Sabi, a startup developing an interface between the brain and artificial intelligence built into a baseball cap, announced on October 9, 2026, that it has closed a $50 million seed round. The deal is led by Khosla Ventures, investor Vinod Khosla’s fund, with participation from Accel, Initialized Capital, DST Global, Collaborative Fund, Ascend and Kevin Weil, OpenAI’s former chief product officer.
According to Forbes, which cited a person familiar with the deal, the round values the company at $600 million. Sabi has not confirmed that figure in its official announcement. For a seed round, a company’s first institutional funding, the amount is extraordinarily high and reflects investors’ appetite for devices that promise new ways to interact with artificial intelligence assistants.
Sabi’s pitch is ambitious. The company describes its first product as a cap with sensors able to read neural signals through hair, without gel, surgery or direct contact with the scalp. Its goal is to turn imagined speech, meaning the words a person thinks without saying them, into text that appears on a connected screen, and to transform a trained mental command into an instruction for an artificial intelligence assistant.
Forbes describes a prototype lined with up to 100,000 sensors, each a chip between one and five millimeters across. The signals are interpreted by a model trained on about 100,000 hours of labeled neural data. In its announcement, Sabi speaks of “more than 100,000 hours” of labeled non-invasive recordings and says this is the largest disclosed dataset of its kind, a comparison that comes from the company itself.
The most striking figure is accuracy, and it should be read with caution. According to Forbes, chief executive Rahul Chhabra says an early, wired, helmet-like prototype can transcribe people’s thoughts with up to 77% accuracy in a lab setting, for users who have worn it for more than two weeks. This is a company claim about a device that predates the final product, and no data has been published that would allow independent verification.
The company says it will spend the money on expanding its dataset, developing a custom chip, building its first devices and hiring. Sabi plans to demonstrate early versions of these interactions at CES in Las Vegas in January 2027 and is preparing units for members of its early-access waitlist. Forbes reports that the cap will not be available until next year, with no specific launch date.
Privacy is a central issue for a product that works with signals from the brain. Sabi says neural data is encrypted starting at the sensor and that its system is designed to process that data while it stays encrypted. Chhabra has insisted that only the user will have access to their data. For now, these are design commitments that will need to be checked once the product reaches the market and can be examined by independent experts.
The scientific community is skeptical. Forbes calls the concept “nascent and largely unproven” and quotes Nicholas Hatsopoulos, a neurology professor at the University of Chicago, who notes that signals passing through the skull and scalp are of lower quality and that getting reliable results that way still seems a long stretch. Vinod Khosla himself acknowledges that early versions “won’t be perfect.”
The skepticism has a basis. The brain-computer interfaces that have achieved the most striking advances in decoding speech tend to be invasive: they require surgically implanting electrodes in the brain, which captures much cleaner signals. Non-invasive techniques, such as electroencephalography, are safer and more comfortable, but they record weaker, noisier activity. Sabi’s bet is to make up for that limitation with many sensors and with artificial intelligence models trained on huge volumes of data.
The timing is no accident. The technology industry is looking for the next big device for the age of artificial intelligence assistants, and smart glasses, pendants and other always-connected gadgets have emerged. The idea of controlling an assistant without speaking aloud or typing is very appealing, for example in public spaces or for people with speech difficulties, although Sabi presents its product as a consumer interface rather than a medical device.
Kevin Weil’s presence among the investors has also drawn attention, given his experience building consumer artificial intelligence products. And Khosla Ventures’ lead fits the track record of a fund that often backs high-risk, long-term technology projects. Even so, the history of consumer devices is full of attractive ideas that never made it from the lab into everyday life.
There are several things this round does not prove. It does not prove that the cap works as described, because the capability and accuracy claims come from the company and from a lab prototype. Nor does it confirm the valuation, which comes from an anonymous source cited by Forbes, or guarantee that a product will be on sale by a specific date: the CES 2027 appearance is a planned demonstration, not a launch.
What it does confirm is that investors are willing to fund generously the interfaces that could change how we talk to machines. If Sabi gets close to what it promises in real-world conditions, the keyboard and the voice would no longer be the only ways to give an assistant instructions. If it does not, it will be another example of how hard it is to read the mind from outside the skull. The January demonstration will be the first public test.
