As of June 2026, Lila is in talks to raise approximately $2 billion in a Series B at a pre-money valuation of roughly $8.5 billion, implying a post-money valuation of approximately $10.5 billion. The round is being anchored by the California Public Employees' Retirement System (CalPERS) and NVentures. Terms are not final.

Lila's prior round was a Series A closed at $350M in October 2025, valuing the company above $1.3 billion. The Series A was announced in two tranches: an initial $235M in September 2025 with investors including Flagship Pioneering, General Catalyst, March Capital, ARK Venture Fund, Altitude Life Science Ventures, and Abu Dhabi Investment Authority, followed by a $115M extension in October 2025 with new participants including Braidwell, Collective Global, NVentures, Analog Devices, IQT, Dauntless Ventures, Catalio Capital Management, and Pennant Investors.

Before the Series A, Lila raised $200M in committed seed capital at its March 2025 unveiling, led by Flagship Pioneering. Total disclosed funding across seed and Series A is $550M, before any close of the reported Series B.

Lila is building an autonomous science platform that takes a research problem, generates hypotheses, designs experiments, runs those experiments in a physical robotic lab, reads the results, and decides what to try next without a human scientist in the loop for each iteration.

At the core is a proprietary scientific reasoning model called Lila Iris. Rather than only predicting which molecule or material might work from existing data, it proposes experiment plans, selects protocols, and directs robotic instruments to execute those plans in real time. The physical execution layer is what Lila calls an AI Science Factory (AISF), a facility where robotic arms, mobile lab robots, screening stations, imaging systems, and analytical instruments are orchestrated by AI.

A typical workflow starts when a customer brings a problem such as optimizing an antibody for binding and manufacturability, finding a catalyst with better selectivity under commercial reactor conditions, or discovering a corrosion-resistant coating for extreme environments. Lila Iris proposes candidate designs and experiment formats, the AISF runs those experiments, and the results come back as both scientific data and what Lila calls scientific tokens, structured outputs that feed directly into the next training cycle for Lila Iris. The system then proposes the next round and repeats the loop.

By the time of its Series A in October 2025, Lila's first AISF had already closed the loop across hundreds of thousands of AI-driven experiments. As of mid-2026, a 200,000-square-foot AISF facility in Cambridge was under construction, with a planned Q3 2026 opening.

Lila sells the platform through two offerings. Catalyst gives external R&D teams access to Lila Iris and AISF capacity to accelerate their own programs, positioning the product as on-demand scientific infrastructure rather than a fixed internal lab. Creation is a more outcome-oriented engagement in which Lila runs a focused discovery campaign end-to-end and delivers validated molecules, materials, protocols, and data packages, in some cases including the foundation for a new company or platform built around the resulting IP. The platform spans therapeutics, biotech, chemicals, energy and environment, advanced materials, oil and gas, and aerospace and defense.

Lila uses a B2B model aimed at large organizations with expensive, slow R&D cycles, including pharma, chemicals, industrials, energy producers, and defense-linked programs. The value proposition is a shorter design-test-learn loop in domains where each experimental iteration is costly enough that faster iteration has large economic value.

Monetization runs through two lanes. Catalyst charges for platform access and Lab-as-a-Service capacity, where customers pay to use Lila Iris and AISF infrastructure on an on-demand basis rather than building their own autonomous lab stack. Creation charges for end-to-end discovery campaigns, with economics likely structured around program fees, milestone deliveries, and, in some cases, IP participation or equity in new companies formed around the resulting assets.

The cost structure differs from software-first peers. Lila owns and operates physical AI Science Factories, employs specialized scientific staff, and runs frontier model training and inference at scale. Fixed costs are high and gross margins are likely well below what a pure SaaS business would carry, closer to a capital-intensive research platform than a software company. The strategic logic is a self-reinforcing flywheel: more customer programs generate more experiments, more experiments produce proprietary scientific tokens, those tokens improve Lila Iris, and a better model attracts more customers and justifies more AISF buildout. Each commercial engagement serves two functions, generating revenue while improving the underlying platform, which differs structurally from a contract research organization, where each project ends without compounding the platform's intelligence.

The go-to-market motion is likely land-and-expand: start with one program or business unit, prove faster cycle times and better hit rates, then extend into adjacent workflows or domains. The Catalyst and Creation split broadens the buyer set, some customers want to accelerate their own programs, while others want to outsource the discovery loop entirely, without requiring separate technology stacks.

The competitive field for autonomous science platforms is stratifying across three layers: scientific reasoning models, lab orchestration software, and physical autonomous lab capacity. Lila's strategy is to own all three as a unified stack, which is more ambitious than most rivals and requires execution across more dimensions at once.

The closest structural analog to Lila is Atinary, which markets self-driving labs that combine AI, robotics, and closed-loop experimentation. Atinary's March 2026 Boston lab launch pushed it beyond software into physical lab ownership, integrating hardware from ABB, Agilent, Bruker, Chemspeed, and Mettler-Toledo.

Atinary is more open and partner-centric, with an emphasis on no-code deployment and hardware interoperability, while Lila is more vertically integrated around its own proprietary model and facilities. That openness can appeal to enterprises that want autonomous lab capability without committing to a single external platform.

