**********Hardware & infra: Power shortages/Broadcom-Google-Anthropic TPU/Nvidia Siemens/Jetson/AMD Ryzen/Firmus $505M/Marvell/Broadcom ASIC/Anthropic TPU/Google/Oracle/Fireworks/Intel/Ironwood/IBM Arm/China chips/CapEx/Akamai/custom/HBM/AI design/inference shift/Meta spend/Nutanix buy-stacks**********
Key Questions
What TPU deals did Anthropic secure?
Anthropic secured multi-GW TPU deals with Google and Broadcom through 2027, with Broadcom's deal at a $30B run-rate. Broadcom expanded AI chip deals for TPU capacity with Google until 2031.
What is causing power shortages for data centers?
Power shortages and a $650B capex crisis are stalling data centers, with hyperscalers facing over $1T in costs. Firmus raised $505M to expand AI data center infrastructure.
What is Broadcom's AI revenue outlook?
Broadcom reported a $73B backlog and projects $8.4B in Q1 2026 AI revenue. It inked expanded AI chip deals with Google and Anthropic.
What shift is happening in AI inference chips?
Hyperscalers are shifting to TPUs, ASICs, and Groq for inference. Nvidia focuses on SchedMD, Siemens, Jetson Thor, and RTX for local use.
What are AMD's recent AI advancements?
AMD's Ryzen AI Max chips now have a worthy Linux partner with staggering benchmarks. They support AI workloads efficiently.
What is Fireworks' capacity?
Fireworks offers 15T capacity for AI inference.
Why will neoclouds buy AI stacks?
Most neoclouds, sovereigns, and enterprises will buy, not build, their AI stacks, as per Nutanix strategy.
What verification solution did Siemens and Nvidia unveil?
Siemens and Nvidia unveiled an AI chip verification solution capable of trillions of cycles in days.
Anthropic multi-GW TPU deals Google/Broadcom '27 ($30B run-rate); power crisis stalls DCs ($650B capex/$1T+ hyperscalers/Firmus $505M); Broadcom expansions (Google 2031/$73B backlog, $8.4B Q1'26 AI rev); inference chip shift (hyperscalers TPU/ASICs/Groq); Nvidia SchedMD/Siemens/Jetson Thor/RTX local; AMD Ryzen AI; Fireworks 15T; Nutanix buy-not-build for neoclouds.