Pomegra
Provision Labs, Anjali Sardana's physical AI venture, acquired Mumbai hardware startup Basilisk Industries to control how robot-training data is captured.
Key Takeaways
- Provision Labs acquired Basilisk Industries on November 2, 2026. The price and structure are undisclosed.
- Basilisk founder Shourya Watwani will run Provision's hardware work and its in-house post-processing models.
- Provision backers Lachy Groom, Glade Brook, General Catalyst and Bain Capital Ventures also back Pronto.
Lead
Provision Labs, the San Francisco physical AI company set up by Pronto founder Anjali Sardana, has acquired Basilisk Industries, a hardware startup founded in Mumbai in 2023. Sardana announced the deal on November 2, 2026. Neither side disclosed the amount, the mix of cash and stock, or the size of the Basilisk team joining Provision.
The deal comes roughly a week after Provision Labs surfaced publicly in late October. The company collects real-world data used to train and test AI systems that operate robots and other machines. It is Sardana's second company, and she continues as chief executive of Pronto.
What Did Provision Labs Actually Buy?
Provision bought a small hardware and perception team that was already working on its products. Basilisk had built a 3D hand-tracking model, redesigned Provision's perception hardware, and assembled an early version of a lightweight tactile glove that records touch, which cameras cannot capture.
Watwani will lead hardware at Provision and work on its post-processing models, which turn raw captures into training-ready data. With no disclosed price, the deal reads as a strategic purchase of capability rather than an acquihire of a standalone product. Basilisk was already doing contract-style work for Provision before the sale.
Why Does Owning the Hardware Matter?
Owning the hardware lets Provision decide what gets captured and how, instead of depending on third-party devices. In robot training, the sensor determines the dataset. A camera-only rig records motion but misses grip force and contact, and a glove can record both.
Touch data is scarce. Video of human tasks is plentiful, but synchronized tactile recordings are not, and robotics developers building manipulation models have few places to buy them. Provision is betting that a vertically integrated data pipeline, from sensor to processed output, will be harder to copy than software alone.
Where Does the Money Come From?
Provision has not disclosed its funding, customers or commercial partnerships. Its named backers are Lachy Groom, Glade Brook, General Catalyst and Bain Capital Ventures. All four also invested in Pronto, so the new venture draws on an existing investor circle rather than a fresh set of outside checks.
Pronto itself gives a sense of that circle's appetite. The home-services startup, founded in 2025, extended its Series B in May 2026 with $20 million of new money on top of an initial $25 million, bringing the round to $45 million. The extension came at a $200 million valuation, roughly double the mark from about a month earlier, with cumulative funding near $60 million.
What Is the Sector Context?
Physical AI funding is accelerating. Dealroom data puts venture investment in the category at $9.9 billion in the third quarter of 2026, double the year-earlier figure. Money is flowing toward robot makers and also toward the data suppliers that feed them.
That makes data collection a contested layer. Provision competes for the same developer budgets as other firms that gather egocentric video, teleoperation recordings and sensor logs. A tactile glove is a differentiator only if developers prove willing to pay for touch data at scale, and no public evidence of that demand exists yet.
What Are the Privacy Questions?
The acquisition lands beside an unresolved debate over how Sardana's companies source their data. Pronto piloted a programme in which service professionals wore cameras while working in customers' homes, which raised privacy and consent concerns. Pronto later expanded the effort under an opt-in service called Pronto Verified, which uses anonymized footage for AI training.
Provision says it sources data from professionals around the world. The tactile glove adds a new category of personal data, and regulators in several markets are still working out how they treat biometric and behavioral recordings. Any customer in Europe or California will likely ask how consent is documented.
Hiring and Footprint
Provision is hiring across San Francisco, Toronto and Bengaluru for computer vision, perception, tactile data, environment reconstruction and hardware roles. Bengaluru gives it an engineering base in India, while the Basilisk team in Mumbai adds hardware skills that are harder to recruit in the US at short notice.
Outlook
Provision Labs has spent its first weeks buying the part of the stack it did not control. The next tests are commercial: whether named robotics developers sign on, whether the tactile glove moves from prototype to production, and whether the consent framework holds up. Until the company discloses funding and customers, the deal is best read as a bet that touch data becomes the scarce input in robot training.
