Valve founder Gabe Newell’s neural chip company Starfish Neuroscience announced it’s developing a custom chip designed for next-generation, minimally invasive brain-computer interfaces—and it may be coming sooner than you think.

The company announced in a blog update that it’s creating a custom, ultra-low power neural chip in collaboration with R&D leader imec.

Starfish says the chip is intended for future wireless, battery-free brain implants capable of reading and stimulating neural activity in multiple areas simultaneously—a key requirement for treating complex neurological disorders involving circuit-level dysfunction. That’s the ‘read and write’ functions we’ve heard Newell speak about in previous talks on the subject.

Mike Armbinder (left) and Gabe Newel (right) | Image courtesy Valve

The project aims to overcome current limitations of minimally-invasive neural interface implants, which are often bulky, power-hungry, and difficult to scale across multiple brain regions.

Current clinical technologies, like Elon Musk’s Neuralink (approved by the FDA in 2023), typically focus on single-region intervention in the brain’s motor cortex. In contrast, Starfish hopes to reduce surgical burden through miniaturization, making implants easier to place across multiple sites.

SEE ALSO
What We Know So Far About Anduril's 'Eagle Eye' Military XR Headset and Founder's Reunion With Meta

And at just 2 × 4mm, Starfish’s chip is tiny. If you never imaged reading a brain chip spec sheet from a company founded by Valve’s Gabe Newell, well, welcome to the future. Starfish’s first brain chip boasts:

  • Low power: 1.1 mW total power consumption during normal recording 
  • Physically small: 2 x 4mm (0.3mm pitch BGA) 
  • Capable of both recording (spikes and LFP) & stimulation (biphasic pulses) 
  • 32 electrode sites, 16 simultaneous recording channels at 18.75kHz 
  • 1 current source for stimulating on arbitrary pairs of electrodes 
  • Onboard impedance monitoring and stim voltage transient measurement 
  • Digital onboard data processing and spike detection allows the device to operate via low-bandwidth wireless interfaces. 
  • Fabricated in TSMC 55nm process

It’s still early days though. The company is now calling for early-stage collaborators—particularly those working in wireless power delivery, communication, and implantable neural devices—to explore novel applications of this technology ahead of its expected availability in late 2025.

As Newell has long suggested, the real potential lies beyond medicine, noting back in 2023 that “we’re way closer to ‘the Matrix’ than people realize.”

“I think connecting to people’s motor cortex and visual cortex is going to be way easier than people expected and doing things like […] reading and writing to somebody’s motor cortex is way more of a tractable problem than making people feel ‘cold’. And you never would have guessed that,” Newell said in a 2023 interview with IGN. “And I never would have guessed that before going into it. It turns out your brain has really good interfaces for some things and really badly designed, kludgy interfaces for doing other things. And the fact that your immune system gets involved in your perception of temperature means there’s all sorts of weird parts of your brain that participate in the sensation of being cold, whereas your motor cortex [or] your visual cortex are much more tractable problems.”

In 2019, prior to his departure from Valve, the company’s Principal Experimental Psychologist Mike Ambinder also gave some insight into how brain-computer interface might inform immersive games.

“We can measure responses to in-game stimuli. And we’re not always getting [data] reliably, but we’re starting to figure out how. Think about what you’d want to know about your players. There’s a long list of things we can get right now with current technology, current generation analysis, and current generation experimentation,” Armbinder said in his GDC 2019 talk, which was entitled Brain-Computer Interfaces: One Possible Future for How We Play.


Thanks to Brad ‘SadlyItsBradley’ Lynch for pointing us to the news.

Newsletter graphic

This article may contain affiliate links. If you click an affiliate link and buy a product we may receive a small commission which helps support the publication. More information.

