ins-$1.2M · 13F 13+/10-×0.0 · short↓0.22
Current snapshotPublic companyResearch currentFresh
Schrodinger
AI Discover · Healthspan Infrastructure
SignalHOLDReviewed 25 Oct 2026
AI Discover exposure in Healthspan Infrastructure.
AI Discover / Physics-based drug discovery: In the AI-drug-discovery stack, physics-based simulation is the *ground-truth training-data layer* that pure ML "black-box" models lack. Schrödinger is one of the only companies with 30+ years of validated quantum/statistical-mechanics methods (FEP+) that compute molecular properties from first principles rather than fitting them from possibly-hallucinated data.
Catalysts
Schrodinger Q2 2026 earnings; FEP+ / computational platform customer metrics; pipeline (MALT1 inhibitor, others); pharma partnership revenue
Invalidation
Thesis breaks if software revenue keeps declining (-10% YoY, hosted transition masks weak spot demand), op loss $41.5M/quarter persists to 2028, or OpenEye/XtalPi/Recursion outcompete the physics+AI platform. (vault 2026-08)
Latest change
· Target moved from $18.33 to $18.29.
Capital in context
Mapped funds and investors
Direct relationships supported by published evidence.
Reported holdings as-of
| Fund / investor | As of | Fund portfolio weight | Evidence |
|---|---|---|---|
| ARK Genomic Revolution ETF ARKG | 2.67% | Source |
Disclosed investment history
No sourced relationships mapped yet.
Showing up to four relationships in each group. ETF weights describe the fund’s portfolio, not ownership of this company. Private investment history does not establish current ownership.
Neutral — No investor/analyst mentions of the company found. [X search Aug 2026]
Lower-confidence snapshotTarget price: $29.05
Dated research targetMoat & position
- Evidence Grade
- Inferred
- Moat Durability
- Capacity cycle
- Toll Road Tier
- Strong proxy
Timestamped research update
Deep research update
2,296 words · Updated 19 Oct 2026 · 7 sources
Schrödinger sells a physics-based molecular simulation software platform (FEP+, Glide, Maestro, LiveDesign) plus an agentic AI co-scientist layer ("Bunsen") used by biopharma to predict binding affinity, optimize lead molecules, and de-risk drug discovery before wet-lab synthesis. It also runs a partnered drug-discovery business (milestones + royalties) and a small grant-funded "contribution" line.
- Total revenue $58.9M (+8% YoY); H1 2026 $117.5M; FY2025 $255.9M
- Software revenue $32.5M (−10% YoY); Drug discovery $23.0M (+65%); Contribution $3.4M (−29%)
- Software gross margin 71% (down from ~80%); blended gross profit $32.4M
- Operating loss −$41.5M (Q2); R&D $41.0M; OpEx $74.0M (−6% YoY)
- Net income +$6.0M (Q2, inflated by $48.9M one-time Ajax/Lilly equity mark-to-market gain; operating line still deeply negative)
- Cash + marketable securities $418.8M
- FY2026 guidance: ACV $218–228M (+10–15%); drug discovery revenue $65–75M (raised from $55–65M); OpEx < 2025; Q3 ACV (ex-contribution) $41–45M
- monetization model: Software subscriptions (term licenses/SaaS-like) plus drug-discovery collaborations: research funding, milestones, royalties/licensing. (10-K FY2024, Business; sec.gov ; accession see EDGAR index)
- revenue mix: FY2024 total $198.1M: Software $159.1M (~80%); Drug discovery $39.0M (~20%). (10-K FY2024, revenue/segment note)
- who pays: Pharma/biotech R&D budgets and academic/government grants for software; pharma/biotech collaborators for drug discovery. (10-K FY2024, Business)
- margin_value_capture: Software carries high gross margin and recurring revenue; drug discovery is lower-margin/milestone-based; overall GAAP operating margin negative. Value accrues via software subscriptions and downstream milestones/royalties. (10-K FY2024, MD&A/Risk Factors)
- customer classes: Pharma & Biotech, academic/government research institutions, materials/chemicals companies. (10-K FY2024, Business)
Schrödinger, Inc. operates a physics-based computational platform for molecular design and is organized around two reportable segments according to its SEC filings:
- Software segment — commercial licenses, hosted/SaaS software, maintenance, and professional services for computational chemistry, cheminformatics, and materials science.
