Sabi, a San Francisco startup building brain-computer interfaces disguised as baseball caps, said Wednesday it raised $50 million in a seed round led by Khosla Ventures. The round, which included Accel, Initialized Capital, DST Global, Collaborative Fund, Ascend and Kevin Weil, OpenAI's former chief product officer, valued the company at $600 million post-money, according to Forbes, though the company's press release did not disclose the valuation.
The financing underscores a growing investor bet on noninvasive alternatives to surgical brain implants. While companies like Neuralink and Synchron require surgery to place electrodes on or inside brain tissue, Sabi embeds custom EEG sensors inside ordinary headwear. The technology is designed to translate imagined speech and intent into AI commands by reading signals through hair and scalp, without breaking skin.
The sensors themselves are miniature, with what the company describes as a 1-to-5-millimeter footprint, housed inside a cap that looks unremarkable from the outside. They can capture brain signals through hair with gaps of up to 5 millimeters and connect wirelessly to devices, similar to how Bluetooth earbuds pair with phones. Sabi claims its custom EEG chip runs more than 50 times as power-efficiently as Texas Instruments' ADS1299, the industry standard.
Behind that efficiency claim sits a training process the company says drew on more than 100,000 hours of labeled, noninvasive recordings, what it calls the largest disclosed dataset of its kind. Wired reported in April that Sabi collected 100,000 hours from 100 volunteers while testing an earlier beanie prototype. That volume of data, if accurate, represents a substantial head start in a field where neural signal interpretation remains notoriously finicky.
CEO Rahul Chhabra and CTO Atmadeep Banerjee co-founded the company. Banerjee co-authored MindEye, a 2023 NeurIPS paper on decoding images from fMRI brain scans, lending some academic credibility to the venture. Sabi says its team includes former leaders from Kernel, Apple, Microsoft, Meta, Adobe and Nike, along with neuroscientists from Stanford, though the company has not detailed their specific roles.
In a post indexed by Techmeme, Chhabra thanked TSMC "for helping fabricate the world's first chip purpose-built for consumer BCI." TSMC's involvement is claimed in the founder's X post but has not been verified by TSMC or mentioned in the company's press release, and TSMC did not respond to requests for clarification on the scope of that partnership.
The new capital will fund expansion of Sabi's neural dataset, complete custom chip development and manufacture early devices. The company also plans to grow its neuroscience, hardware and machine learning teams and prepare units for participants on its early-access waitlist. Sabi aims to debut an early prototype at CES 2027 in Las Vegas, according to the announcement.
The cap is designed to enable imagined speech-to-text, thought-to-prompt and intent prediction, though those capabilities remain largely theoretical for now. Wired reported in April that Sabi initially targeted roughly 30 words per minute, with speed expected to improve as users adapt. The company has not disclosed error rates or how well the system performs across different users, metrics that will likely determine whether the technology can move beyond tech demos.
On privacy, Sabi says neural data gets encrypted at the sensor and inference occurs while data remains encrypted, with user controls over recording and app permissions. How those controls function in practice remains to be seen.
"You cannot perform a billion brain surgeries to replace a billion keyboards," Vinod Khosla, founder of Khosla Ventures, said in the announcement. "Every computing era is defined by its interface. Sabi is building one designed for the age of AI."
Khosla told Forbes that "very simply, thought-to-computer input is a sexy idea." In April, he told Wired that noninvasive approaches offer the only path to scale: "If you're going to have a billion people use BCI ... it can't be invasive."
That thesis positions Sabi against a divided market. Surgical implant companies like Neuralink, Paradromics and Precision Neuroscience are pursuing higher-fidelity signals through direct electrode placement on or in brain tissue. Consumer-focused startups like Neurable and OpenBCI have shipped EEG headphones and development kits, but Wired reported that calibration requirements remain a usability hurdle. Most require time-consuming setup processes that derail the casual user experience.
JoJo Platt, a neurotech consultant, told Wired in April that consumer BCIs must be "ready to go out of the box" and work consistently to succeed. Sabi's challenge will be proving it can deliver on that standard, particularly given how much individual brain signals vary.
"AI can write, reason and act, yet we still communicate with it by tapping on keyboards and screens," Chhabra said in the announcement. "The goal is a cap people can wear in ordinary life and use when they intentionally want to communicate with AI."
The company's website shows product renders and hosts a waitlist for early access, though no pricing or specific availability timeline has been disclosed. Whether a $600 million valuation on a seed round proves prescient or premature will depend largely on whether Sabi can navigate the gap between controlled demonstrations and devices that work reliably in messy, real-world conditions.