Well before the first modern XR products hit the market, Scott recognized the potential of the technology and set out to understand and document its growth. He has been professionally reporting on the space for nearly a decade as Editor at Road to VR, authoring more than 4,000 articles on the topic. Scott brings that seasoned insight to his reporting from major industry events across the globe.
  • XRC

    The "Squib" tech from Kathryn Bigelow's "Strange Days" movie is getting closer…

    • Christian Schildwaechter

      Not really. Squib is supposed to connect to the human sensory system and memory, and while the sensory system in theory is sort of feasible, because our senses are connected throughout the body with neurons that could be tapped into, memory is strictly brain-internally and very holographic in the sense that information it isn't stored in one location like in computers, but stored in the strength of connections between neurons that may also be involved in holding lots of other informations. We barely understand it, let alone would be able to read or write it.

      And neither is something Starfish Neuroscience is even trying to do. Their focus on the motor cortex that handles movement is much simpler than for examples transferring vision or touch. Each finger tip has around 3000 nerve endings, so to simulate touch, you'd have to send 3000 signals. But you can control muscles even through the skin by sending a strong enough signal/voltage that basically makes the muscle contract involuntarily.

      So if you are in a room with a gigantic pile of Lego bricks, memory is all the brick in the heap you cannot really access, touch is the ones on top that you can see, and the motor cortex is the light switch in the corner. Squib is reading the position and color of each brick and being able to change them without touching the pile, Starfish Neuroscience wants to push and read the light switch by poking it with a stick. Actually 32 sticks for poking and reading 16 light switches. Still very impressive, really useful when the connection to the motor cortex is broken, but very far from reading or uploading memory, or VR via BCI. For the foreseeable future, we will have to fill our memories the conventional way and strap large and heavy plastic boxes to our faces for immersion.

      https://uploads.disquscdn.com/images/56cb5c19f10c27205fef4d8726ba0bd54d894c0c15f5f904539fc8d65ac9bf4a.jpg

      • XRC

        Ah Christian, don't bring that science stuff in here….let us dream ;)

        • Christian Schildwaechter

          Hm, how about Play for Dream and Pimax Dream electrode add-ons for neurofeedback plus lots of vodka as a compromise? You get the dream, the brain-computer-interface, the immersion from VR, which has been proven to create similar memories as real world events, something flat games rarely achieve, the well researched training effects from neurofeedback already used with VR to influence the creation of new neuronal connections, and ethanol fueled memory alteration.

          • I can already see Jaap announcing Vodka addon for Pimax in the next Youtube video

          • Christian Schildwaechter

            They could add a small bottle to both the front and back of the HMD with a small pipe leading to the mouth, and a small pump to adjust the distribution of vodka between them, so the HMD is always perfectly balanced on the users head, even while looking straigt up or down.

  • Foreign Devil

    Valves logo was always a guy with a faucet knob implanted into his skull. . .seems appropriate.

  • Crazy. Do we know anything about how these are powered? I know I sure wouldn't want a lithium-ion battery implanted in my skull the way they have a tendency to spontaneously burn up… It'd kill you instantly.

    • Christian Schildwaechter

      10hp internal combustion engine burning hydrogen extracted from mitochondria during the Krebs cycle with subcutaneous solar cells implanted on top of the skull. This of course adds heat to your already warm head containing a brain that constantly consumes 20W, so you'll have to resort to a buzz cut, both for the cooling and light to run the solar cells. They went for this solution because someone's cousin told them that lithium batteries explode all the time, which must lead to millions of deaths each year, given that there are now several billion smartphones in use, all powered by lithium cells.

      So even though about 1.25mn pacemakers driven by lithium-iodine batteries are implemented each year, and apparently none of them ever exploded since they switched to them from batteries driven by plutonium decay (no joke) in 1972, this seemed the safer option. The implants are significantly smaller than the Hindenburg zeppelin also containing highly flammable hydrogen, so a repetition of the Lakehurst disaster seems unlikely.

      • Wow, you're so wonderfully kind and friendly and nice. Thanks for adding such positivity to this website. People must really love to have you around.

