Tag
The author argues that accumulating efficiency gains will drive consumer hardware prices up while data center hardware prices fall, quoting a report of a capacity surge with providers offering training compute at heavily discounted rates.
Creators complain that Hongguo’s revenue sharing for comic dramas has dropped sharply, with 10 million views earning only about 500 yuan—barely enough to cover AI computing costs—and are seeking resources and opportunities to exchange experiences on taking comic dramas overseas.
Rohan Paul cites Epoch AI data showing total compute of AI chips in use grows 3.4x yearly, but warns that if new chip purchases slow to 2.5x growth, overall compute growth will follow.
Workhorse Group and UK-based RHAELM announced a strategic partnership to commercialize TerraPod, Workhorse's modular AI data center platform, including a $10M spec unit agreement and a non-binding LOI for a potential 150-unit initial order.
Modders have unlocked hidden AI processing capabilities on Nvidia RTX 20, 30, and 40 series GPUs, nearly doubling performance in some benchmarks by accessing disabled hardware resources.
OpenAI reportedly plans to purchase Google's TPUs to expand AI compute capacity and reduce dependence on Nvidia, while major cloud players prepare to spend up to $600 billion on AI infrastructure this year.
Epoch AI launched the AI Chip Imports explorer, estimating how much AI compute (in H100-equivalents) has arrived in each country since 2022, finding the US received 72% of global AI compute while China lags at 10% despite producing more chips.
General Compute 获得 1 亿美元 A 轮融资，由 Blackbird Ventures 和 Abstract Ventures 领投，估值升至 8 亿美元。该公司专注于使用 SambaNova、Positron、Cerebras 等推理芯片，并通过改造比特币矿场和租用现有数据中心来提供 AI 算力。
Commentary on the reported Anthropic–SpaceX compute deal, arguing the $3.1B/month rate isn't guaranteed through 2029 but that big compute providers and NVIDIA stand to win while neoclouds may suffer.
SpaceX 和 Tesla 已成立特别委员会讨论合并，SpaceX Q3 超智能业务营收达 75.6 亿美元（大部分来自向 Anthropic 和 Google 出租算力），而 Tesla 则连续第二季度利润下滑，凸显 Musk 推动合并背后的 AI 与控制权动机。
Modular 宣布将 MAX 随 Modular 26.7 源码开放，以打破 AI 对单一硬件厂商或芯片的锁定，推动 AI 算力民主化。
Elon Musk replies that achieving online and reliable compute at scale is extremely difficult, highlighting ongoing AI infrastructure challenges.
A commentator shares their shift from skepticism to buying $SPCX stock, citing growing evidence that SpaceX is serious about its AI and orbital datacenter plans despite exaggerated-seeming timelines.
報告指出中國 55% 的 AI 算力 token 消耗用於短劇與影音生成，僅 15% 投入研發，引發「過半算力造爽劇」討論，字節跳動則否認該比例。
据报道，Nscale 与 Anthropic 正就在一个月前 450 亿美元算力协议之外再签一笔 450 亿美元的算力交易进行谈判，若达成将进一步加码 Anthropic 近一年来巨额算力采购的步伐。
Elon Musk shares a post about the experience of trying to rent AI/superintelligence compute, highlighting ongoing challenges in securing sufficient compute resources.
Intel SVP Cho Min-sung confirmed that Razor Lake-AX SoCs will launch in 2028 with 700 TOPS of AI compute, a 3.88x increase over Panther Lake, targeting premium devices and the Physical AI market against NVIDIA Jetson and AMD Ryzen Halo platforms.
Elon Musk replied that SpaceX may need more than 10GW of AI compute by the end of 2027, as Nvidia now counts SpaceX among its cloud-scale customers and both companies' engineers are reportedly designing a server for space.
Character.ai的破产文件显示其AI商业模式根本无法盈利：每月AI算力开销高达1380万美元，而订阅收入峰值仅280万美元，即便大幅削减成本后每月仍亏损800万美元，成为AI泡沫破裂的早期标志性事件。
A Twitter exchange speculating whether GPUs will eventually rent more GPUs to train improved versions of themselves, while noting that insufficient compute and token generation will remain the bottleneck for AI progress for years.