According to @AnthropicAI, the company plans to expand its use of Google TPUs by securing approximately one million TPUs and more than a gigawatt of capacity in 2026, source: @AnthropicAI, Oct 23, 2025. The announcement specifies a 2026 timeline and quantifies scale at roughly 1,000,000 TPUs and over 1 GW, but does not include pricing, contract structure, or model details, limiting immediate financial modeling from the statement alone, source: @AnthropicAI, Oct 23, 2025. The statement mentions Google TPUs and does not reference cryptocurrencies or blockchain, and it provides no direct crypto-market catalyst in the disclosure, source: @AnthropicAI, Oct 23, 2025.

According to @DeepLearningAI, Alibaba released Wan 2.2, an open-weights video generation family that adds a mixture-of-experts architecture (source: @DeepLearningAI on X, Aug 26, 2025). Variants include a five-billion-parameter text/image-to-video model that can run on consumer GPUs (source: @DeepLearningAI on X, Aug 26, 2025). The MoE setup uses two experts, including a high-noise expert, while the second expert was referenced but not detailed in the provided source text (source: @DeepLearningAI on X, Aug 26, 2025). For traders, the open-weights release and consumer GPU compatibility are concrete datapoints to track for AI infrastructure themes in crypto markets, as broader accessibility to video AI tooling can influence developer activity and on-chain usage in decentralized compute and storage ecosystems, without implying price impact (context based on the same source details about open weights and consumer GPU support: @DeepLearningAI on X, Aug 26, 2025).