Cloud laboratories

01Overview and value chain

Markers: [EC: FDA 21 CFR Part 11 & GxP data integrity | OECD: Biotech & health | Regulator: FDA (USA), EMA (EU), NMPA (China)]

Cloud laboratories are multi-tenant, remotely operated wet-lab facilities that execute user-defined experiments submitted through an API or web portal and return structured, machine-readable data. A protocol written in a high-level language (Strateos SBOL/Protocol, Emerald Cloud Lab Symbolic Language) compiles down to instrument commands run on a shared robotic fleet — liquid handlers, incubators, plate readers, bioreactors, cell sorters — that many users rent by the run rather than own. The model decouples experiment design from physical bench work, lets a biotech run wet-lab campaigns without a lab, and is the substrate for the AI-driven autonomous (self-driving) lab, in which machine-learning planners close the loop between result and next experiment. The platform is mid-transformation along four axes: full-stack automation that takes a single facility from sample-in to structured-data-out with no human intervention between; API-as-a-service bioinformatics that lets computational teams call wet biology like a function; closed-loop autonomous labs (Arctoris Ulysses, XtalPi Science) that design and iterate their own experiments; and cloud-connected bioprocess (Culture Biosciences Nexxys) that extends the model from bench assays to upstream bioreactor operation.

The key directions of cloud laboratories are:

1. API-driven full-stack wet labs (Strateos, Emerald Cloud Lab): facilities whose entire protocol — from cell thaw to assay readout — is specified in code and executed on a robotic fleet, returning structured JSON data via API.
2. Closed-loop autonomous drug-discovery labs (Arctoris Ulysses, XtalPi Science): platforms that couple robotic execution with machine-learning planning to design, run and interpret their own assay cycles, with AI agents (XtalPi Genius Agents) driving the experiment loop.
3. Cloud-connected bioprocess (Culture Biosciences Nexxys): extending the cloud-lab model from bench assays to upstream bioreactor operation, letting process-development teams run parallel fermentation campaigns remotely.
4. Protocol-as-code and reproducibility (Symbolic Language, SBOL): unambiguous, version-controlled protocol descriptions that make every run exactly repeatable across sites — the software-engineering discipline that cloud labs bring to wet biology.

Sectoral value chain

[protocol-as-code] ──> [compile to instrument commands] ──> [robotic execution on shared fleet]
                                                            │
                                                  (inline analytics + structured data)
                                                            │
                                                            ▼
[AI planning / next cycle] <─── [structured JSON results] <─────┘

Value chain levels

Cross-cutting technologies of the sector:

• Cloud laboratory (Cloud Laboratory): remotely operated, API-driven, multi-tenant wet-lab facility returning structured data.
• Robotic autonomous lab (Robotic Autonomous Lab): closed-loop platform coupling robotics, analytics and ML planning for self-driving experiment cycles.
• High-throughput screening (High-Throughput Screening): the assay layer whose parallelisation makes cloud labs economical.

02US

The United States is the centre of the cloud-laboratory category: three of the four leading full-stack platforms are US-based. Strateos (San Diego, formerly Transcriptic) pioneered the API-driven wet lab; Emerald Cloud Lab (Austin, TX and South San Francisco) operates the most fully automated single facility; Culture Biosciences (South San Francisco) leads the cloud-connected bioprocess segment. FDA’s 21 CFR Part 11 data-integrity framework is what qualifies cloud-lab output for regulated pharmaceutical work.

API-driven wet labs, cloud bioprocess, FDA data integrity

• Strateos: the API-driven cloud lab with the Science Board marketplace; compiles user protocols to instrument commands on a shared robotic fleet.
• Emerald Cloud Lab: the most fully automated single facility (ECL1), executing end-to-end protocols in the Emerald Cloud Lab Symbolic Language.
• Culture Biosciences — Nexxys: cloud-connected bioprocess platform letting process-development teams run parallel fermentation campaigns remotely.
• FDA 21 CFR Part 11: electronic-records and audit-trail standard that qualifies cloud-lab output for GxP pharmaceutical R&D.

03CN

China is the demand-side growth market and home to the leading AI-integrated autonomous lab: XtalPi (晶泰科技, HKEX:2228, Shenzhen and Cambridge, MA) couples its XtalPi Science platform — Genius Agents (AI agents addressing long-workflow hallucination) and a physical autonomous lab — to close the loop between computational drug design and robotic wet-lab validation. Domestic pharmaceutical and CRO demand for remote, reproducible wet-lab capacity is the growth driver, with NMPA ICH-aligned data-integrity expectations framing the regulated route.

