Google's Gemini for Science reshapes how research gets done

The News

Google used its annual I/O developer conference this week to draw a sprawl of scientific AI projects under a single banner called Gemini for Science. The package, unveiled on Tuesday, bundles tools that until now lived as separate research releases from Google DeepMind, the company's London-based AI lab.

The line-up is broad. WeatherNext, a forecasting system credited with flagging Hurricane Melissa's approach to Jamaica ahead of conventional models, gets its newest version in November. AlphaGenome and AlphaEarth Foundations, released last summer for genetics and Earth science, sit beside two tools that are not yet public: AlphaEvolve, which optimises algorithms, and an AI Co-Scientist built to generate research hypotheses.

Demis Hassabis, the Google DeepMind chief executive, set the tone with a bold keynote line, telling the audience the industry is "standing in the foothills of the singularity". In a separate interview he was more measured, describing AI as "this amazing tool to help scientists" for the coming decade before the systems mature into something closer to collaborators.

Why It Matters

The shift here is less about any single tool than about packaging. DeepMind has spent five years building scientific credibility on the back of AlphaFold, whose protein-structure predictions have now been used by roughly 3 million researchers worldwide. Folding that goodwill into a branded product line signals that Google wants science to be a recurring reputation engine, not a string of one-off papers.

The money is following. Isomorphic Labs, the DeepMind spinout chasing AI-designed drugs, has raised a $2 billion Series B, a sum that would have looked absurd for a drug-discovery startup before AlphaFold existed. Five years on from AlphaFold's debut, the gap between a striking demo and an approved medicine remains wide, and investors are betting heavily it will close.

There is a quieter signal too. John Jumper, the Nobel laureate behind AlphaFold, has moved on to AI coding rather than science-specific tools, and an OpenAI reasoning model recently disproved a mathematics conjecture using GPT-5.5 with no specialist training. The frontier may be drifting from bespoke science models towards general-purpose reasoning.

Indian Angle

For India, the most immediate prize is weather. The economy still turns on the monsoon, and the India Meteorological Department has spent years upgrading its forecasting. A maturing WeatherNext, validated on a real hurricane, could sharpen cyclone warnings in the Bay of Bengal and the rainfall forecasts that shape sowing decisions for millions of farmers.

Drug discovery is the second front. Indian pharma built a vast export business on generics, copying molecules after patents expire. AI-designed drugs threaten to reset that contest towards original discovery, where firms such as Sun Pharma, Dr Reddy's and Biocon have historically been thinner. The $2 billion behind Isomorphic Labs is a warning that the next decade of pharma value may accrue to whoever owns the discovery platform, not the manufacturing line.

Then there is sovereignty. The IndiaAI Mission and the Department of Science and Technology have pushed for home-grown research infrastructure. If the most capable scientific AI sits inside a Gemini subscription, Indian universities and public labs face a familiar choice: pay for frontier tools or risk slipping behind on the discovery curve.

FAQ

What is Gemini for Science?

It is a package announced at Google I/O 2026 that groups Google DeepMind's scientific AI tools, including WeatherNext, AlphaGenome and AlphaEarth Foundations, under one banner. Two further tools, AlphaEvolve and an AI Co-Scientist, were announced at the same event but are not yet publicly available.

When does the new WeatherNext version arrive?

Google said the newest version of WeatherNext is due in November. Earlier versions were credited with flagging Hurricane Melissa's path towards Jamaica ahead of conventional models, which is why meteorological agencies are watching it closely.

Why does Isomorphic Labs matter to Indian pharma?

Isomorphic Labs, a Google DeepMind spinout, raised a $2 billion Series B to pursue AI-designed drugs. That scale of capital signals a shift towards original drug discovery, an area where India's generics-heavy pharmaceutical sector has traditionally been less dominant than in low-cost manufacturing.

Where can I read the original announcement?

MIT Technology Review covered the Google I/O science announcements in detail, including the keynote remarks from Demis Hassabis. The link to its full report appears in the attribution note at the foot of this article.

This story was reported by MIT Technology Review. Read the full original coverage at MIT Technology Review.