AI Reading for Tuesday October 14

Major vendors and hyperscalers have adopted an annual AI-accelerator release cadence through the late 2020s, with AMD, Nvidia, Google, Amazon, Meta, Microsoft and Intel announcing roadmap chips including AMD MI430X (H2 2026) and Nvidia Rubin (late 2026). - Tom's Hardware

Data factory of the near future: Nvidia unveils Vera Rubin NVL144 and Kyber racks, open standards for 800V DC racks with future 576‑GPU clusters, gains adoption for its Ethernet standard with ~1.6Tbps per port. - SiliconANGLE

Nvidia starts shipping DGX Spark desktop AI “supercomputers” for consumers… can probably run some form of qwen3-235b-a22b, if you get 2 then maybe can run 405b parameter models. - The Verge

ASML warns of delays after China restricts rare earth exports - ctech

AMD will provide Oracle with 50,000 MI450 AI GPUs delivered in Helios rack servers (with next‑generation Epyc Venice CPUs and Pensando Vulcano networking), with deployment starting in the second half of 2026 and extending into 2027. - Yahoo Finance

Google will invest $15 billion through 2030 to build a 1‑gigawatt data center and AI hub on India's Bay of Bengal, in partnership with Bharti Airtel and AdaniConneX, its largest investment in India and largest outside the U.S. - TechCrunch

OpenAI and Buenos Aires-based Sur Energy signed a nonbinding letter of intent to build "Stargate Argentina," a proposed £19.5 billion ($25 billion) AI data centre with up to 500 MW of compute capacity under Argentina's RIGI investment regime. - www.ibtimes.co.uk

OpenAI is working with SoftBank-owned Arm to develop a CPU for an AI chip tied to a Broadcom effort, with TSMC set to manufacture the chip as part of OpenAI’s plan to build custom silicon for its data centers. So they will bake-off Nvidia, AMD and their own chips. - The Information

The two companies said Monday that they’re jointly building and deploying 10 gigawatts of custom artificial intelligence accelerators as part of a broader effort across the industry to scale AI infrastructure. Financial terms weren’t disclosed. - CNBC

Does Broadcom's bet on OpenAI exposes the company to concentrated strategic and financial risk? - The Wall Street Journal

Financial details not disclosed, it's just a press release and it boosts the stock and they are not stupid, should be pay as you go, in case at the end Nvidia blows it away.

As others catch up on models, OpenAI turns to deals in attempt to lock in dominance - Bloomberg

A proposed class action filed Monday alleges Microsoft violated antitrust laws by restricting compute to OpenAI after a $13 billion investment, forcing OpenAI to pay higher processing costs and causing ChatGPT to be priced up to 200 times above competitors. - Bloomberg Law

'AI users' sue Microsoft in antitrust class action over OpenAI deal - Reuters

Former Apple CEO John Sculley said OpenAI+Jony Ive might be "the first real competitor" to Apple, asserted Apple has not been strong in AI compared with OpenAI, Google, Meta and Amazon, and urged an AI-driven shift from apps to agentic services. - NDTV Profit

A leaked milestone table shared on X says Google plans a Gemini 3.0 announcement on October 22 - Android Authority

Gemini one-shots a Mac HTML interface. - Reddit

Localities fight data centers over impact on power, water; Michigan township settles AI data center suits, caps water use, secures millions amid growing resistance to $1T U.S. data center expansion. - Tom's Hardware

Oracle announced general availability of Oracle AI Database 26ai and Autonomous AI Lakehouse, embedding native AI functions including vector search, ONNX/LLM and Apache Iceberg support, available on OCI, AWS, Azure, GCP and Exadata Cloud@Customer, with free upgrades from 23ai. - SiliconANGLE

Microsoft announced MAI-Image-1, an in-house text-to-image model it says generates photorealistic images faster than rivals; LMArena ranks it ninth, and Microsoft will gather feedback before integrating it into Copilot, Bing Image Creator and other products. - BGR

Microsoft warns of 'shadow AI' after a UK survey found 71% of employees have used unapproved consumer AI at work, 51% still do, with 49% for workplace communications, 40% for reports, 22% for finance, and 32% citing privacy concerns. - The Register

Most investors told a Bank of America survey that AI stocks are in a bubble. - Bloomberg

Ex-Intel CEO Pat Gelsinger calls AI a bubble with limited business payoffs, but might not pop for several years. - Business Insider

Hinton: cuts to basic-science funding and threats to withdraw federal research support from universities risk costing the US its AI edge to China within 20 years. - Business Insider

Kenneth Rogoff argues that AI-driven growth will raise capital's share and reduce labor's share of output, so labor-derived tax revenues will grow less than GDP, making AI unlikely to generate equivalent public-revenue increases and leaving rich countries with debt. - Project Syndicate

A survey found more than one in three Americans—nearly half of those under 35—have used AI chatbots for medical concerns, which can aid symptom research, self-management and care navigation but cannot replace professional diagnosis or treatment. - The Washington Post

ChatGPT can serve as a personal shopper and deal hunter by providing personalized product recommendations and price comparisons when given detailed prompts, though users should verify results and expect further integration with real-time data and privacy considerations. - The Washington Post

Visa launched the Trusted Agent Protocol to cryptographically verify AI shopping agents using HTTP Message Signature (with Cloudflare), available now on GitHub, after agentic traffic surged 4,700% with traditional fraud running $40B annually (Oct 2022–Sep 2023). - VentureBeat

California Gov. Gavin Newsom signed SB 243, the first U.S. state law regulating AI companion chatbots, effective January 1, 2026; it requires age verification, AI labeling, self-harm/crisis protocols, minor protections and penalties up to $250,000 per illegal deepfake offense. - The Verge

Newsom vetoes AI bill that would have barred use of potentially harmful chatbots by minors, said it might have resulted in banning all minors from all chatbots. - SFGate

California Gov. Gavin Newsom vetoed SB 771, a bill by Sen. Henry Stern that would have fined social platforms up to $1 million (doubled for minors) for intentionally promoting content violating state civil‑rights laws. - Politico

Fetii launched NeuralShuttle, an AI-powered shuttle that trains on historical trip data to predict event-driven demand and automatically dispatch 15-seat vans to stadiums and festivals, with riders reserving via app or walking on and paying by QR. - Forbes

If 15-passenger vans can use the bus lane on 2nd Ave and you can call them like an elevator saying where you are going and they can optimize the pickups and dropoff, that sounds like a big win.

A prominent beer competition introduced an AI-judging tool without warning, used judges' evaluation notes to train the system, plans a consumer recommendation app, and threatened legal action against a judge who published an open letter criticizing AI. - 404 Media

Tonal is pursuing growth through retail expansion and AI-powered Pilates offerings. - Bloomberg

'Digital twin' market research studies use LLMs in place of real consumers and are said to be equally accurate. - VentureBeat

Robotin's R2, billed as the world's first fully autonomous carpet-cleaning robot, combines deep vacuuming, three-stage automated washing and intelligent drying, cleans about 400 sq ft per hour, and is offered to Kickstarter backers for $799 with March 2026 shipping. - New Atlas

Follow the latest AI headlines via SkynetAndChill.com on Bluesky