- Drug discovery segment — collaboration revenue, milestone payments, preclinical/clinical program economics, and equity interests in co-founded or partnered drug discovery companies.
The exact revenue split by segment for FY2023 or TTM must be pulled from the most recent 10-K/10-Q segment footnote. Revenue mix by segment (%) needs primary source validation. I will not state a hard percentage here because the SEC segment note is the only acceptable source under the requested Tier 1 standard.
SDGR’s monetization model is hybrid, and the following structure is disclosed across its filings and investor materials:
- Software licenses and subscriptions: customers pay annual license or subscription fees for access to Schrödinger’s molecular modeling platform, including products such as Maestro, FEP+, Desmond, and LiveDesign.
- Hosted/cloud usage: enterprise customers can access the platform through hosted environments, which may involve usage-based or capacity-based fees.
- Drug discovery collaborations: biopharma partners pay upfront amounts, research funding, and success-based milestones; SDGR may also receive royalties on products developed through the collaboration.
- Equity stakes in co-founded companies: SDGR has historically used its platform to co-found or seed drug discovery companies and may hold minority equity interests.
The precise revenue recognition policy for each stream — including whether milestones are recognized point-in-time or over time — needs primary source validation from the 10-K Revenue Recognition footnote.
Because no reliable seat-level or contract-level data is available in this session, I provide an illustrative software unit economics model rather than a reported SDGR metric. This is not a valuation input and must not be confused with SDGR’s reported financials.
Based on SDGR’s public disclosures and SEC filings, customer classes are:
- Large pharmaceutical companies — buy enterprise licenses or hosted deployments for computational and medicinal chemistry groups.
- Biotechnology companies — often use the platform for early drug discovery programs and may enter drug discovery collaborations.
- Academic institutions — use licensed software for research and teaching, often under academic pricing models.
- Materials science and chemicals companies — use the Materials Science Suite for polymers, catalysts, semiconductors, and specialty chemicals.
- Government laboratories — use the platform for fundamental research or applied materials/drug work.
- Computational chemistry leader — owns software evaluation, technical validation, and workflow integration.
- Head of medicinal chemistry — cares about design cycle time and candidate quality.
- Chief scientific officer or head of discovery — approves multi-year enterprise deployments or collaboration agreements.
- CIO/IT procurement — negotiates hosted/SaaS terms and security.
- Institutional technology transfer office — for academic licenses.
I cannot cite a named customer from an SEC filing without direct retrieval. The customer-concentration footnote in the 10-K may or may not name a revenue concentration. Any named customer claim must be sourced from the specific 10-K exhibit or concentration note. That claim is currently needs primary source validation.
- Cost-of-moat = R&D intensity. Q2 2026 R&D expense $41.0M on $58.9M revenue = 69.6% of revenue spent defending the physics lead (FY2025: $173.1M / $255.9M = 67.7%). Schrödinger must out-spend AI-native competitors on method science to stay ahead — this is the economic limit on how fast it can reach profitability.
- SaaS-transition gross-margin trade. Software GM compressed to 71% in Q2 2026 (from ~80% historical perpetual) because hosted revenue (47% of software revenue, up from 23% late-2025) is recognized ratably. The transition forfeits ~9pts of GM (~$2.9M/quarter on $32.5M software revenue) in exchange for recurring, compute-metered revenue that scales with Bunsen throughput.
- Bookings-vs-recognition decoupling. Q2 ACV $29.6M (+27% YoY) vs software revenue $32.5M (−10% YoY) — the hosted shift makes bookings (ACV) the forward indicator while revenue optically declines. ACV is the honest demand signal and it is accelerating.
- Runway. Cash + equivalents + restricted cash + marketable securities $418.8M vs Q2 operating loss $41.5M → ~2.5 years of runway at current burn (annualized ~$166M). Management guides to positive adjusted EBITDA by 2028, so the bridge requires ACV compounding to out-grow R&D.
- Milestone-as-free-option economics. Drug discovery revenue $23.0M (+65% YoY) was driven by a $10M Ajax Therapeutics collaboration milestone (Lilly acquired Ajax; Schrödinger held equity). $10M ≈ 17% of total Q2 revenue, captured with zero clinical-trial capex — the partnered model is the capital-light upside.
