👀In Case You Missed It
Ilya Sutskever says the “age of scaling” is ending, arguing frontier AI progress will now come from pure research, efficiency breakthroughs, and new architectures rather than brute-force compute — a major philosophical pivot for long-horizon AGI labs
Anthropic claims AI could double U.S. labour-productivity growth, accelerating knowledge work, automating multi-step workflows, and reshaping services-sector economics at national scale.
NVIDIA hits back at the TPU wave, insisting Blackwell is still “a generation ahead” in flexibility, throughput and multi-model inference — signalling tension as Google, Meta, and China all explore non-GPU paths.
China blocks ByteDance from using Nvidia chips domestically, intensifying Beijing’s sovereign-silicon push and accelerating offshore model training as a workaround to U.S. export controls.
Read more here
🧠 Cloud & Infrastructure Watch
OpenAI confirmed a data breach at an analytics partner that exposed user names and emails (but not chat content), putting AI platform data governance and vendor risk squarely back in the spotlight.
Read MoreMIT’s new Iceberg Index estimates AI can already handle tasks worth ~11.7% of US wages, suggesting the true automation impact is buried in office and professional work rather than just headline tech layoffs.
Read MoreUniversity of Utah rolled out ChatGPT Edu campus-wide, a sign that entire universities are now standardising on AI platforms rather than leaving usage to individual departments.
Read MoreSAP launched an EU AI Cloud offering, giving European enterprises a region-locked environment for building and running AI workloads under stricter data-sovereignty and compliance rules.
Read MoreAI shopping assistants are going mainstream as Walmart, Amazon, Target, Google, Meta and OpenAI all ship shopping bots, turning retail into one of the first sectors to run on agentic AI at scale.
Read MoreCommentators continue to argue over whether there’s an AI bubble or not, but a common thread is emerging: the broader economy is becoming structurally dependent on AI capex and cloud spend.
Read MoreJeff Bezos’ Project Prometheus quietly acquired agentic-computing startup General Agents, reinforcing the idea that “workflow orchestration” and AI agents will be a core infra layer, not just a UX feature.
Read More
💻 Hardware & Semiconductor Moves
China’s tech giants are increasingly training models offshore to tap Nvidia chips, while regulators at home block firms like ByteDance from buying Nvidia GPUs — accelerating work on domestic alternatives.
Read MoreA Chinese startup founded by a former Google engineer claims its new TPU-style ASIC runs 1.5× faster than Nvidia’s A100 and 42% more efficiently, signalling a serious local push to loosen Nvidia’s grip.
Read MoreReports note that “AI trade” is splintering as Google’s TPUs and new Ironwood generation challenge Nvidia’s dominance, with both firms publicly downplaying the other’s advantages
Read MoreHP plans to cut 4,000–6,000 jobs by 2028 while ramping AI across its product lines, a classic example of hardware vendors using AI both as a growth story and a justification for restructuring.
Read MoreChinese planners warn of a speculative bubble in humanoid robotics, even as robots and AI begin to reshape the country’s industrial base — early signs of overbuild risk in yet another AI-adjacent hardware segment.
Read MoreNew commentary pieces argue that Nvidia may be overspending and over-extended, while short-sellers like Michael Burry openly frame AI hardware as the next big bubble to bet against.
Read More
