According to Bloomberg, Microsoft plans to begin deploying a new generation of Maia 300 chips at gigawatt scales in data centers next year. It is expected that the number of token that can be processed per megawatt by these chips will be 40% higher than that of mainstream GPU systems. Microsoft stated that these chips will be prioritized for internal AI services rather than being leased out for computing power, in order to reduce dependence on NVIDIA GPU and control inference costs.

Meta's Superintelligence Lab released two new advanced models, Muse Wonder and Muse Wonder Full, on Monday, stating that this is the company's first attempt at cutting-edge models. Meta indicates that these two models perform better in benchmark tests and complex tasks, among which Wonder Full is only available to subscribers of Muse Code and is not provided through API.

Lowest Fee Bitcoin ATMs Announces the Launch of Over 400 Bitcoin ATM in the United States, with a Uniform 5% Fee; Supports Cash Purchases of Bitcoin, Ethereum, and USD-Backed Stablecoins USDT and USDC; Also Launches Online Pre-Registration to Shorten First-Visit Time.

PYTH rose nearly 20% after breaking through the $0.08148 resistance level during trading. The article states that the golden cross formed in September has strengthened the bullish outlook, with the next target resistance level at $0.13021; if $0.08148 is lost, it may fall back to $0.05589 and $0.03703.

The article states that even if AI company purchases a large amount of NVIDIA GPU and builds data centers, it may still not be able to operate at full capacity due to the local power grid's inability to provide sufficient electricity. Berkeley Lab predicts that by 2030, data center electricity consumption in the United States could account for 9.5% to 15.3% of the country's total electricity consumption, and transformers, transmission lines, grid connection approvals, and cooling systems are becoming important bottlenecks for the expansion of AI.