Robotics SINGLE SOURCE

NVIDIA Says Kazakhtelecom Has Deployed 400 H200 and 480 B300 GPUs

First reported 23 Oct, 02:18 UTCSource: Unite.AILast editorial activity 23 Oct, 02:18 UTCWrite a comment ↓

A close-up view of modern GPU units, ideal for gaming and tech visuals. — Illustrative photo.Photo by Nana Dua on PexelsNVIDIA logo shown for identification only; no affiliation with or endorsement of WORLDTECH is implied.

NVIDIA says Kazakhtelecom, which it describes as the nation's largest telecommunications company, has already put 400 H200 and 480 B300 GPUs into operation in Astana and Almaty. In the same post, inDrive and Pony.ai are reported to be exploring robotaxi services, with Astana as the first city. The figures come from a post bylined to NVIDIA Vice President Rev Lebaredian, as reported by Unite.AI.

Capacity already in place and planned

According to the post, the installed hardware is split between the two cities, with 400 H200 units and 480 B300 units running across Astana and Almaty. NVIDIA describes these chips as processors built for many calculations at once, used for graphics and AI.

Next year, Kazakhtelecom intends to bring more than 1,000 GPUs online at each of two further locations: the Data Center Valley and the Ekibastuz region, where Vera Rubin GPUs are to be used. NVIDIA says the infrastructure is designed to support diverse workloads that enterprises, developers and institutions can run at scale.

Robotaxi exploration in Astana

The post also reports that inDrive and Pony.ai are working together to examine robotaxi services in Kazakhstan, with Astana as the initial deployment city. The companies will assess whether Pony.ai's seventh-generation robotaxis can be offered through the inDrive app, backed by NVIDIA AI infrastructure built on the Hyperion sensor and compute platform with the Halos safety system.

Pony.ai says it has worked with NVIDIA since 2017, introduced computing units based on DRIVE Orin chips in 2022, and in 2025 scaled to what it calls the world's first Level 4 domain controller with four DRIVE Orin X chips for its mass-produced seventh-generation robotaxi. The platform can link two DRIVE Thor chips for a combined maximum of 4,000 FP4 TFLOPS. A separate April 25, 2026 announcement from Pony.ai described an upgraded Level 4 domain controller developed with NVIDIA on the DRIVE Hyperion platform.

Framing and what is not yet settled

Lebaredian's post presents the announcements as evidence that Kazakhstan's AI ecosystem is moving from planning to deployment, spanning what NVIDIA calls the full AI stack of infrastructure, models and applications. It argues that more local capacity lets developers and companies train, fine-tune and serve more capable models, and that successful applications in turn create demand for more compute.

These statements come from NVIDIA's own post and from the companies involved, and the robotaxi element is described as exploration rather than a launch. The account rests on a single outlet's report of that post.

Open questions

When, if ever, will the Astana robotaxi exploration become a commercial service?
Will the planned GPU sites in the Data Center Valley and Ekibastuz be completed on the announced timeline?

Key facts

NVIDIA says Kazakhtelecom has deployed 400 H200 and 480 B300 GPUs in Astana and Almaty.

Kazakhtelecom plans to add more than 1,000 GPUs at each of two further sites next year.

The two planned sites are the Data Center Valley and the Ekibastuz region, where Vera Rubin GPUs are to be used.

inDrive and Pony.ai are exploring robotaxi services in Kazakhstan, with Astana as the first deployment city.