Big Tech Pushing into Drug Discovery

Amazon Web Services moved further into pharmaceutical research on April 14, launching Amazon Bio Discovery, a software product that lets scientists design and test drug candidates inside Amazon’s cloud tools. (aboutamazon.com) Drug discovery starts with finding a biological target, then screening huge numbers of molecules to see which ones might bind to it like a key fitting a lock. Amazon said its new product gives researchers access to specialized biological foundation models and routes promising candidates to lab partners for synthesis and testing. (aws.amazon.com) Amazon is not alone. Google DeepMind’s drug-focused spinout Isomorphic Labs said in January 2024 that collaborations with Eli Lilly and Novartis could be worth nearly $3 billion, excluding royalties, and chief executive Demis Hassabis said in Davos in January 2026 that the company now expects its first clinical trials by the end of 2026. (isomorphiclabs.com) (marketscreener.com) Microsoft has taken a partnership route. In February 2024, 1910 Genetics said it signed a five-year commercial agreement with Microsoft to build an artificial-intelligence and high-performance-computing platform for drug discovery and development. (1910.ai) The technical shift is that tech companies are selling more than storage and chips. Amazon’s product includes “lab-in-the-loop” workflows, where software proposes molecules, outside labs run experiments, and the results feed back into the model for the next round. (aws.amazon.com) Google’s side of the field grew out of AlphaFold, the DeepMind system for predicting the three-dimensional shapes of proteins. Google DeepMind says AlphaFold 3 predicts structures and interactions across proteins and other biological molecules, which is useful because drug design often depends on knowing exactly where and how a molecule will attach. (deepmind.google) Drug research is a tempting target for automation because it is slow and expensive. PhRMA says developing one new medicine takes 10 to 15 years on average and costs about $2.6 billion including failures, while the United States Department of Health and Human Services says the clinical phase accounts for 69 percent of overall research and development costs. (phrma.org) (aspe.hhs.gov) Pharmaceutical companies are also building the computing muscle to use these systems at scale. Roche said on March 16, 2026 that it expanded its artificial-intelligence infrastructure with 2,176 high-performance graphics processing units across the United States and Europe to support work from discovery through commercialization. (roche.com) The pitch from tech companies is speed: fewer manual handoffs, more simulation before wet-lab work, and faster narrowing from millions of candidates to a short list. The open question is whether those gains produce medicines that survive clinical trials, where most drug candidates still fail. (aws.amazon.com) (phrma.org) For now, the clearest change is where the big platforms are positioning themselves. Amazon, Google, Microsoft, and chip suppliers such as NVIDIA are no longer just selling infrastructure to drugmakers; they are building tools that sit inside the research workflow itself. (aboutamazon.com) (isomorphiclabs.com) (microsoft.com) (nvidia.com)