Roche speeds up drug development and doubles down on Nvidia

Pharmaceutical giant Roche said today it is launching a new large-scale “AI factory” with thousands of the latest Nvidia chips to accelerate the development of new drugs and diagnostics.

Roche said the company has ramped up its Artificial Intelligence (AI) capacity by purchasing 2,176 Nvidia Blackwell GPUs, which will be deployed across the US and Europe. That brings its total GPUs to more than 3,500, which it claims is the greatest number owned by any pharmaceutical company. Roche, which made the announcement in conjunction with Nvidia’s GTC conference in San Jose, hopes to use that technological firepower to discover, develop, manufacture and commercialize therapies faster.

While Roche, which has a market cap of $340 billion, is based in Switzerland, its US subsidiary Genentech is well known for developing drugs to treat cancer and immune system diseases, including Herceptin (for breast cancer) and Avastin (for colon and lung cancer). Roche brought in $78 billion in revenue last year at current exchange rates.

The company declined to disclose how much it is spending on this effort, but it is “a substantial sum,” says Aviv Regev, executive vice president and head of Genentech research and early development. “It does highlight how important AI is for our business.”

The Future of Healthcare, by Keynote Speaker Matthew Griffin

Nvidia CEO Jensen Huang has said its Blackwell chips cost between $30,000 and $40,000 a pop, which would place the company’s investment at $65 million or more for the chips.

Roche’s announcement follows that of Eli Lilly, which earlier this year partnered with Nvidia on the creation of a $1 billion AI innovation lab in the San Francisco Bay Area. Lilly has already had a big win in using AI to crank up manufacturing of its popular GLP-1s.

Pharmaceutical makers and VC-backed startups alike are betting heavily that AI will change the way that new drugs are discovered, developed and manufactured. Hopes are riding high that AI could help find therapies for biological targets that had previously been considered “undruggable,” as well as accelerate the lengthy clinical trials process. Because developing a new drug can cost $1 billion and take 10 years, any way to speed up that process is extremely valuable - both for patients who need new treatments now, and for drug makers facing a ticking patent clock to make money off their innovations.

Roche said earlier this year that its AI tools had helped its scientists design a specialized molecule for oncology treatments 25% faster and with a structure that would not have been possible without AI. In another cancer treatment that is now entering human trials, researchers used AI to ensure that the therapy didn’t cause an immune system response that would reduce its chance of success.

“We used AI to predict if the molecule will be immunogenic and to clean it up,” Regev says. “That was done very quickly by AI.”

Like other big drug developers, Roche is both using AI internally and working with outside collaborators. Last year, for example, it announced a partnership with VC-backed startup Manifold Bio, paying it $55 million up front to find new ways to transport medicines to the brain with the help of AI. If it hits certain research, clinical and commercial milestones, the deal could be worth up to $2 billion.

Given the speed at which AI is moving, Regev says that it’s no longer about where the payoff will happen first, but about embedding it throughout the entire drug discovery, development and manufacturing process at the same time.

“It’s a little bit everything, everywhere, all at once,” says Regev, who joined Genentech in August 2020 and is also a member of Roche’s expanded corporate executive committee. “I think that’s a big shift in thinking. Before people were thinking there’s something in one area and everything else continues as before.”

Genentech first announced its collaboration with Nvidia in November 2023. At the World Economic Forum earlier this year, Nvidia cofounder and CEO Jensen Huang called out drug research as an area that will be transformed by AI.

“We’re going to see some really great, big breakthroughs,” he said at the time.