Jukan@jukan05Fun fact: the combined capex of the four major U.S. hyperscalers has now surpassed $700 billion. That’s an increase of $300 billion year-over-year, or roughly 70% growth.Opens with an observation
Jukan@jukan05Wow, this is genuinely awful. AI accelerator development takes two years until launch, and yet they evaluate the performance of the people working on it every six months? This is exactly why MTIA is shit.Opens with an observation
Jukan@jukan05Kim Yong-beom, Chief of Staff for Policy: “Samsung Electronics and Broadcom have signed an MOU covering the supply of $200 billion (KRW 290 trillion) worth of advanced memory semiconductors over the next five years, as well as foundry cooperation for AI chip production.”Opens with an observation
Jukan@jukan05Very interesting. China’s big tech companies are renting data centers in Southeast Asia to train their AI models. DeepSeek is still optimizing and developing Huawei’s AI chips domestically in China. For inference, they are increasingly relying on Chinese-made chips.Opens with an observation
Jukan@jukan05I’m curious whether Apple’s FFN NAND-like approach reduces the mobile DRAM requirement needed for on-device AI. If so… why doesn’t Nvidia use this kind of technology? Wouldn’t that mean the 128GB in N1X is overkill?Opens with an observation
Jukan@jukan05Microsoft is considering an investment of less than $10 billion in OpenAI. Amazon is considering an investment of up to $20 billion in OpenAI. NVIDIA is considering an investment of up to $30 billion in OpenAI.Opens with a number