In the AI-drug-discovery stack, physics-based simulation is the ground-truth training-data layer that pure ML "black-box" models lack. Schrödinger is one of the only companies with 30+ years of validated quantum/statistical-mechanics methods (FEP+) that compute molecular properties from first principles rather than fitting them from possibly-hallucinated data. The CEO's own framing (Q2 2026 call): "the need to generate ground-truth data has never been greater" — Schrödinger positions itself as the definitive computational infrastructure that constrains and trains biopharma AI.
- FEP+ accuracy at scale, independently validated. Q2 2026 8-K EX-99.1 reports a Journal of Chemical Information and Modeling publication validating FEP+ for binding-affinity prediction and macrocycle/cyclic-peptide optimization across 230+ unique compounds on 5 diverse, challenging targets — the explicit claim is that FEP+ lets teams "bypass low-probability chemical synthesis, substantially compress development timelines, and drive cost-effective pipeline advancements for historically undruggable targets."
- Hyperscaler + GPU endorsement. NVIDIA and Google Cloud will provide a full-stack AI platform, infrastructure, and NVIDIA BioNeMo Agent Toolkit access for early Bunsen customers (EX-99.1) — the physics platform is being positioned as the compute substrate for biopharma AI workloads.
- Big-pharma lock-in signal. Bristol Myers Squibb signed a strategic software agreement to deploy Bunsen across its research organization, and BMS will co-develop novel functionality with Schrödinger (EX-99.1) — a named-account expansion, not a point license.
- Switching cost = method revalidation. Adopting FEP+ is not a tool swap; it requires re-running validation protocols (binding-affinity benchmarks against internal assay data) across a discovery org. A 230-compound/5-target validation cycle is the floor cost of re-proving any competing force field.
- product_service: Schrödinger computational platform for physics-based molecular simulation and drug discovery, plus collaborations. (10-K FY2024, Business)
- technical function: Predicts molecular properties/designs using quantum mechanics, molecular dynamics, free-energy perturbation, and ML; integrated tools such as Maestro, Glide, Desmond. (10-K FY2024, Business)
- why it is hard: Decades of validated physics methods, proprietary force fields, curated data, compute infrastructure, and high switching costs in R&D workflows. (10-K FY2024, Business/Risk Factors)
- substitutes: In-house computational chemistry, open-source MD/QM tools, other CADD software (BIOVIA, OpenEye, Cresset, MOE), AI drug-discovery platforms, CROs. (10-K FY2024, Risk Factors/Competition)
- pricing power: MED — software subscriptions are embedded/high-retention, but budgets/competition cap pricing; drug-discovery economics UNPROVEN. (10-K FY2024, Risk Factors)
Schrödinger sells a computational chemistry platform that combines physics-based molecular simulation with enterprise cheminformatics. The platform’s best-known commercial components include:
- Maestro — molecular modeling and visualization interface.
- FEP+ — free-energy perturbation calculations for predicting protein–ligand binding affinity.
- Desmond — molecular dynamics engine.
- LiveDesign — web-based enterprise informatics and collaborative molecular design environment.
- Materials Science Suite — tools for materials design, including polymers, catalysts, and electronic materials.
These product names are drawn from the company’s own platform descriptions and are Tier 2 context, not SEC financial data.
The platform applies physics-based force fields and molecular dynamics to predict how molecules interact with biological targets such as proteins. Instead of relying solely on statistical or machine-learning patterns, SDGR’s core approach uses physics-based simulations to estimate binding affinity, solubility, permeability, and other molecular properties. FEP+ specifically calculates relative or absolute binding free energies, which medicinal chemists use to rank virtual compounds before synthesis.
This is complementary to, and partially substitutable with, machine-learning and generative-AI drug discovery tools. Many competitors now combine physics-based methods with neural networks and large datasets, so the technical frontier is contested.
Substitutes include:
- Open-source cheminformatics and molecular dynamics tools — lower cost but generally less integrated and less supported at enterprise scale.
- Commercial competitors — OpenEye Scientific (acquired by Cadence Design Systems), Chemical Computing Group, Dassault Systèmes BIOVIA, and academic-origin packages.
