Nvidia's Week: Taiwan HQ Milestone & AI Vision

Weekly Nvidia Brief — November 12, 2025

A concise roundup of Nvidia’s past 7 days, with clear links to the company’s broader “AI factory” vision and what it means for AI overall. Slight focus on GPU competitors, as requested.

  1. Taipei clears path for Nvidia’s Taiwan HQ (decision week underway)
  • What’s new: Taipei’s deputy mayor moved to end a land-use contract by Friday, Nov 14, to free the Beitou–Shilin site for Nvidia’s planned Taiwan headquarters; council review targeted Nov 12. (focustaiwan.tw)
  • Nvidia’s bigger vision: A flagship base near TSMC and key suppliers strengthens supply-chain coordination for chips, packaging, and networking—the backbone of Nvidia’s “AI factory” rollout across regions. (focustaiwan.tw)
  • AI impact: More local engineering and partner support in Taiwan should speed system design and delivery for global AI capacity. (taipeitimes.com)
  1. Huang: Blackwell demand “very strong”; wafers requested from TSMC; no China sales talks
  • What’s new: In Taiwan, Jensen Huang said demand for Blackwell is “very strong,” asked TSMC for more wafers, noted memory remains tight, and reiterated there are no active discussions to sell Blackwell to China. (reuters.com)
  • Nvidia’s bigger vision: Securing more foundry supply and memory underpins Nvidia’s full‑stack expansion (GPU, CPU, networking) and keeps Blackwell ramps on track for AI factories. (reuters.com)
  • AI impact: More Blackwell supply accelerates training and inference availability for frontier and agentic models outside China, reinforcing a bifurcated compute market. (reuters.com)
  1. AMD turns up the heat: bullish 2030 targets and OpenAI tie-up color Analyst Day
  • What’s new: At its Nov 11 Analyst Day, AMD projected a steep ramp in data‑center chips and highlighted a multiyear OpenAI deal, signaling a more aggressive push against Nvidia in AI accelerators. (reuters.com)
  • Nvidia’s bigger vision: A stronger AMD keeps pricing and innovation pressure high—pushing Nvidia to differentiate on system‑level performance (NVLink/Spectrum, CUDA‑X, enterprise stack) rather than chip specs alone. (reuters.com)
  • AI impact: Healthier multi‑vendor supply should add total compute to the market, easing scarcity and broadening access for model training and deployment. (reuters.com)
  1. Google’s Ironwood TPU goes GA, targeting the “age of inference”
  • What’s new: Google moved its Ironwood (7th‑gen) TPU into general availability last week, emphasizing large‑scale, efficient inference. (blockchain.news)
  • Nvidia’s bigger vision: Competitor inference platforms nudge Nvidia to sharpen its end‑to‑end offering—from Blackwell training to cost‑efficient inference on both data center GPUs and RTX edge. (blog.google)
  • AI impact: With TPUs and GPUs competing, customers get more options to serve ever‑larger models at lower cost and power, hastening AI adoption in production. (blockchain.news)
  1. What to watch next (near-term catalysts)
  • Nov 19 earnings (Q3 FY26): bookings, Blackwell/GB300 shipment cadence, and commentary on export rules and supply. (nvidianews.nvidia.com)

Quick take

  • Nvidia spent the week shoring up supply and regional presence (Taiwan HQ steps; TSMC wafers), while rivals sharpened their pitches (AMD Analyst Day; Google TPU GA). This dynamic supports Nvidia’s long‑term “AI factory” strategy but also signals a more pluralistic compute landscape—good for the overall pace and affordability of AI. (reuters.com)

Sources are linked inline.