Most people know Gabe Newell as the man behind Steam, the store where hundreds of millions of people buy video games. Fewer know that for years he has funded a quieter obsession: connecting brains and computers. His company, Starfish Neuroscience, has now shown its first concrete piece of hardware — and it looks almost nothing like the brain implants you have read about.
Instead of one big chip in one spot, Starfish wants many tiny implants sprinkled across the brain, each smaller than a grain of rice, each running without a battery. The idea sounds strange until you hear the medical logic behind it.
What was actually announced
On May 20, 2025, Starfish neuroengineer Nate Cermak published the company's first technical blog post, describing a custom electrophysiology chip built in partnership with imec, the Belgian microelectronics research institute. The specifications are unusually aggressive:
- Physical size: 2 x 4 millimeters, in a package with 0.3 mm ball-grid-array pitch.
- Power: 1.1 milliwatts during normal recording — far below the tens of milliwatts typical of commercial electrophysiology front-ends, and well under the roughly 6 milliwatts reported for Neuralink's N1 implant.
- Recording: 16 simultaneous channels from 32 electrode sites, sampling spikes and local field potentials at 18.75 kHz.
- Stimulation: capable of delivering biphasic pulses, not just listening.
- Processing: onboard spike detection and data reduction, so the implant does not need to stream everything it hears.
- Manufacturing: TSMC 55-nanometer process.
The company said its first chips were expected to arrive in late 2025, and it is openly looking for collaborators in wireless power delivery, communication, and custom implanted neural interfaces.
Why "many small implants" instead of "one big chip"
Today's clinical neural interfaces mostly talk to a single brain region. Deep brain stimulation for Parkinson's disease, for example, targets one nucleus with fixed settings. But many neurological disorders are circuit problems — misregulated conversations between several connected regions, not a fault in one place.
Starfish's argument runs like this:
- To read and write to multiple connected brain regions at once, you either build one large, complex implant reaching everywhere, or you distribute several small ones.
- Big implants mean big surgery, bulky electronics, and heavy power demands.
- Small, wireless, battery-free implants could be placed with far less invasive procedures, each drawing power from outside the body.
The chip itself is only a building block. A complete system still needs electrodes, leads, packaging, wireless power and data links through tissue, and a surgical delivery method — none of which Starfish claims to have finished. The company is candid that much of the design is a "half product" offered to partners rather than a finished implant.
The Newell factor
Valve, the game company Newell co-founded, explored brain-computer interface (脑机接口) ideas publicly as far back as the 2019 Game Developers Conference, and Newell incorporated Starfish the same year. Gamers regularly imagine mind-controlled games; Starfish's own materials, by contrast, consistently emphasize medicine — new therapies for disorders that involve multiple brain regions, alongside separate research programs in precision TMS and targeted tumor heating.
The gaming connection still matters, just not the way headlines imply. Newell brings long-horizon funding and an engineer's patience to a field where academic grants run out fast. Consumer dreams can wait; the chip strategy is aimed at researchers and medical-device partners first.
Reality check: where this stands in 2026
There is a wide gap between a chip announcement and a clinical product, and Starfish sits firmly on the early side of it.
- No human has received a Starfish implant, and the company's own materials say distributed interfaces of this kind do not yet exist clinically.
- As independent reviews of the public record have noted, no Starfish-sponsored clinical trial appears in public registries, and no regulatory clearance exists for the announced chip.
- Even if the hardware performs as specified, validating a multi-region wireless implant in living brains is a multi-year effort.
That is normal for neurotechnology. What makes the announcement worth watching is the architecture: if the future of therapy is circuit-level, someone has to build the small, low-power, battery-free pieces first. Starfish is one of the few teams publishing real specifications for those pieces.
According to Li An, chief scientist at BrainNet (脑机网), the shift from single-site electrodes to distributed, multi-region interfaces is one of the more credible engineering bets in the field, because it targets how brain disorders actually behave rather than how implants are convenient to build.
Key takeaways
- Starfish Neuroscience, co-founded by Valve's Gabe Newell, unveiled a 2 x 4 mm, 1.1 mW recording-and-stimulation chip designed as a building block for battery-free, wireless implants.
- Its bet is architectural: many small implants reaching several brain regions at once, instead of one large device, to treat circuit-level disorders.
- The work remains pre-clinical — no human implant, no registered trial, no regulatory clearance as of late 2026.
Honest limitations
- The chip is a component, not a product; wireless power delivery, implant packaging, and surgical delivery are still unfinished pieces.
- Miniaturization trades away things clinicians value, such as high channel counts; 16 simultaneous channels is modest compared with research-grade systems.
- Nothing here has been tested in humans, and timelines in neurotechnology routinely slip by years.
What readers can do now
- If you follow implant tech, evaluate announcements by asking one question: has anything been implanted in a person, and where is the trial registration? It cuts through most hype.
- Researchers or medical-device teams working on wireless power or neural interfaces can engage Starfish directly — the company is explicitly seeking collaborators.
- Patients and families should treat early-stage programs like this as background research, not near-term treatment options; approved and recruiting therapies are the practical starting point.