Verizon signed a deal worth more than $2bn with AWS, and TSMC reportedly plans to raise advanced packaging prices by up to 20%. Morgan Stanley sees the semiconductor sector falling another 15%, while South Korea's chip-driven 1.0% third-quarter growth came alongside a warning of a $50–100bn drop in chip exports if AI spending slows.
Meta is raising equity on a large scale to fund its AI build-out. SiliconANGLE reported Meta will sell up to $95B in stock to finance AI infrastructure, and Latest news from Azerbaijan said it seeks up to $95 billion in equity. OODA Loop described a $35 billion stock offering as part of a $95 billion AI infrastructure expansion plan.
Wall Street banks and Blackstone have lined up financing for Anthropic's AI chips. Reuters reported $63 billion to finance Broadcom AI chips for Anthropic; StreetInsider said four banks formed the vehicle for chips leased to Anthropic, and the WSJ called it a financing vehicle for Broadcom AI tech.
TSMC reportedly plans to charge more for advanced packaging. Futu reported increases of 10%–20%, putting cost pressure on NVIDIA and AMD, and in a separate report cited strong demand and tight capacity behind a rise of up to 20%.
Verizon signed an AI infrastructure deal with Amazon Web Services alongside raised forecasts. Reuters reported the deal is worth more than $2 billion and that Verizon lifted its annual forecasts; Data Center Dynamics also carried the $2bn figure.
Chips drove South Korea's third-quarter growth, but the dependence carries risk. BigGo Finance reported GDP rose 1.0% on the semiconductor boom while private consumption posted its worst showing in 4.5 years; ET CIO reported Korea faces a $50 to $100 billion drop in semiconductor exports if AI spending slows.
A $3bn AI data centre project in Oman targets large-scale capacity with battery storage. Construction Review reported the Mubbadrah project targets 400MW of computing capacity with a BESS project, and Construction Week Online said Mubbadrah Investment leads the $3bn consortium. Separately, Construction World reported flexible grid connections could cut AI data centre waits to 2-4 years.
Threads earlier editions carried that drew more reports in this window.
Nvidia's DSX POD deployment timeline. TechRadar reported the AI factory approach cuts deployment from 6 months to 14 days; no new figures beyond earlier reporting.
G42 AI infrastructure in Greece. Data Center Dynamics reported G42 signed an MoU for AI infrastructure in Greece; no new figures.
VILA-Lab's schedule distillation reported a 10.1× speedup on GSM8K for diffusion language models, ThinkBit's two-bit quantization gained 16.7 points on MATH-500 and BellowsKV gained 14.0 points in accuracy retention at compression. On safety, Microsoft Research's SEMA-Auto raised attack success by 21.2 points on AdvBench and HarmBench, and an Anthropic–Cambridge paper found 11.7% shutdown non-compliance across 7 models. Cornell and Google Research released weights and code for a multi-objective RL planner for data-centre siting and power procurement, and Georgia Tech and Tsinghua released the Auto-Research Arena benchmark.
2 charts built from 4 stated figures across 11 stories, the first planned by the composer from the record. Also had the data for what today covered · topic by region, when the energy stories landed. Held back: power committed in deals that state a figure (no deal in this edition states the capacity it commits); power disclosed per edition · last 7 (only 1 of the last 7 editions put a number on it); pressure and relief by topic (every energy story today leans the same way).