UAE’s G42 To Build 8 Exoflop AI Supercomputer In India
MEA said that the supercomputer cluster will be open for access to both public and private sector entities for research, application development, and
UAE-based technology firm G42 will build an 8 exaflop AI supercomputer cluster in India, a project slated to be delivered in partnership with US-based AI chipmaker Cerebras, the Mohamed Bin Zayed University of Artificial Intelligence and C-DAC, according to reporting published on February 21, 2026. The Ministry of External Affairs (MEA) said the cluster will be available to both public and private sector entities for research, application development and commercial use. The announced 8 exaflop capacity is roughly 19 times the combined peak 410 petaflop compute capacity of India’s two flagship AI supercomputers, AIRAWAT and PARAM.
"The supercomputer cluster will be open for access to both public and private sector entities for research, application development, and commercial use," the MEA said, summarising the intended scope of access for the new facility.
Project partners and technical scale
The delivery consortium named in media reports comprises G42, Cerebras, Mohamed Bin Zayed University of Artificial Intelligence and C-DAC. Cerebras is identified as a US-based AI chipmaker; Mohamed Bin Zayed University of Artificial Intelligence appears as an academic partner; and C-DAC is listed among the domestic collaborators tasked with integration and deployment. Together, these organisations are expected to assemble the hardware, networking and software stack required to reach an aggregate peak performance of 8 exaflops.
- Lead developer: G42 (UAE)
- Hardware/AI silicon partner: Cerebras (US)
- Academic partner: Mohamed Bin Zayed University of Artificial Intelligence
- Local systems integration/government partner: C-DAC
- Announced peak capacity: 8 exaflops (~8,000 petaflops)
- Comparison cited: almost 19X the combined 410 petaflops of AIRAWAT and PARAM
Public reporting emphasises the scale gap the new cluster will create: India’s currently prominent national systems—AIRAWAT and PARAM—have a combined peak compute capacity reported at 410 petaflops, while an 8 exaflop installation equates to multiple thousands of petaflops. The MEA statement positions the facility as a shared national asset intended to support both government research and commercially oriented development by private firms.
Outlook
The announcement signals a notable increase in available AI compute in India should the project proceed to completion as described. By designating the cluster for use by both public and private entities, the MEA’s statement implies an intent to broaden access beyond a small set of national research labs, offering capacity for universities, startups and corporations to run large-scale AI training and inference workloads. How access will be governed, priced, or scheduled was not detailed in the published summary.
Reporting on February 21, 2026, lists G42 and its partners as the delivery team for the project; further technical specifications, timelines and procurement arrangements are yet to be disclosed in the available public summary.