AI-integrated autonomous lab, pharmaceutical R&D demand, NMPA data integrity

• XtalPi (晶泰科技): XtalPi Science platform coupling Genius Agents (AI planning) with a physical autonomous lab for closed-loop drug-discovery validation, HKEX:2228.
• Domestic pharmaceutical R&D demand: biotech and CRO adoption of remote, reproducible wet-lab capacity under cost and reproducibility pressure.
• NMPA ICH-aligned data integrity: harmonised 21 CFR Part 11-equivalent electronic-records expectations qualifying cloud-lab output for regulated Chinese pharmaceutical R&D.

04EU

Europe anchors the closed-loop autonomous drug-discovery segment: Arctoris (London, UK, with US operations) operates the Ulysses platform, which couples closed-loop robotics and machine learning to automate end-to-end drug-discovery workflows. EMA’s GxP and ICH-aligned data-integrity framework define the regulated route for cloud-lab output in European pharmaceutical R&D.

closed-loop autonomous labs, EMA GxP data integrity

• Arctoris — Ulysses platform: closed-loop robotics + machine learning automating end-to-end drug-discovery workflows, with Boston and London operations.
• EMA GxP & ICH data integrity: the EU electronic-records and audit-trail framework qualifying cloud-lab output for regulated pharmaceutical R&D.

05Leading companies and research institutes

06Tech stack and innovations

The stack combines protocol-as-code, a shared robotic fleet, structured data return and an ML planning layer into one programmable substrate.

1. Protocol-as-code (Protocol-as-Code):
   • high-level symbolic languages (Emerald Cloud Lab Symbolic Language, SBOL, Strateos Protocol) that compile unambiguously to instrument commands;
   • version-controlled protocols make every run exactly repeatable across sites and time.
2. Shared robotic fleet (Shared Robotic Fleet):
   • liquid handlers, incubators, plate readers, cell sorters and bioreactors scheduled across many tenants;
   • full-stack automation (Emerald ECL1) takes a facility from sample-in to structured-data-out with no human intervention between.
3. Structured-data return (Structured-Data Return):
   • raw assay data normalised to structured JSON and delivered via API or dashboard;
   • enables computational teams to call wet biology as a function and feed results straight into ML pipelines.
4. ML planning & the autonomous loop (ML Planning & Autonomous Loop):
   • machine-learning agents (Arctoris Ulysses, XtalPi Genius Agents) interpret results and propose the next run, closing the self-driving lab loop.

07Value chains and production pipelines

Industrial pipeline of a cloud-laboratory run (21 CFR Part 11)

┌───────────────────────────┐      ┌───────────────────────────┐
│ 1. Protocol-as-code       │ ───> │ 2. Compile & schedule     │
└───────────────────────────┘      └───────────────────────────┘
                                                  │
                                                  ▼
┌───────────────────────────┐      ┌───────────────────────────┐
│ 4. Structured JSON return │ <─── │ 3. Robotic execution      │
└───────────────────────────┘      └───────────────────────────┘
              │
              ▼
┌───────────────────────────┐      ┌───────────────────────────┐
│ 5. ML planning (next run) │ ───> │ 6. GxP archive & audit    │
└───────────────────────────┘      └───────────────────────────┘

Stage 1: Protocol-as-code

The user writes the experiment in a high-level symbolic language (SBOL, Emerald Symbolic Language, Strateos Protocol) that specifies every step, reagent and instrument setting unambiguously.

Stage 2: Compile and schedule

The protocol compiles to instrument commands and is scheduled onto the shared robotic fleet, queued alongside other tenants’ runs for capacity-optimal execution.

Stage 3: Robotic execution

Liquid handlers, incubators, plate readers, cell sorters or bioreactors execute the run with no human intervention between steps, capturing every action in an electronic run log.

Stage 4: Structured-data return

Inline analytics (plate readers, cell counters, mass spec, bioprocess sensors) generate raw data that is normalised to structured JSON and delivered via API or dashboard.

Stage 5: ML planning

For autonomous platforms, a machine-learning agent (Arctoris Ulysses, XtalPi Genius Agents) interprets the structured results and proposes the next run, closing the self-driving lab loop.

Stage 6: GxP archive and audit

Every run’s electronic records, instrument telemetry and full provenance are archived under 21 CFR Part 11, producing a traceable, release-ready data package for regulated work.