NAll news

All news

Wednesday47 items

  8. Yuchen JinOpinion
    AI score80

    DeepSeek's single cluster processes 155 trillion tokens per day, generating $38 in revenue per GPU-hour

    Yuchen Jin, citing data disclosed by DeepSeek, said that a single cluster currently processes 155 trillion tokens per day and generates $38 in revenue per GPU-hour (off-peak pricing). He noted that DeepSeek has not only open-sourced its inference engine but also disclosed revenue figures, though it did not specify which GPU models were used; DeepInfer supports both NVIDIA Blackwell and Huawei Ascend. Based on an estimated B300 rental price of $7 per GPU-hour, the gross margin is approximately 82%.

    Same story · 4 reports

    #English text is being prepared.#DeepSeek#Reasoning

  9. swyxOpinion
    AI score65

    English translation is being prepared.

    Quoted postDeepSeek · @deepseek_ai

    🚀 Introducing DeepInfer, our in-house inference framework, now fully open source! Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. We’re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable. 🔗 https://t.co/JGAik1kgk0 1/4

    Same story · 29 reports

    #Industry news#Reasoning#DeepSeek

  14. Latent SpaceOpinion
    AI score52

    SF Compute CEO Evan Conrad on Programmable Data Centers and the Compute Market

    SF Compute CEO Evan Conrad said on the Latent Space podcast that the compute market must first achieve physical delivery—that is, actually delivering usable GPU compute—before discussing financialization; otherwise, market manipulation and basis risk are likely to arise. He said SF Compute has already put 5% of its capacity on the order book, and that customers who spend $100 to buy capacity can on average recover $25 by reselling it, with hundreds of millions of dollars in trading volume already completed. He also said the company does not hold hardware in order to remain neutral, and that it already supports customers in partitioning InfiniBand fabric online themselves.

    Same story · 2 reports

    #Industry news#English text is being prepared.#English text is being prepared.

  15. ayi_ai_notesOpinion
    AI score52

    English translation is being prepared.

    Quoted postEnglish text is being prepared. · @aoaoao

    https://t.co/X8x8FyW6oR

    #English text is being prepared.#Open-source ecosystem#Agents

  16. peter_steinbergerOpinion
    AI score60

    English translation is being prepared.

    Quoted postVoidZero · @voidzerodev

    Introducing tsrs: a Rust port of the TypeScript compiler 🦀 We are not forking TypeScript! ◆ In close collaborating with the TypeScript team ◆ Up to 8x faster than TS 7.0 & ~70% less memory ◆ Runs in WASM, so it works in your browser & on Workers ◆ Soon part of the Oxc toolchain npx tsrs https://t.co/qmmxF3YnQy

    #Industry news#English text is being prepared.#Open-source ecosystem

  17. One Useful ThingOpinion
    AI score57

    Ethan Mollick reviews 30 months of AI development in charts

    On the release day of his new book Co-Existence, Ethan Mollick used charts to review the AI progress in the 30 months since his previous book Co-Intelligence was published. He said that since GPT-4, 20 models have successively taken the top spot, and 180 models were stronger than GPT-4 at release, 19 of which can run on modern smartphones; the METR long-task benchmark showed that the amount of human work AI can complete increased from 4 minutes in 2024 to 17 hours in April 2026, and the benchmark was subsequently discontinued because it could no longer measure accurately. He proposed a third mode of collaboration beyond centaur and cyborg: conjurer, that is, summoning agents, assigning tasks, and evaluating results. The charts in this article were completed by Opus 5.5 in Claude Projects orchestrating multiple agents.

    Same story · 2 reports

    #English text is being prepared.#Agents#English text is being prepared.

Tuesday42 items

  1. AYi AI NotesOpinion
    AI score52

    Altman Responds to the $1.4 Trillion Compute Bill: Better to Buy Too Much Than Too Little

    On the BG2 podcast, investor Brad Gerstner, in front of Microsoft's CEO, questioned how OpenAI could sign a $1.4 trillion compute bill on $13 billion in annual revenue. Sam Altman responded that if Gerstner thought the risk was too great and wanted to sell his shares, he could help find a buyer right away, and said he looked forward to the company going public so the naysayers could short it. The author sums up Altman's logic as: at this stage, the real danger is not buying too much compute and being unable to pay the bill, but stockpiling too little compute and being unable to build the next-generation model, and lists four supporting paths: ChatGPT subscription growth, building its own AI cloud service, developing its own AI consumer hardware, and fully automated scientific research; the author also points out that when Altman announced the AI cloud service, Microsoft's CEO was right beside him, the competition between the two sides is already out in the open, and closing a $1.4 trillion gap in four years remains an unproven gamble.

    Quoted postAYi · @AYi_AInotes

    English text is being prepared.

    #Industry news#OpenAI#Microsoft

  2. 36Kr · AIOpinion
    AI score53

    iFlytek's non-recurring net loss in the first three quarters was 763 million yuan, with its book profit turned positive by subsidies and equity gains

    In the first three quarters, iFlytek's revenue was 18.187 billion yuan, and net profit attributable to shareholders was 21.8872 million yuan, but non-recurring net profit showed a loss of 763 million yuan, more than double the 338 million yuan loss in the same period last year. The book turnaround to positive was mainly due to 471 million yuan in government subsidies, 370 million yuan in equity appreciation of invested companies, and 84 million yuan in asset disposal income, totaling about 785 million yuan. The company's R&D expenses in the first three quarters were 3.850 billion yuan, accounting for more than 21% of revenue, and it attributed the widened non-recurring loss to continued investment in the self-controlled large model route; in public domestic large model tenders in the third quarter, iFlytek ranked first in both the number and amount of bids won.

    #Industry news#English text is being prepared.#English text is being prepared.