Max Jaderberg told WIRED Health in London that Isomorphic Labs has a broad pipeline of AI-designed medicines heading toward clinical trials, and one compound already has FDA clearance to enter human testing , a milestone that is either the beginning of something transformative or the most expensive science experiment in history.
Drug discovery has historically been a game of enormous waste. A compound enters preclinical testing, survives animal studies, clears regulatory review for a Phase 1 trial, and still has roughly a one-in-ten chance of making it to approval. The average timeline runs twelve to fifteen years. The average cost exceeds $2 billion. Isomorphic Labs, the AI drug design company spun out of Google DeepMind in 2021, is betting that all of those numbers are wrong , not marginally wrong, but structurally wrong , and that the right AI models can compress each stage of that process by factors that make pharmaceutical incumbents uncomfortable.
Jaderberg, now President at Isomorphic Labs after previously serving as Chief AI Officer, reflected on five years of progress at the WIRED Health conference in London earlier this month. He posted on LinkedIn after the event that what once felt like science fiction is "racing toward clinical reality" and that the company's drug design engine has recently achieved "breakthrough performance on the foundational capabilities essential to drug discovery." That engine, formally called IsoDDE and detailed in a company announcement in February 2026, represents a substantial evolution beyond AlphaFold 3, the protein structure prediction model that won Demis Hassabis a share of the 2024 Nobel Prize in Chemistry. IsoDDE is not just predicting how molecules fold , it is designing new ones from scratch, targeting disease mechanisms that traditional medicinal chemistry has struggled to address.
On January 28, 2026, the FDA cleared ISM8969 for human clinical trials, making it one of the first drugs designed by AI to reach that threshold. As Forbes noted in its February analysis, what makes ISM8969 unusual is not just that AI designed it, but that the AI identified a molecular interaction that human researchers had not been prioritizing , the kind of non-obvious path through biological complexity that Isomorphic's models are specifically built to find. Isomorphic has confirmed its initial clinical pipeline focuses on oncology candidates, which makes regulatory sense: cancer indications often have faster pathways and more clearly defined endpoints than chronic disease categories. Partnerships with Eli Lilly and Novartis, reported to total $3 billion in deal value, provide both validation and the clinical infrastructure to run trials that a standalone AI startup could not manage alone.
Demis Hassabis signaled at the World Economic Forum in Davos in January 2026 that the company had pushed its original target , AI-designed drugs in clinical trials by end of 2025 , back by roughly twelve months, with the revised goal being end of 2026. That slip drew some criticism, but ISM8969's FDA clearance in late January suggests the timeline compression was real, just slightly delayed. The $600 million financing round closed in March 2025, led by Thrive Capital, providing the runway to advance programs into human studies without depending entirely on partnership economics.
The Concern Side of the Ledger
The question the headline poses , concern or good? , deserves a direct answer, and the honest one is: both, and the proportions depend heavily on what happens next. The concern is not that AI should not design drugs. It is that the validation loop in drug development is brutally slow. A model can generate a structurally elegant compound that binds precisely to its target and still fail in human trials because of toxicity the AI did not model, off-target effects no computational framework predicted, or pharmacokinetic behavior that only emerges in a living system at scale. Every AI drug discovery company, Isomorphic included, is in the early innings of learning what its models get wrong. That knowledge only comes from clinical data, which takes years to generate and carries real risk for trial participants.
The structural concern is concentration. Isomorphic Labs is an Alphabet subsidiary with $600 million in recent funding, Nobel-winning technology, and partnership agreements with two of the largest pharmaceutical companies on earth. If AI-designed drug discovery works at scale, the competitive dynamics of the pharmaceutical industry will reshape around whoever controls the most capable models and the most proprietary biological training data. Smaller biotech companies and academic drug discovery programs operate on timelines and budgets that cannot match that infrastructure. As Jaderberg acknowledged at WIRED Health, moving toward the mission of solving all disease will require moving the entire industry, not just Isomorphic's internal pipeline. Whether that movement is collaborative or consolidating will be one of the defining questions of the next decade in medicine.
For now, the honest frame is cautious optimism with clear eyes about the gap between a promising compound and a proven drug. ISM8969 is in human trials. That matters. The pipeline Jaderberg described is real and growing. But the pharmaceutical graveyard is full of compounds that looked exactly this promising at exactly this stage. AI does not change biology. It changes how fast we learn which bets are worth making.
Also read: Anthropic's Mythos is a real threat to crypto infrastructure, just not in the way the panic suggests • GPT-5.5 lands as OpenAI accelerates its model release cadence to near-monthly • Sam Altman has apologized to Tumbler Ridge and now the AI industry faces its most consequential liability question yet
