UK neocloud firm Nscale disclosed in documents tied to its US IPO that a ByteDance subsidiary called Spring accounted for $24 million of its $33 million 2025 revenue, giving Spring access to 2,304 Nvidia B200 chips at Nscale's Glomfjord, Norway facility. The ByteDance link surfaced not in Nscale's S-1 but in a supporting $105 million loan document with Macquarie, and the Financial Times reports the arrangement was legal under current US export rules.
tomshardware.com
· 2026-09-23
Hugo Vergnes built a config-driven training framework called little-lm and used it to train a 3.8-billion-parameter language model from scratch on 65 billion tokens, taking 43 hours on eight rented B200 GPUs at a total cost of $998. The model scored 0.384 on the CORE benchmark, outperforming Andrej Karpathy's nanochat d32 model, which cost about the same to train but scored 0.310.
hugovergnes.github.io
· 2026-09-10
A technical blog post walks through building a custom attention kernel for Nvidia's Blackwell B200 GPU entirely from scratch, using CUDA and some PTX code. Through a 14-step progression of optimizations illustrated with 60 diagrams, the kernel reaches 94.4% of FlashAttention-4's performance on 4K, 8K, and 16K sequence shapes, and is demonstrated by plugging it into a video-generation model.
iaroslavelistratov.github.io
· 2026-09-02
South Korea's government will roll out a free generative AI program called 'AI for All' starting with beta testing in September, expanding to the whole population later this year. Three consortia, including groups led by the nation's top telecom carriers and Kakao's operator, will build apps that connect to government systems for tasks like booking medical appointments, apartment hunting, and tax filing, with no token limits on usage.
techspot.com
· 2026-08-29
GoKawiil