G42

Core42’s new program separates software builders from resellers, but both routes rely on a cloud stack combining Microsoft Azure services with Core42 controls. The launch is a distribution signal—not yet evidence of customer conversion, recurring revenue or delegated regulatory responsibility.

Abu Dhabi's G42 describes itself through a family of operating companies — Core42, Khazna Data Centers and Presight among them — and its public narrative is usually told in headline units: gigawatts of planned AI campus, hundreds of thousands of authorized chips, national-scale cloud ambitions. Yet the part of the group that can actually be audited by outsiders is far smaller, and structurally revealing. Only one of its principal operating companies, the ADX-listed analytics firm Presight AI Holding, publishes revenue, profit and backlog in reviewed financial statements. The layers that physically deliver the AI infrastructure story — the data-center operator Khazna and the cloud-and-compute company Core42 — publish no revenue at all, and no consolidated G42 financials exist in public sources. This article maps G42's internal division of labor through the only evidence available: audited numbers at the edges, self-reported capacity in the middle, and partnership announcements at the compute frontier — and argues that the group's scale narrative rests on its least-audited layers.

The most consequential number in Abu Dhabi's AI infrastructure story is not 5 gigawatts, 500,000 chips, or 35,000 chips. It is a date: April 6, 2027. Under a US Commerce Department rule effective July 10, 2026, two entities in the Group 42 orbit — Group 42 Holding Ltd d/b/a G42 (including G42 Cloud Technology LLC) and Core42 Technology Projects LLC d/b/a Core42 — may receive advanced computing items without an individual export license. That entitlement expires automatically on April 6, 2027 unless those entities become US companies, a condition reported by law firm Covington & Burling in its analysis of the Bureau of Industry and Security final rule. This article examines what that sunset clause changes about G42's operating footprint, and why it now matters more than any announcement about campus scale.

Abu Dhabi's G42 sits at the intersection of Gulf sovereign capital, US export-control policy and one of the fastest AI infrastructure buildouts anywhere outside the United States. The public record around the company is dense with large numbers — 500,000 chips a year in draft, 5GW of planned campus, 1GW of announced cluster — but the numbers that describe what has actually been authorized, built and energized are much smaller, and they arrive in a different order than the announcements did. This article organizes that public record into a stage-tagged ledger: what is approved, what is under construction, and what is operating. [BTW's research draws on company statements, wire-service reporting and US government documents; the load-bearing sources below were inspected only through search-result snapshots, so each claim is reported as a claim rather than an independently verified document fact.](https://btw.media/en/directory/g42)

Reuters says the 5GW project may shift from one Abu Dhabi campus to multiple UAE sites, while Stargate UAE's first 200MW remains targeted for 2026.

A multi-site Stargate UAE could reduce local outage risk, but resilience will depend on how workloads recover between locations.

G42 is becoming one of the clearest tests of whether sovereign AI can be sold as an operating capability rather than a political slogan. Its partnerships, cloud platforms, data center buildout, health and public-sector reach, and regulated-compute commitments give it a serious claim on the market, but the value is proved only when a sensitive workload reaches a running state with an auditable data boundary, clear control of infrastructure, secure partner access, accountable use of models, and costs that still make sense after supervision, integration, review, maintenance, and rollback are counted.

Cisco and G42's October 2025 expansion is not just another AI partnership announcement. Cisco is taking the networking, security, observability and integration role for a G42-deployed AI cluster using AMD MI350X GPUs inside G42's Regulated Technology Environment. That makes the event a test of whether the US-UAE AI corridor can turn export-controlled compute, trusted US suppliers and UAE sovereign-scale infrastructure into a governed deployment model rather than a political headline.