Drug discovery costs billions and takes a decade per viable molecule. Most candidates still fail. While AI startups have built better tools for this process, they’ve mostly served researchers who already have the technical skills to use complex software.
SandboxAQ thinks the real problem isn’t the models themselves. It’s making them easy to use. The company has partnered with Anthropic to put its scientific AI directly into Claude, letting researchers run powerful drug discovery simulations through simple conversation instead of specialized computing setups.
What makes SandboxAQ different from other AI drug companies
Founded five years ago as an Alphabet spinout, SandboxAQ has former Google CEO Eric Schmidt as chairman. The company has raised over $950 million and runs several business lines, including cybersecurity.
But its most interesting work involves large quantitative models, or LQMs. Unlike typical AI trained on text patterns, these models are “physics-grounded” – built on actual rules of the physical world. They can run quantum chemistry calculations and simulate how molecules move and react at the atomic level.
This matters because it tells researchers how candidate drugs might behave before anyone touches a test tube.
Targeting the $50 trillion quantitative economy
“Trained on real-world lab data and scientific equations, LQMs are AI models engineered for the quantitative economy, a $50+ trillion sector spanning biopharma, financial services, energy, and advanced materials,” the company said in a release.
That positioning sets SandboxAQ apart from well-funded competitors like Chai Discovery and Isomorphic Labs, which focus on building better scientific models. SandboxAQ cares more about who can actually use the technology.
“For the first time, we have a frontier [quantitative] model on a frontier LLM that someone can access in natural language,” Nadia Harhen, SandboxAQ’s general manager of AI simulation, told TechCrunch.
Solving problems other software can’t handle
Previously, using SandboxAQ’s models required users to provide their own computing infrastructure. The Claude integration removes that barrier.
SandboxAQ’s customers are typically computational scientists, research scientists, or experimentalists at large pharmaceutical or industrial companies. They’re hunting for new materials that can become actual products.
“Our customers come to us because they’ve tried all the other software out there, and the complexity of their problem is such that it didn’t work or didn’t yield positive results for them when that translation went to take place in the real world,” Harhen said.
The move reflects a broader shift in enterprise AI – from building more powerful models to making existing ones more accessible to non-technical users. In drug discovery, where failed experiments cost millions, that accessibility could determine which tools actually get used in the lab.