- In The Loop AI
- Posts
- NVIDIA is moving data servers to the US
NVIDIA is moving data servers to the US
Meta to resume AI training of European public users
Algot Persson
April 15, 2025
In partnership with
In Today’s Issue:
NVIDIA is moving data servers to the US
Meta to resume AI training of European public users
Read time: 3 minutes
NVIDIA is moving data servers to the US
NVIDIA is moving some of its data servers to the USA to mitigate its supply chain risks. It will produce the new Blackwell B100 and B200 AI GPUs, with TSMC manufacturing them inside the USA as well. This move of re-shoring is expected to take 12–15 months before it’s fully operational. The GPUs will be fabricated at TSMC’s advanced Fab 21 facility in Phoenix, Arizona, using a combination of 4nm (N4P) and 3nm (N3) process technologies. In parallel, NVIDIA is partnering with Foxconn and Wistron to assemble full AI supercomputers in Texas, further localizing the production of its end-to-end compute stack. The company’s domestic expansion comes in response to intensifying demand for generative AI infrastructure and increasing geopolitical concerns, particularly around Taiwan’s chip supply. By anchoring its high-value chip production and system integration in the U.S., NVIDIA aims to create a more resilient and politically insulated supply chain while maintaining technological leadership in the global AI arms race.
Jensen Huang
Technical details:
Total investment: $500 billion over four years
Chip production location: TSMC Fab 21, Phoenix, Arizona
Assembly location: Foxconn and Wistron facilities, Texas
Chip models: Blackwell B100 and B200 AI GPUs
Production status: TSMC has begun volume production of Blackwell chips
Manufacturing ramp-up: Supercomputer assembly scheduled to scale in 12–15 months
Product application: AI datacenters, hyperscale compute, and inference systems
Manufacturing partner nodes: 4nm N4P process technology for B100, 3nm N3 for B200
Find out why 1M+ professionals read Superhuman AI daily.
In 2 years you will be working for AI
Or an AI will be working for you
Here's how you can future-proof yourself:
Join the Superhuman AI newsletter – read by 1M+ people at top companies
Master AI tools, tutorials, and news in just 3 minutes a day
Become 10X more productive using AI
Join 1,000,000+ pros at companies like Google, Meta, and Amazon that are using AI to get ahead.
Meta to resume AI training of European public users
Meta
Meta is resuming its AI training efforts in Europe after a pause last year due to legal pressure from privacy advocates. Starting this month, the company will begin using publicly available posts and comments from adult users in the EU to train its AI models, including Meta AI, which recently launched across European markets.
In a blog post, Meta confirmed that user interactions with its AI assistant — such as questions and prompts — will also be used to improve model performance. Private messages, however, are excluded from training.
This move follows a December decision by EU privacy regulators affirming that Meta’s data usage approach meets GDPR requirements. The company had previously faced complaints from NOYB, the privacy group led by Max Schrems, which urged regulators to halt Meta’s model training.
Meta also noted that Google and OpenAI have already used European public data to train their AI systems. EU users will soon start receiving notifications and have the option to opt out via an objection form, which Meta says it will honor.
How was today's newsletter? |
Reply