- AI-native drug discovery platforms — XtalPi, Insilico Medicine, Genesis Therapeutics, and similar players that emphasize machine-learning approaches.
- Internal bespoke platforms at very large pharma companies that build or assemble their own computational tools.
The platform’s defensibility relies on:
- Long development history and accumulated validation across targets and chemical space.
- Physics-based accuracy in free-energy predictions, which requires specialized force fields, simulation protocols, and high-performance computing infrastructure.
- Workflow integration into medicinal chemistry decision-making, creating switching costs.
- Enterprise collaboration tooling through LiveDesign, which embeds the software into multi-team workflows.
These are qualitative moat factors. No moat claim should be treated as quantitative unless supported by customer-retention or gross-dollar-retention data from SEC filings, which needs primary source validation.
Software pricing power is theoretically present if the platform shortens cycle times and reduces synthesis spend. However, pricing power depends on:
- Competitive pressure from AI-native tools,
- Budget cycles at large pharma,
- The demonstration of return on investment within customer organizations,
- The mix of named-user vs. enterprise licenses.
Whether SDGR can raise prices without losing share needs primary source validation from revenue retention, renewal rate, or average selling price disclosures. The 10-K may disclose volume/price impacts only qualitatively, so external evidence may be required.
- Software revenue is declining (−10% YoY) — the hosted transition masks weak spot demand and means reported revenue understates nothing; it is genuinely the number falling while ACV (bookings) rises.
- Operating loss $41.5M/quarter is structural until ~2028 — this is a show-me profitability story, not a cash-flow compounder today.
- The Q2 "profit" is an artifact — $6.0M net income came from a $48.9M one-time equity mark-to-market on Ajax/Lilly, not operations.
- Competitive encroachment. OpenEye (now Cadence) sells a competing physics-based force field; XtalPi fuses physics + AI; Recursion and Absci attack from the pure-AI/phenomics side. The physics moat is real but not a monopoly, and AI-native approaches are converging on the accuracy bar.
- Contribution revenue is grant-funded and lumpy (Gates Foundation toxicology + battery grants, −29% YoY) — not a durable franchise line.
software revenue −10% YoY (hosted-transition mask); operating loss $41.5M/quarter structural until ~2028; Q2 net income $6.0M is a $48.9M Ajax/Lilly mark-to-market artifact, not operations.
Verdict: STRENGTHENED (unchanged). No new catalyst to re-rate; this update closes the toll-road + scoring-writeback gap on an already-differentiated note.
The SDGR thesis would be wrong or broken if any of the following were confirmed in primary filings:
- Software segment growth stalls or declines — especially if license or hosted revenue fails to grow while R&D spend continues to rise.
- Customer concentration deepens — if a single customer accounts for a material portion of revenue and that customer fails to renew or expands a competing internal platform.
- Gross margin falls materially — either in software, indicating pricing pressure or cloud/royalty cost inflation, or in drug discovery, indicating uneconomic collaboration terms.
- Internal drug discovery pipeline fails — if lead programs such as MALT1, CDC7, or Wee1/Myt1 inhibitors produce negative clinical or preclinical data, the market may recalibrate the company’s optionality.
- Competitors demonstrate faster time-to-value — if AI-native platforms show comparable binding-affinity prediction accuracy at lower cost, SDGR’s software renewals and pricing power would be threatened.
- Collaboration economics deteriorate — if new partnerships show lower upfronts, lower milestones, or reduced royalties, the drug discovery segment’s value would be impaired.
- Cash burn accelerates without commercial inflection — if operating cash outflows rise with no corresponding software growth or milestone income, the balance-sheet cushion becomes a bear case.
Any of these disconfirming signals must be confirmed against SEC filings before changing conviction.
Primary and supporting material
Sources
7 sources preserved from the latest qualified research update.
- sec.govSchrodinger FY2025 Form 10 KOpen source ↗
- sec.govSchrodinger Q2 2026 earnings release, Exhibit 99.1Open source ↗
- sec.govsec.govOpen source ↗
- data.sec.govdata.sec.govOpen source ↗
- ir.schrodinger.comir.schrodinger.comOpen source ↗
- schrodinger.comschrodinger.comOpen source ↗
- sec.govsec.govOpen source ↗