not a growth bet.