Merge Labs · Grey Matter

Merge Labs is a brain-computer interface research company, co-founded by Sam Altman, that came out of stealth on 15 January 2026 with a 252 million dollar seed round led by OpenAI, and that aims to read and modulate the brain with ultrasound and engineered molecules instead of implanted electrodes.

Merge Labs. Merge Labs is a brain-computer interface research company, co-founded by Sam Altman, that came out of stealth on 15 January 2026 with a 252 million dollar seed round led by OpenAI, and that aims to read and modulate the brain with ultrasound and engineered molecules instead of implanted electrodes.

Its bet is on reach. Functional ultrasound already maps activity at a fraction of a millimetre across the whole depth of the brain, but only where the skull is thinned or removed; Merge's plan is to pair the sound with molecules that make neurons easier to see and to address through it, the line of work its co-founder Mikhail Shapiro is known for, with proteins that echo under ultrasound. The company describes itself as a research lab that thinks in decades, with medical uses first.

Its founders mix the two worlds. The neurotechnology researchers Mikhail Shapiro, Tyson Aflalo and Sumner Norman joined Altman, Alex Blania and Sandro Herbig; the company grew out of Forest Neurotech, a non-profit research organisation founded in 2023.

Its money came with a partner. OpenAI wrote the largest cheque, alongside Bain Capital and Gabe Newell, at a valuation of about 850 million dollars, and it agreed to collaborate on scientific foundation models, which made the deal one between two companies Altman helped found.

Its hardware is borrowed silicon. On 1 September 2026 it took a multi-year licence to Butterfly Network's ultrasound-on-chip, through the Butterfly Embedded business, with Butterfly as its exclusive chip-based ultrasound partner.

Merge Labs bets that coverage will beat contact.

Electrodes hear a few thousand neurons up close, and the wager is that sound, helped by molecules, can reach far more of the brain without anyone opening the skull.

Questions: What would a molecular brain interface add to ultrasound? Ultrasound sees blood flow, a slow and blurred echo of what neurons do. Optogenetics found a way around a similar limit for light: give chosen neurons a gene for a protein that responds to it, and light can switch them on or off with cell-type precision. The ultrasound version would give neurons proteins that echo sound or respond to it; Mikhail Shapiro, a co-founder of Merge Labs, showed in 2014 that tiny gas-filled protein shells made by microbes return an ultrasound signal. Merge Labs has said it combines such molecular approaches with ultrasound, without publishing which molecules or how they would reach a human brain. Why does Merge Labs build on ultrasound when implanted electrodes already work? Implanted electrodes hear a few thousand neurons in a patch of cortex a few millimetres across, and each new patch costs another operation. Sound travels through the whole depth of the brain, and functional ultrasound already maps activity at a fraction of a millimetre in animals and through skull windows in people. Merge Labs is betting that coverage of many regions at once will matter more than single cells, and that the skull, which still blocks good images, can be dealt with by better chips and by molecules that make neurons answer to sound. Its licence of Butterfly Network's ultrasound-on-chip in September 2026 is the hardware side of that bet.