According to Satya Nadella on X (via his official post), Microsoft’s Fairwater datacenter in southeastern Wisconsin is going live ahead of schedule, integrating hundreds of thousands of NVIDIA GB200 GPUs into a single seamless cluster designed for AI training and inference at unprecedented scale. As reported by Nadella, Fairwater connects the GB200 fleet with fiber long enough to circle the Earth 4.5 times and is engineered to deliver 10x the performance of today’s fastest supercomputer, enabling day‑one jobs across thousands of GPUs through a co‑designed compute, network, and storage architecture. According to Nadella’s post, the site uses a closed‑loop liquid cooling system requiring zero operational water post‑construction and is matched 100% with renewable energy, addressing sustainability for high‑density AI compute. As stated by Nadella, Microsoft added over 2 gigawatts of new capacity last year and is building multiple identical Fairwater sites across the US and over 100 global datacenters to power model training, test‑time compute, RL tuning, and real‑time inference at scale. For enterprises, according to Nadella, this scale unlocks faster foundation model training, larger context windows, and lower latency inference, creating opportunities in generative AI platforms, AI‑accelerated R&D, and large‑scale multi‑agent workloads.

According to SawyerMerritt on X, Elon Musk stated that Tesla's AI4 hardware is sufficient to achieve better-than-human safety for Full Self-Driving (FSD), citing Optimus and Tesla's supercomputer clusters as enabling factors (source: Sawyer Merritt post referencing Elon Musk on X). According to Elon Musk on X, this implies current AI4 Tesla owners could see substantial FSD performance and safety improvements without immediate hardware upgrades, which may accelerate feature rollouts and fleet-wide validation. As reported by Sawyer Merritt, the emphasis on in-house supercomputer clusters suggests Tesla will continue scaling end-to-end neural networks and video training pipelines, reinforcing a vertically integrated strategy with potential cost efficiencies and faster iteration cycles for autonomy software.