← Sticker Wall AGENT · DETAIL

Insilico Medicine AI drug discovery pipeline BD collaboration, first-time profitability with $106.3 million revenue in H1

Workflow: The company's proprietary generative AI platform receives target or disease requirements daily, inputs protein structure

AGENT

Key Fields

FIELD STAMPS

IndustrySaaS / Enterprise Software

RegionMulti-region（中国大陆/海外）

ScaleSME

ChannelOnline

🔧 Workflow

The company's proprietary generative AI platform receives target or disease requirements daily, inputs protein structure and bioactivity data, and outputs a list of candidate molecules. Human medicinal chemists screen and score them, and the AI regenerates based on feedback, iterating until a preclinical candidate is selected. For small practitioners, one can focus on segments like target intelligence synthesis or molecular docking scoring as an outsourced service, producing a daily target feasibility report or candidate molecule ranking table for pharmaceutical companies to use in project initiation.

🛠 Setup Requirements

At the corporate level, it is necessary to build large model training and wet-lab platforms. Individual beginners do not need to replicate the entire system; they can start by running a target intelligence summary workflow using public tools. A foundation in medicinal chemistry or bioinformatics is required, and mastering Python to call public databases like PDB and ChEMBL is sufficient to get started. Tool subscriptions are billed based on API calls, and building a minimum viable workflow takes about two to four weeks.

🧰 Toolchain

🔧 Pharma.AI
🔧 Chemistry42
🔧 PandaOmics
🔧 PDB Database
🔧 ChEMBL Database

💰 Revenue

① Partnered pharmaceutical company BD licensing and milestone payments (the company's primary revenue): Pharma companies pay based on upfront licensing fees + R&D milestones + sales royalties. A single milestone of $5 million has been disclosed. The company's total revenue for H1 2026 was $106.3 million, a year-on-year increase of 287.2%, with a gross margin of 90.3%, averaging about $17.72 million per month. Specific revenue-sharing ratios for each line are not public (official company disclosure as of H1 2026); ② Individual/small team replication layer: Delivering target feasibility reports and candidate molecule ranking tables to small pharma or CROs on a per-project or per-task basis. Unit prices are not provided, monthly order volumes cannot be verified, and there is no public data on the scale of this segment (case-based, not externally verified), nor its share of total revenue; ③ Platform and data ecosystem: Breaking down target intelligence, molecular generation, and preclinical assessment into separately sellable service segments, charging pharma companies annual subscriptions or per-project licensing fees. Service pricing is not provided, and no figures are available for this segment's share; ④ Opportunity items—Tripartite collaborations extending into new modalities like in vivo CAR-T; no percentage share provided.

💸 Cost

Corporate-level training and experimental costs are extremely high; individual beginners only need to cover API fees and database subscriptions, approximately 500 to 2,000 RMB per month. If using open-source models run locally, costs can be further reduced to just computing resource expenses.

⏱ Time Investment

The company employs a full-time R&D team; individuals can start by dedicating 10 to 15 hours per week to run a small-scale target intelligence workflow, increasing to over 20 hours per week once orders stabilize.

🚀 Getting Started

The first step for beginners is to register on PandaOmics or a public AI drug platform and run a complete demonstration from disease to candidate molecule for a known target. Organize this into a reusable intelligence report template and send it to small pharma companies or CROs to verify their willingness to pay. Start with one-off paid reports, and once a client base is accumulated, consider subscription models or monthly outsourcing packages.

🔑 Keys to Success

✅ Proprietary data flywheel: Real-world experimental feedback feeds back into model iteration, forming a data moat that competitors cannot easily replicate.
✅ Pipeline value layering: Proprietary candidates combined with BD collaboration revenue sharing, diversifying the risk of failure in any single project.
✅ Clear human oversight: Medicinal chemists perform final screening and set constraints to avoid the 'black box' trust crisis.
✅ Modular output: Breaking down AI drug discovery capabilities into multiple separately sellable service segments such as targets, molecules, and preclinical assessments.

⚠️ 风险

⚠️ Clinical failure of candidates leads to sunk costs, especially given the long investment cycle of proprietary pipelines.
⚠️ Long negotiation cycles for partnerships can easily lead to cash flow breaks for small teams; advance payments or installment settlements are required.
⚠️ Models trained on public data suffer from severe homogenization; if individuals lack exclusive data or unique workflows, they are easily squeezed by price wars.

📌 Real Cases

📌 Insilico Medicine achieved $106.3 million in revenue and first-time profitability in H1 2026, validating the commercial path of AI drug discovery and proving that AI-designed molecules can consistently translate into partnership revenue.
📌 Insilico Medicine reports H1 net profit and significant year-on-year revenue growth, with its AI-training-AI drug discovery super-intelligent system successfully completing 13 out of 31 new drug candidates (42%).
📌 Insilico Medicine invested in NeoX Biotech and partnered with Quantum Sand to form a tripartite collaboration, entering the in vivo CAR-T market, demonstrating the extension of AI drug discovery platforms into new modalities like cell therapy.
📌 Pharma.AI showcased platform upgrades at its Q2 2026 summer webinar, with continuously iterating AI drug design tools becoming a key product carrier for external collaborations.

Sources