    • VR5

      From the article:

      Starfish says the chip is intended for future wireless, battery-free brain implants

      The company is now calling for early-stage collaborators—particularly those working in wireless power delivery

      Sounds like induction. So clearly no batteries, lithium-ion or other. And although you overstate the inflammation risk, I wouldn't want a battery (or even a chip for that matter) in my head either.

      • Thank you for giving a reasonable response and being nice about it! :)

        Battery free is very cool if they can pull it off! While L-ion batteries may be very safe if developed from a reputable source, there is no way to predict over time what changes in composition or design might lead to an unknown problem. For instance, all the caps that are bad and prone to leakage in 90's era game consoles. Or the Galaxy Note 7 fiasco. Accidents are also unpredictable and could damage the battery and cause ignition.

        A chip in the brain does sound crazy, but at the same time, people get bionic hips, screws in bones, and all kinds of tech things implanted in surgery. I don't know that I want to be a gen 1 adopter, but I think it can be made safe.

        • VR5

          Those examples of surgery all are results of medical conditions, where the risk is offset by the affliction it cures. Basically you're already in a bad place which makes the risk worth it.

          The chance to die or be worse off in health due to surgery (or the implant) is very low but not zero. But the affliction is already there, 100%. So it's worth it to take the small risk for the improvement it promises.

          Taking that risk as a healthy person, even if only a small percentage might be negatively impacted, seems not worth it. Or crazy, as you put it. Why risk your life for entertainment?

          I guess a better comparison would be bungee jumping or sky diving. Calculable risk for enjoyment. So yeah, I see people using the tech when it becomes available. But not mainstream.

          • Also very true. I guess it will depend on what benefits the tech offers as it matures. As you said, entertainment isn't exactly the best case to take the risk for, but there may be other benefits down the road, who knows. It's very interesting to watch, for certain.

      • Christian Schildwaechter

        Induction to transfer energy through the skull, but that energy still has to come from somewhere. Given that these devices will first be implemented in patients with severe disabilities, they may all be unable to move and thus a tethered solution would be possible, but obviously a BCI for healthy people would have to be mobile and therefore battery driven, only the battery wouldn't be implanted.

        There are many possible reasons not to implant a battery, requiring a more complex solution. One being high energy consumption, another that brain surgery is very risky and regularly replacing the battery undesirable. Pacemakers have to be replaced every 10-15 years, requiring another operation.

        The primary reason not to implement a battery in the brain is probably a lack of space, and that adding a solid body there that couldn't be anchored would increase the danger of potentially deadly bleeding/swelling. There are numerous different battery technologies based on the very reactive lithium, incl. types that in contrast to lithium-ion will never combust even when the cell is ruptured, with a perfect safety record, so that's not really the issue. Lots of (external) medical devices still use lithium-ion though due to its benefits, and several of these had to be recalled due to either faulty batteries or design errors that increased the danger of them igniting.

  • ZarathustraDK

    Part of me wonder how easy/how accurate it would be to make moter-cortex-controlled handtracking, as in the headset having sensors down the motor-cortex side of the head and using those to do handtracking instead of cameras.

    I mean, in the Deckard's case, it's going to have SLAM-tracked controllers which could be used as calibration-points for motor-cortex sensing-data to create a transform for hand/arm tracking (and leg-tracking for that matter). After training, it'd be a simple matter of moving your limbs to produce the same output for the sensors to pick up and translate into movement in VR, solving camera occlusion problems.

    Before you knock it, consider we have projects like Motoc that can calibrate playspaces between standalone headsets and lighthouse-based trackers simply by dancing around holding the tracker to the front of the headset for 10 seconds. Yeah, it's much simpler triangulating a playspace, but then again, ultimately, we're talking about recording information and correlating it to some other information to build a model. In that way it is not different. Also, it's simple one-way communication, so it should be much cheaper than the "matrix-like" 2-way communication the article's getting at.

  • VR5

    The image showing Armbinder holding a drill to Newel's head is hilarious, lol. Very interesting tech although I would never get such implants and doubt there's a (mainstream) market in gaming for this. But I'd like them to make it work on a tech level.