Abu Dhabi-based CNTXT AI opened a new Silicon Valley office, and it’s using it as much to bring robots home as to sell into the US, CEO and co-founder Mohammad Abu Sheikh tells EnterpriseAM. The office, in Milpitas, California, doubles as a deployment center, where the startup tests its clients’ robots and partners’ hardware before deploying them. The Wall Street Journal first reported the US expansion.
Most of its data supply clients are in the Valley: About 90% of clients in that part of the business are in Silicon Valley, mostly hyperscalers and robotics labs, Abu Sheikh says. Most of its USD 300 mn in committed revenue comes from US hyperscalers, AI developers, and large enterprises, making it necessary for the startup to operate in the same time zone, Abu Sheikh says. Across the whole company, CNTXT serves more than 250 government and enterprise customers in the Middle East and the US.
The startup plans to hire a senior vice president of revenue, a vice president of marketing, and a commercial team of 15 to 25 people in the first year of its US operations, Abu Sheikh confirmed.
What CNTXT actually does: The two-year-old company has three businesses. The biggest sells training data: it labels and annotates data for AI labs, hyperscalers, and enterprise AI teams, combining automation with human reviewers. The second is its own models, led by Munsit, an Arabic voice AI that CNTXT says is trained on more than 110k hours of audio across more than 25 dialects and now has more than 300k users. That’s roughly double the c. 150k it reported when it raised its USD 60 mn series A in June. The third and fastest-growing business is physical AI: CNTXT runs robotics training facilities and a network that records people doing real-world tasks, data that robotics labs use to teach machines to work outside the lab.
The wager is on a data shortage: AI models have outgrown the data available to train them, Abu Sheikh argues, and the gap is widest for robots that have to work in kitchens and warehouses. “Buyers are no longer short of compute or hardware. They are short of real-world data they are legally allowed to use, and that is what we supply,” he tells us. He told the WSJ the Gulf’s advantage is its large blue-collar workforce and its hotels, warehouses, and data centers, which match those of the US and Europe. He also cited supportive regulators.
And on a deployment gap: Lots of data is being collected, but very few humanoid robots are deployed anywhere yet, Abu Sheikh says. CNTXT wants to close that gap at home, deploying clients’ robots in Gulf sites where it gathered their training data. The office also works the other way, giving US companies a way into Gulf AI projects.
What’s next: Europe, including the UK, is CNTXT’s next market, and Abu Sheikh says China is “also interesting.”
The bigger picture
CNTXT AI isn’t the only UAE AI company setting up shop in the US: G42’s sovereign cloud arm Core42 is more than tripling capacity at its Buffalo, New York site, from 18 MW to 60 MW, and the site serves as the hub for its US operations. It has also leased space in a Minneapolis office tower being converted into an AI data center, adding to G42’s operations in California, Texas, and New York. G42 itself incorporated G42 USA in Delaware last year. Last month, it was again reported to be exploring a sale of a majority stake to US investors to secure long-term chip access, though a US outpost looks the more likely outcome. G42-backed Presight already sells into the US through partners, and Abu Dhabi’s own government is next: several US cities are already looking at the TAMM super app, and its first customer under the new TAMM Global company is expected to be American.
GO DEEPER- Abu Sheikh spoke to us for My Morning Routine and on why homegrown developers are pivoting to native Arabic data.