If general models take over the robot “brain,” manufacturers' buying criteria shift toward reliability, dexterity and integration.
AI NEWS
Manufacturing AI News
Chinese and English are the same day. This page lists every item, each with the original link, the author, and the engagement.
Latest day
November 2, 2026
5 items for 2026-11-02, ranked by combined engagement.
Top 3
Unconfirmed Hundreds of thousands of new GPUs per month at the top shapes downstream API pricing and availability, and explains persistent memory and power crunches.
The memory crunch raises prices for servers, PCs and phones; businesses buying local AI hardware should budget for longer lead times.
All
Freda Duan after talking to 50 robotics insiders: “Astra is robotics' GPT-3 moment,” with data budgets reportedly nearing $1B per frontier labUnconfirmed
If general models take over the robot “brain,” manufacturers' buying criteria shift toward reliability, dexterity and integration.
CollapseFreda Duan after talking to 50 robotics insiders: “Astra is robotics' GPT-3 moment,” with data budgets reportedly nearing $1B per frontier lab
Investor Freda Duan shared 10 takes after talking with 50 people at the robotics frontier, linking a Robonomics essay. She argues OpenAI's Astra is robotics' “GPT-3 moment” and that robotics is becoming just another vertical for frontier labs, leaving robotics foundation-model startups a narrower skills-and-motor-control niche. She says word is OpenAI reorganized its 30–40-person robotics team around Astra, with similar moves at Tesla Optimus (hearsay). She also cites million-hour data procurement programs, some possibly reaching tens of millions of hours, talk of budgets near $1B per frontier lab, and notes a 100B-parameter edge model needs about 50GB just for 4-bit weights — a possible memory supercycle driver.
Musk names two new SpaceX supercomputer clusters — Gun Range (late Nov) and Spam (Dec), ~420K GB300s combined (secondhand)
Hundreds of thousands of new GPUs per month at the top shapes downstream API pricing and availability, and explains persistent memory and power crunches.
CollapseMusk names two new SpaceX supercomputer clusters — Gun Range (late Nov) and Spam (Dec), ~420K GB300s combined (secondhand)
Sawyer Merritt reports that Elon Musk has named two upcoming SpaceX supercomputer clusters: Gun Range, due in late November, and Spam, due in December. By his account the two together hold about 420,000 Nvidia GB300 GPUs, or roughly $35 billion-plus in GPUs. The figures are Sawyer's summary rather than a full official SpaceX disclosure, and the dollar amount is his estimate. If accurate, it shows compute build-outs are still accelerating, and that shortages of GPUs, memory and power are unlikely to ease soon.
Bloomberg: Altman meets Samsung's top brass in Korea on memory chips and AI infrastructure — no new supply deal announced
The memory crunch raises prices for servers, PCs and phones; businesses buying local AI hardware should budget for longer lead times.
CollapseBloomberg: Altman meets Samsung's top brass in Korea on memory chips and AI infrastructure — no new supply deal announced
Bloomberg reports OpenAI CEO Sam Altman met Samsung Group's top executives in South Korea on Saturday to deepen cooperation on memory chips and AI infrastructure as global shortages intensify. Tom's Hardware says about 50 senior Samsung executives attended; Samsung confirmed the talks built on an earlier letter of intent to supply memory for Stargate but emphasized AI transformation and future business rather than announcing a new supply agreement. The article cites Counterpoint: Samsung and SK hynix held about 63% of global DRAM revenue and roughly 83% of HBM revenue in Q2; TrendForce sees DRAM supply 1–2% short this year, and Micron has warned the shortage could last into 2028.
Bloomberg: Hon Hai is counting on US expansion and an OpenAI partnership to capture half the AI server market within two years
AI server production concentrating in the US and contract makers tying directly to labs will reshape order flows and tier-2/3 supplier opportunities.
CollapseBloomberg: Hon Hai is counting on US expansion and an OpenAI partnership to capture half the AI server market within two years
Bloomberg reports that Hon Hai (Foxconn), a key Nvidia production partner, is counting on steady US expansion and a partnership with OpenAI to hit its goal of capturing half the AI server market within two years. The headline references Chairman Young Liu on AI demand; specific capacity and partnership terms could not be verified from the full article here. It shows AI server assembly shifting to the US and contract manufacturers tying up directly with model companies.
NVIDIA: AI agents now steer the Allen Telescope Array to search for alien signals, processing 2.4 Tbps in real time
Agents directly operating instruments applies equally to factory lines, test rigs and labs — the key is exposing equipment state as one agent-readable source of truth.
CollapseNVIDIA: AI agents now steer the Allen Telescope Array to search for alien signals, processing 2.4 Tbps in real time
NVIDIA says the SETI Institute and Breakthrough Listen have brought agents into daily observing at the Allen Telescope Array in Northern California. Its Oct 30 blog says agents can steer antennas, calibrate and focus, fix data errors and run AI models on the stream from a single natural-language request; Holoscan's DAQIRI library pulls terabits per second of sensor data straight into GPU memory, while the SETI Institute's CyberEther framework does signal processing with cuFFT and cuBLAS. A web orchestrator called Nexus acts as the agent harness, and the system handles 2.4 Tbps on RTX PRO 6000 Blackwell GPUs with on-prem inference. The blog says the pipeline achieved the first real-time AI detection of a pulsar in the Crab Nebula, and the team is now evaluating open models such as Nemotron.