Strateos and Emerald Cloud Lab compete on the lab-as-a-service axis. Strateos offers cloud-accessible robotic labs and private lab control software deployable into customer-owned environments, while Emerald Cloud Lab runs a remote facility with over 200 instrument models under a single software interface. Both are more operationally familiar to enterprise buyers than Lila, but neither combines a frontier scientific reasoning model with the physical execution layer.

Isomorphic Labs is a meaningful threat in drug discovery despite being narrower than Lila. Its Drug Design Engine and expanded partnerships with Novartis and Eli Lilly point to a strategy centered on computational molecule design rather than general autonomous experimentation. If pharma buyers decide that model quality at the design layer matters more than owning a general-purpose scientific OS, Isomorphic can win high-value budgets without matching Lila's robotics footprint.

Recursion, which combined with Exscientia in late 2024, now presents a vertically integrated TechBio platform spanning large-scale automated wet labs, multimodal foundation models, and one of the largest compute footprints in the sector. Relative to Lila, Recursion is more therapeutics-centric and more clinically grounded, with an established track record among biotech buyers.

Insilico Medicine uses automation-generated data to improve its models but is optimized for therapeutics rather than cross-domain science. LabGenius takes a similar closed-loop ML-plus-robotic-experimentation approach but focuses more narrowly on antibody and nanobody optimization, making it a specialist threat in that workflow.

Bruker's combination of Chemspeed and SciY, announced in February 2026, launched an open self-driving lab platform that combines robotics, analytics, data infrastructure, and AI orchestration on a vendor-agnostic backbone. For Lila, this is the clearest case of an incumbent broadening into its territory: Bruker can sell autonomous-lab infrastructure into existing enterprise environments with the compliance posture, installed base, and enterprise trust of a major instrumentation company.

HighRes Biosolutions markets CellarioOS as a lab orchestration platform connecting instruments, protocols, and software across labs and sites. Its 2026 partnership with Opentrons to enable agent-to-agent workflows moves it up-stack toward the control plane of automated labs while leaving AI model choice open to the customer, a modular approach that can capture the orchestration layer without requiring customers to commit to a single proprietary science platform.

Benchling is not a direct competitor today, but it is relevant to watch as it expands from a system of record and collaboration layer for biotech R&D into experiment orchestration. If it continues moving from documentation into execution, it could become more relevant in the same enterprise R&D accounts Lila is targeting.

Lila's TAM expansion runs in two directions: broadening the scientific domains it serves and increasing the value captured within each engagement, from platform access to asset creation and company formation.

The Catalyst and Creation split expands TAM in two ways. Catalyst packages Lila's AISF infrastructure as an on-demand resource that R&D organizations can use without building their own autonomous lab stack, extending the addressable market to incumbents that want faster experimentation without a greenfield infrastructure investment.

Creation expands monetization by delivering validated assets, IP, and de-risked technical roadmaps rather than only model outputs or lab throughput. That creates a path to participate in discovery economics through milestone economics, IP licensing, or equity in new companies formed around the resulting science, rather than relying only on access or services revenue.

Lila's initial positioning spanned life sciences, chemistry, and materials, but by 2026 its industry coverage extends to oil and gas, energy and environment, advanced materials, biotech, and aerospace and defense. That broadens the buyer base from pharma and biotech R&D teams to industrial manufacturers, energy producers, defense primes, and specialty chemical companies.

Within biology, Lila is moving up the stack from discovery into adjacent workflows such as gene products, bioprocessing, reagents, and assays under real manufacturing constraints. This extends the platform from candidate identification to manufacturability and reproducibility, where development bottlenecks often emerge. Defense and national security are a separate expansion vector: Lila's Series A investor commentary tied the platform to U.S. resilience across materials discovery for computing, energy, and infrastructure, and that customer set can support premium budgets and long-term partnerships without requiring a separate scientific stack.

Lila has announced hubs in Boston, San Francisco, and London, and its Series A framing referenced global growth plans. Geographic presence matters because Lila's delivery model depends on physical AI Science Factories rather than cloud-only software.

The planned 200,000-square-foot Cambridge AISF, targeted for Q3 2026 opening, is the clearest near-term capacity lever. More factory capacity means more experimental bandwidth, more scientific tokens generated per unit time, and a faster-improving Lila Iris, which supports higher customer throughput and broader cross-domain model performance. Investor participation from NVentures and Analog Devices also points to a potential partnership channel around compute and instrumentation that could improve AISF deployment economics and lower per-experiment cost as the network scales.

Capital intensity: Lila's moat depends on scaling physical AI Science Factories, robotic instrumentation, and frontier model training at the same time, and if customer conversion or utilization lags facility buildout, the company could carry a heavy fixed-cost base for an extended period before its data flywheel and commercial model reach self-sustaining scale.

Generalist tradeoff: Lila's core strategic bet is that one general platform can outperform narrower domain-specific tools across therapeutics, materials, chemicals, and energy, but focused rivals like Isomorphic Labs in drug design, LabGenius in antibody optimization, and Atinary in chemistry and catalysis can still win the highest-value subsegments by going deeper on domain-specific data, workflows, and buyer trust before Lila's cross-domain transfer learning proves an advantage.

Trust bottleneck: Lila sells into high-stakes, often regulated domains where IP ownership, data security, scientific reproducibility, and compliance shape buying decisions, which means commercial adoption may depend less on model quality alone and more on whether large pharma, industrial, and defense customers will let an external autonomous system design and execute consequential experiments inside sensitive programs.