Jyunmi Hatcher

And in December, Aspuru Guzik's team published a method for giving these robots something they've never had before.

Using simple webcams, the robots can now watch experiments as they happen, observing, for example, whether crystals are forming in reaction plates and respond to what they see.

The robot identifies which experiments produce results worth characterizing and skips the rest.

In Aspuru Guzik's words, with computer eyes, the robot can actually see what happens in the reaction and then act on it.

The commercial world has noticed all of these advancements and the scale of investment signals that self-driving labs are moving from academic curiosity to industrial infrastructure.

Lila Sciences, a startup in Cambridge, Massachusetts, has built what it calls the AI Science Factory, roughly 22,000 square meters of automated lab space designed to provide research and development services to pharmaceutical companies, material science firms, and other research-intensive organizations.

The UK government has already invested in a project using Lila's self-driving robot to synthesize and improve the stability of nanoparticles.

Periodic Labs, launched in 2025 in San Francisco, was co-founded by Liam Feddes, one of the creators of ChatGPT and OpenAI, and Ekendogus Chubuk, who previously led materials and chemistry research at Google DeepMind.

Their automated lab can mix powders, heat them in a furnace, and characterize the products.

The company aims to perform 1,000 experiments per day, but Kubik says success depends not on throughput, but on how effectively the AI analyzes results to guide the next round of experiments.

Pharmaceutical companies are already seeing returns.

Scientists at OpenAI and Ginkgo Bioworks, a biotech company, tested more than 30,000 experimental conditions over six months using a cloud lab integrated with the GPT-5 LLM.

The result...

a 27% improvement in protein yield, and a 40% reduction in cost per gram compared to existing state-of-the-art models.

Novartis, the Swiss pharmaceutical giant, has developed an autonomous platform called MicroCycle that can synthesize, purify, and test compounds, analyze the results, and choose what to make next.

The complete drug discovery cycle, all automated.

the economic argument is becoming harder to ignore.

King estimates his new Genesis robot will be at least an order of magnitude cheaper than human labor for the same experimental output.

The Acceleration Consortium stated goal, compressing materials discovery from 20 years to 100 million, down to one year and one million.

It is ambitious, but the cost curves are real.