--- title: "Nvidia’s top AI chips are coming to a Centurion data centre" url: https://daily.dev/posts/nvidia-s-top-ai-chips-are-coming-to-a-centurion-data-centre-xjh4vshjt source_url: https://techcentral.co.za/nvidias-top-ai-chips-are-coming-to-a-centurion-data-centre/285394 type: article source: "TechCentral" published: 2026-08-26T13:02:15.828Z updated: 2026-08-26T13:09:43.523Z tags: ["cloud", "nvidia", "gpu", "ai-infrastructure"] reading_time: 5 upvotes: 0 comments: 0 language: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Nvidia’s top AI chips are coming to a Centurion data centre **[TechCentral](https://daily.dev/sources/techcentral)** · 5 min read · 0 upvotes · 0 comments ## Summary South African start-up Stratos Lab, with Dubai-based integrator Ecoblox and host Digital Parks Africa, is deploying Nvidia B300 Blackwell Ultra GPUs at a Centurion data centre, claiming Africa's most powerful AI cloud. Phase one is 50 Giga Computing servers with 400 B300 GPUs (7.2 exaflops FP4), costing about $25-million, going live within three months. The 'most powerful' claim is contested: rival Cassava Technologies has more but older Hopper-generation chips, and on a like-for-like FP8 basis Cassava's cluster actually leads by roughly 3x. Each rack draws up to 65kW with rear-door liquid cooling, and the venture emphasizes data sovereignty and local ownership over foreign clouds. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How many Nvidia B300 GPUs is Stratos Lab deploying in its Centurion data centre and what is the total compute? Stratos Lab's first phase deploys 400 Nvidia B300 Blackwell Ultra GPUs across 50 Giga Computing servers, eight GPUs per server, hosted at Digital Parks Africa's Samrand campus in Centurion, South Africa. The partners cite 7.2 exaflops of combined FP4 throughput, costing about $25-million for the GPU compute alone, with deployment expected within three months. _Track real-world GPU cluster deployments and capacity claims like this one on daily.dev._ ### Is Stratos Lab's B300 GPU cluster actually more powerful than Cassava Technologies' GPU cluster in Africa? Not on a like-for-like basis. Stratos Lab's 7.2 exaflops figure is measured in FP4, a low-precision metric; restated in the higher-precision FP8, it halves to about 3.6 exaflops, while Cassava's 3,000 Hopper-generation GPUs (H100s and H200s) total roughly 12 exaflops in FP8, more than three times as much raw compute. _Compare GPU cluster performance claims carefully before betting infrastructure decisions on daily.dev._ ### How much power does each rack draw in Stratos Lab's new B300 GPU deployment? Each rack in the Centurion deployment draws up to 65kW, far above standard data centre compute density, and is cooled with a rear-door heat exchanger tied into the building's cooling loop. This high density lets racks sit directly next to each other, meaning available power becomes the limiting factor for expansion rather than floor space. _Follow power and cooling constraints shaping AI data centre buildouts on daily.dev._ ## Similar posts on daily.dev - [Africa’s first Nvidia RTX Pro GPU servers have landed](https://daily.dev/posts/africa-s-first-nvidia-rtx-pro-gpu-servers-have-landed-i4chcla20) · TechCentral · 0 upvotes · 0 comments --- Tags: [#cloud](https://daily.dev/tags/cloud), [#nvidia](https://daily.dev/tags/nvidia), [#gpu](https://daily.dev/tags/gpu), [#ai-infrastructure](https://daily.dev/tags/ai-infrastructure) [View this post on daily.dev](https://daily.dev/posts/nvidia-s-top-ai-chips-are-coming-to-a-centurion-data-centre-xjh4vshjt)