2025's AI-fueled scientific breakthroughs

AI-driven scientific innovation accelerated in 2025 — fueling major progress in fields ranging from robotics to health care.Why it matters: Powerful new models and computing tools are influencing everything from how experiments are designed to how results are analyzed — with much of that momentum coming from the private sector.Here are some of 2025's biggest AI-driven breakthroughs: 1. Making diagnoses for Alzheimer's and related diseases is on the road to becoming faster and cheaper with AI.Researchers at a wide range of universities and health care institutions announced findings this year about how AI will help with future therapies and better detection in primary care.For example, one study found that a specific gene is a cause of Alzheimer's — a discovery the researchers were only able to make because AI helped them visualize the three-dimensional structure of the protein.2. Google released its AlphaGenome model to understand diseases better and lead to drug discovery.The model was made possible by technical advancements that allow it to process long DNA sequences and provide quality predictions. 3. Advancements in humanoid robots' dexterity and human interaction this year could mean AI-enabled robots one day clean homes, keep people company, work in warehouses or provide care in health care settings, some observers are predicting.General-purpose humanoids are still a ways off, but money is pouring into the technology and companies are looking to combine generative AI with physical capabilities.4. Weather forecasting is more powerful than ever thanks to AI.Researchers are combining AI with physics-based climate models to predict extreme weather that may happen every 1,000 years, also known as "gray swan" events.Google released its most advanced forecasting model, which can generate forecasts eight times faster than before.5. High demand for cement alternatives that are cost and emission-efficient led an MIT team to use AI to find new ingredients that can be used in concrete. A machine-learning framework helped the team analyze scientific literature and more than one million rock samples to narrow down viable alternatives.Zoom out: Many of the advancements were enabled by AI tools built outside traditional academic and government research institutions. That has put companies such as Google, Microsoft, Nvidia and OpenAI at the center of a new scientific revolution.Google's investments in AI-driven science stretch back more than a decade. DeepMind's work led to a Nobel-winning breakthrough with AlphaFold2 — which predicts a protein's 3D structure — five years ago.Google-affiliated scientists have been associated with six Nobel Prizes, three awarded in just the last two years.Between the lines: The AI boom is also fueling a new generation of science startups. One buzzy company, Lila Sciences, has declared it's on a mission to "build scientific superintelligence."Backed by deep-pocketed venture capital firms, the startup says it's using specialized AI software to come up with and direct experiments in real-world labs.Another startup, Latent Labs, announced a frontier model earlier this month that it says can help design drugs and accelerate development timelines for pharmaceuticals by reducing wet lab work.Stunning stat: The federal government invested $3.3 billion in non-defense AI research and development in fiscal year 2025, the Center for Strategic and International Studies notes in a report.In the private sector, investments exceeded $109 billion in 2024.What we're watching: President Trump is moving to put his stamp on AI-driven science. An executive order signed last month launching "the Genesis Mission" aims to coordinate research between federal agencies.Two dozen leading AI companies — including Microsoft, Nvidia and Google — have joined the effort.The administration is reportedly looking to accelerate robotics next year as well.