Foghorn Therapeutics told investors on October 1 that it is cutting about 40% of its workforce. In its 8-K filing the company says it expects to end up with roughly 65 full-time employees, to book about $2.3M in severance and benefit charges, and to finish the reduction in the fourth quarter. The stated cause is a clinical decision. AI and automation appear nowhere in the filing or the press release, so this goes in our file for cuts where AI is not the cause.
The sequence in the filing
On September 25, Foghorn and Eli Lilly looked at Phase 1 data for FHD-909, a SMARCA2 inhibitor, and decided not to move the trial into its expansion phase. The filing adds that further activity under the Lilly collaboration agreement is not anticipated. Five days later, on September 30, the board approved what it calls a strategic reprioritization and the workforce reduction. The company announced it the next morning.
The CEO’s explanation is one sentence of biology: the SMARCA2/4 synthetic-lethality idea 「has not translated into the level of efficacy required」. Foghorn is now concentrating on a selective EP300 degrader, an immunology and inflammation asset, a selective CBP degrader and its induced-proximity platform. The company says the changes extend its cash into the second half of 2029.
The headcount math
The 8-K gives a percentage and an end state, not the number of people let go. Working backwards, 65 remaining after a 40% cut implies a starting team of about 108 and roughly 43 roles removed. That is our arithmetic, not a figure Foghorn published. Dividing the $2.3M charge by 43 gives about $53K per role, though the filing does not break the charge down by person.
On raw headcount this is small. BioMarin’s Novato cut filed the day before was 117 roles at a single plant, nearly three times Foghorn’s implied number. On share of workforce it is far heavier. Anchorage Digital’s 17% came a day later, and Foghorn’s share is more than double that.
Why we log it as not-AI
Challenger’s September count put AI at 3,961 of 43,281 announced U.S. cuts, about 9%, fifth among the reasons employers gave. Reading the rest of that table matters as much as reading the AI line. A trial that fails and ends a partner collaboration is a familiar biotech cut. Nothing in Foghorn’s disclosure suggests a model replaced a chemist, and a reader who files this under AI displacement would be adding a cause the company did not state.
The contrast with Novo Nordisk’s September disclosure is the useful part. Novo put a headcount fall and an AI productivity claim in the same deck, so readers could draw their own line between them. Foghorn draws none, and its cut is explained entirely by one program ending.
What this means for a specific career
The people most exposed here are discovery biologists, medicinal and computational chemists, translational scientists and clinical-operations staff tied to the SMARCA2 program. Those roles hold up in a normal biotech hiring market and shrink when a program dies, whatever tools sit on the bench.
The timing is concrete. The company expects to complete the reduction in the fourth quarter, so affected staff will be on the market over the next three months. Protein-degradation and induced-proximity experience is the narrower, more portable skill set, and it is the one Foghorn says it is keeping around its EP300 and CBP programs.
For a reader at a similar-sized biotech, the signal to track is dependence on one partner-funded asset. A single collaboration ending produced a 40% reduction in five days. Remaining staff should note the runway claim: cash into the second half of 2029 is a long way off for a company that just cut its team, and it is the number that will be tested if the EP300 program stalls.
What to watch
Whether Foghorn files a later disclosure giving the exact number of roles cut. Whether other degrader and precision-oncology companies that share partner-funded programs with big pharma cut in the same pattern this quarter. And whether any company in this category starts to cite AI-driven discovery as a reason for a reduction, which would change how we file it. Until a source says so, this one stays labeled a program cut.
