At a health-focused event in New York, Google announced its latest push into healthcare AI with the launch of TxGemma, a suite of AI models designed to streamline the drug discovery process. Set to release later this month under Google’s Health AI Developer Foundations program, TxGemma aims to help researchers better predict the potential of new drug candidates.
According to Google, TxGemma models are trained to understand both natural language and complex therapeutic structures such as chemicals, molecules, and proteins. This dual capability could help scientists analyze vast datasets more efficiently and surface insights that might otherwise take months to uncover.
“The path from drug concept to approval is long and costly,” explained Karen DeSalvo, Google’s chief health officer, in a blog post shared with TechCrunch. “We’re collaborating with the research community to find smarter ways to make this process more efficient. With TxGemma, researchers can ask critical questions about a potential therapy’s safety and effectiveness — even before testing begins.”
While Google describes TxGemma as a collection of “open” models, the company hasn’t clarified if developers will be able to fine-tune, customize, or commercialize the models. Google has yet to respond to inquiries about licensing terms and commercial usage rights.
The move adds momentum to a growing industry trend — the race to use AI for drug discovery. Google itself has already spun out Isomorphic Labs, a company focused on using AI to design new drugs in partnership with major pharmaceutical players like Eli Lilly and Novartis. Isomorphic recently confirmed it plans to begin testing its AI-designed drugs sometime this year.
Despite the optimism, AI’s impact on drug discovery remains a mixed bag. Companies like Exscientia and BenevolentAI, once seen as leaders in this space, have suffered clinical trial setbacks, highlighting the challenges of translating AI-driven predictions into real-world therapies.
Even AlphaFold 3, developed by Google DeepMind, has faced scrutiny over its variable accuracy in predicting protein structures — a critical step in drug development.
Still, pharma companies and investors remain bullish. By some estimates, over 460 AI startups are tackling drug discovery, with more than $60 billion invested so far. The potential payoff is massive: faster drug development timelines, reduced costs, and breakthroughs in treating complex diseases.
With TxGemma, Google joins a growing list of tech companies betting that AI can revolutionize how drugs are discovered and developed. The models aim to empower scientists by offering a new set of tools to predict drug properties early in the research phase — potentially saving years of trial and error.
As pharma’s interest in AI deepens, the question now is whether models like TxGemma can deliver reliable, real-world results. Google plans to release more details when TxGemma becomes available later this month, giving researchers a clearer view of how the technology fits into the future of drug development.
