Jiang Zhuo'er, founder of Laibit Mining Pool, stated that there is currently a 90% confidence that the bear market has ended, and ETH's performance in this round may surpass BTC. Jiang Zhuo'er has bought back ETH sold between $1738 and $1931 at a stop loss of $2100, and sold 50% of the spot ETH when it rose to $2525, with a stop loss set at $2550. For short selling funds, Jiang Zhuo'er plans to buy all BTC if it falls back to the range of 67000 to 72000 US dollars, otherwise buy at the current price no later than the end of October, and emphasizes the importance of execution and risk control in trading.

US President Trump stated in Myrtle Beach, South Carolina that the United States has not built a new automobile factory in 40 years, and there are now more factories under construction than at any time in history. Trump said tariffs are his favorite word and mentioned that last month the US inflation rate fell at the fastest rate in years, and now the number of US jobs is higher than at any time in US history.

Matrixport is associated with a giant whale that closed 40000 ETH and gained a profit of 9.9 million US dollars. The giant whale currently holds a long position of 80000 ETH and 500 BTC, with an unrealized profit of 22.9 million US dollars. Its total profit and loss has changed from a loss of 92.5 million US dollars to a profit of 32.8 million US dollars. （Lookonchain）

According to Ai Yi's monitoring, a bit associated entity that long 120000 ETH has taken profits for the first time in four months. