{"version":1,"type":"story","url":"https://digestai.news/story/nvidia-secures-45-000-grace-gpus-for-kazakhstan-aws-launches-serverles","json":"https://digestai.news/story/nvidia-secures-45-000-grace-gpus-for-kazakhstan-aws-launches-serverles.json","markdown":"https://digestai.news/story/nvidia-secures-45-000-grace-gpus-for-kazakhstan-aws-launches-serverles.md","slug":"nvidia-secures-45-000-grace-gpus-for-kazakhstan-aws-launches-serverles","headline":"Nvidia secures 45,000 Grace GPUs for Kazakhstan; AWS launches serverless SageMaker training","summary":"Nvidia announced that its partner Firebird has obtained 45,000 Grace Blackwell GPUs for a new AI data centre in Kazakhstan, with U.S. export authorization. The first phase is being built in Data Center Valley together with Kazakhtelecom and is slated for commercial launch in the third quarter of 2027. After the second phase, the facility is expected to reach roughly 500 megawatts of AI capacity. Kazakhtelecom has already installed 400 H200 and 480 B300 GPUs in Astana and Almaty and plans to add over 1,000 units in Data Center Valley and more than 1,000 Vera Rubin units in the Ekibastuz region next year.\n\nAWS made its “training sessions” feature in SageMaker AI generally available. The serverless offering lets users write custom training loops, reward calculations and other procedures for four supported models—gpt-oss-20b, gemma-4-31B-it, Qwen3.6-27B and Qwen3.8-27B—while SageMaker handles GPU allocation, distributed execution, failure recovery and checkpointing. Pricing is token‑based, billing pauses during reward scoring, and no capacity reservation is required. User data and custom weights are not reused to train other models.","keyPoints":["Nvidia partner Firebird secured 45,000 Grace Blackwell GPUs for a Kazakhstan data centre targeting Q3 2027 launch.","Kazakhtelecom installed 400 H200 and 480 B300 GPUs and plans over 1,000 additional units in two regions next year.","AWS released serverless SageMaker training sessions for four models, handling GPU management and charging by tokens processed."],"whyItMatters":"The GPU deployment expands high‑performance AI compute in Central Asia, while AWS’s serverless training lowers the technical barrier for custom model development, accelerating AI adoption across industries.","category":{"slug":"hardware","name":"Hardware & Compute","url":"https://digestai.news/category/hardware"},"entities":{"companies":["NVIDIA","Firebird","Kazakhtelecom","AWS"],"models":["gpt-oss-20b","gemma-4-31B-it","Qwen3.6-27B","Qwen3.8-27B"],"people":[]},"firstPublishedAt":"2026-10-22T23:46:00Z","updatedAt":"2026-10-22T23:46:00Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"note.com","title":"Two AI news stories for today: NVIDIA introduces a plan for 45,000 GPUs in Kazakhstan, and AWS begins offering a feature in SageMaker that allows you to write your own training","url":"https://note.com/excavation_koho/n/ne576b5e484ac?hl=en","publishedAt":"2026-10-22T23:46:00Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":{"title":"AWS Expands SageMaker AI Agent Capabilities","url":"https://digestai.news/thread/aws-shows-how-contextual-bandits-lift-conversions-in-acquisition-funnels","storyCount":6},"cite":{"text":"Digest AI, \"Nvidia secures 45,000 Grace GPUs for Kazakhstan; AWS launches serverless SageMaker training\", 22 October 2026, https://digestai.news/story/nvidia-secures-45-000-grace-gpus-for-kazakhstan-aws-launches-serverles","publisher":"Digest AI","title":"Nvidia secures 45,000 Grace GPUs for Kazakhstan; AWS launches serverless SageMaker training","datePublished":"2026-10-22T23:46:00Z","url":"https://digestai.news/story/nvidia-secures-45-000-grace-gpus-for-kazakhstan-aws-launches-serverles"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}