CNTXT AI Books US$300 Million in Committed Revenue, Opens Bay Area Office
The Abu Dhabi company, founded in 2024, says hyperscalers and AI developers are behind most of the committed bookings as it moves into the US market.
Karim El-Sayed Karim El-Sayed covers company news, policy and regulation across the UAE and wider MENA for Anecdoted, with a focus on how new rules and licences reshape how startups operate. karim@anecdoted.com
CNTXT AI has secured more than US$300 million in committed revenue, founder and CEO Mohammad Abu Sheikh said, a figure that comes to about AED1.1 billion. Most of it is tied to US hyperscalers, AI developers and large enterprises.
The Abu Dhabi company started in 2024. It sells AI training data, Arabic-language models and robotics training, and counts more than 250 government and enterprise customers.
Its arrival at that number was quick by any measure. In June 2026 alone, the company closed a US$60 million Series A round, bought the enterprise AI firm Actualize, and put three commercial products on the market.
What the product line looks like
Munsit, CNTXT AI's Arabic voice model, has passed 300,000 users. It was trained on more than 110,000 hours of Arabic audio drawn from over 25 dialects.
The company also operates robotics training facilities where it collects real-world data rather than synthetic sets. That data now covers more than 300 use cases.
Abu Sheikh frames the business around the least glamorous part of the stack: supplying real-world data at the quality and scale that models require. That, he said, is why customers keep coming back.
The US office and who it is chasing
CNTXT AI has opened its first American office in the San Francisco Bay Area. The office is aimed at US government agencies, technology companies, and enterprises that want AI adapted to Middle Eastern markets.
The company's regional partners and customers include the Technology Innovation Institute, Nationwide Middle East Properties and Venture One.
The scale of the committed revenue says something about where AI budgets are actually going. Model builders and cloud providers get the attention, but their output depends on labelled speech, dialect coverage and physical-world recordings that are expensive and slow to assemble. A company that owns that supply, and owns it in Arabic, holds a position that is harder to copy than another model release.
Committed revenue is not the same as collected revenue. How much of the US$300 million converts on schedule will depend on delivery across those 250-plus accounts.
