Underutilization Risk from Facility Buildout

This risk is really about whether Lila fills expensive lab capacity fast enough to turn a giant science factory into a revenue producing machine instead of an idle asset. Lila is not selling software with near zero marginal cost. It is building and operating robotic labs, training frontier models, and staffing specialized scientific workflows, while also expanding into a 200,000 square foot Cambridge facility. That means rent, equipment depreciation, compute, and labor keep running even if customer programs ramp slowly.1 2

• Lila has already raised $550M across seed and Series A to scale AI Science Factories, which shows how much capital the model needs before it reaches steady utilization. Its own business model depends on more experiments generating more data, and more data improving the model, so underused capacity slows both revenue and product learning at the same time.1 2
• Comparable lab infrastructure businesses show why utilization matters. Emerald Cloud Lab markets access to a remote lab with more than 200 instrument models, while Strateos and Atinary also compete by turning robotic capacity into on demand R&D throughput. In this model, empty instrument time is lost inventory, not a deferred software sale.1 3 4
• Lila partly offsets this risk by selling two ways. Catalyst lets customers buy access to the platform, while Creation lets Lila run full discovery programs and capture program fees, milestones, IP participation, or equity. That broadens demand sources, but both offers still require a steady flow of paid experimental work to absorb fixed factory costs.1

The next phase is about turning facility scale into dense program volume. If Lila can keep adding pharma, chemicals, materials, energy, and defense workloads into the same physical network, fixed costs start spreading across more experiments and the data flywheel gets stronger. If that happens, the factory becomes the moat rather than the burden.1 2