Sources: NVIDIA: Kazakhstan expands compute for its AI future; Kazakhstan presidency: meeting on computing infrastructure

Kazakhtelecom GPUs reported deployed

Units by model; supplier account captured October 25

0240480400NVIDIA H200480NVIDIA B300
NVIDIA report; counts do not imply equal computing performance. [1]

Sources: [1] NVIDIA: Kazakhstan expands compute for its AI future · Page dated October 22, 2026; captured body includes October 23 events; 2027 plans remain prospective

A supplier reports installed hardware

NVIDIA’s Kazakhstan update says Kazakhtelecom has deployed 400 H200 and 480 B300 GPUs in Astana and Almaty. It also describes further deployments planned for 2027. The page is dated October 22, while its captured body includes events from October 23; this edition preserves that distinction. These are the supplier’s deployment claims, not an independent site audit. Different GPU models are counted separately because a unit count does not establish equal performance or usable computing output. [1]

The power plan predates this week

An October 1 statement from Kazakhstan’s presidency described the Data Center Valley project in Ekibastuz as envisaging approximately 300 megawatts of computing infrastructure, with potential expansion to one gigawatt. That earlier official statement provides dated context for the new supplier update. It does not establish that either power level was already operating at this edition’s cutoff. A planned electrical capacity is not a measured workload, and it cannot be converted into a number of productive GPUs without additional engineering information. [2]

A manufactured resource needs an operating system around it

The strategic-resource question extends beyond access to raw materials. Advanced computing hardware can become a constrained industrial input, but its usefulness depends on power, cooling, networks, software and people. A country can acquire equipment without yet having a dependable commercial service, just as a project can announce capacity before it is energised. The distinction matters for assessing capital commitments and economic development. This story follows that chain rather than treating every announced megawatt or processor as an interchangeable unit of national productive capability.

Sources & methodology

Bracketed numbers refer to the sources below. Analysis is original editorial interpretation, not a personalized recommendation.

  1. NVIDIA: Kazakhstan expands compute for its AI future · Page dated October 22, 2026; captured body includes October 23 events; 2027 plans remain prospective
  2. Kazakhstan presidency: meeting on computing infrastructure · October 1, 2026; historical context and prospective power capacity