The First AI-Engineered Antibody Just Reached Phase 3

Two global Phase 3 trials. 1,600 patients with severe asthma. One antibody designed by a machine-learning platform. GB-0895 is the first AI-engineered antibody to reach late-stage studies, and the SOLAIRIA program is already recruiting across 33 of 42 planned countries.

The drug working is the easy test. Whether generative biology survives first contact with Phase 3 is the real one.

The molecule: 89 days of half-life by design

TSLP, thymic stromal lymphopoietin, is the cytokine that sits upstream of airway inflammation in asthma. The target is validated. Tezepelumab, the one approved antibody against it, cut severe exacerbations by up to 71% in its Phase 3 program, and the biology of the pathway is well mapped in the clinical literature. What remained open was whether a competitor could build a better molecule on the same target.

Generate's answer rests on its generative biology platform, which couples proprietary machine-learning models with high-throughput experimentation. The antibody was computationally designed to improve potency and half-life while holding specificity, which is what allows the six-month dosing schedule.

Speed is the less obvious claim. Mike Nally, the chief executive, framed the milestone as an antibody engineered with AI that reached Phase 3 within four years. Conventional antibody programs rarely compress discovery to late-stage trials that fast, and that is why the design story, not just the asthma story, is what the market is watching.

The Phase 1 data behind the six-month claim came from 96 patients with mild to moderate asthma, presented at the European Respiratory Society (ERS) congress in September 2025. The antibody was well tolerated across the whole dose range, and its roughly 89-day half-life translates into pharmacology that should hold between two doses a year.

Phase 1 answers a limited question. It says the molecule is safe enough to dose and that the mechanism behaves as designed. It says nothing about efficacy at the scale that matters.

The trial: 1,600 patients, two shots a year

The SOLAIRIA program runs two replicate trials that are near-identical except for geography. Each tests the candidate at 300 mg subcutaneously every six months against placebo over 52 weeks. The primary endpoint is the annualized asthma exacerbation rate, with lung function, symptom control and quality of life tracked alongside.

Replication is the point. One positive trial can be a statistical accident, and that is why late-stage asthma programs usually run two. The cost is two readouts instead of one and a longer wait. The benefit is that a clean pair of results is far harder to dismiss.

Six-month dosing is the differentiation that matters commercially. Severe asthma biologics are chronic therapies and adherence decays between doses. An injection twice a year instead of twelve changes how the drug gets used, how the physician schedules it and how the patient remembers it.

Twelve injections a year become two.

Recruitment is the operational signal to watch. The company reported 33 of 42 planned countries cleared and enrollment underway across all six global regions as of the second quarter. The ClinicalTrials.gov listing for SOLAIRIA-1 was updated at the start of September with recruiting status, which means the clock on the program is running, not stalled.

The money: a $457 million bet on one platform

The company's cash position funds the program through the first half of 2028. Research and development spending ran at $64.3M in the second quarter, up from $59.7M a year earlier, and the company attributes the increase to the SOLAIRIA trials. Net loss widened to $67.3M from $56.7M.

The asymmetry is the point. The company does not need new funding to reach the first efficacy readout, but it has no cash headroom for a long detour. If either trial stumbles and enrollment or analysis slips beyond the runway, the next capital event happens from a position of weakness.

Platform economics sit on top of the clinical story. Revenue for the quarter was $6.3M, mostly from the Amgen and Novartis research collaborations, and Novartis has advanced a biologic designed on the Generate platform into its preclinical pipeline, the first time a biologic from an AI-enabled platform reached that stage. That repeatability claim, not any single molecule, is the basis of the long-term value case.

The thesis: what Phase 3 does to the AI-drug story

As we wrote in August, the self-driving lab is becoming infrastructure. The open question was always whether molecules designed in silico would clear the clinical bar. GB-0895 is the closest thing to an answer so far, and it is roughly two years from a regulatory decision.

The stakes are sector-wide. Dozens of AI-originated drugs are in clinical development across the industry, and the first approval anywhere would compress a timeline the field has been promising for a decade. The candidate is not guaranteed to be that drug. It is simply the most advanced antibody from the generative biology camp, and the first one built for a market where a validated target and an existing competitor already define the bar.

How much of its valuation already prices in a positive Phase 3? The stock trades on the platform story, and a clean readout would re-rate the entire generative biology cohort. A miss would do the opposite, and the conversation would shift back to AI as a tool inside drug development rather than a generator of drugs.

The distinction matters for allocation. Tool companies earn licensing revenue and sell into existing pipelines. Generator companies ask investors to underwrite molecules that do not exist yet. It is trying to be both, which is exactly why the SOLAIRIA outcome carries so much weight.

Does the first AI-designed antibody get approved by 2028?

Development scenarios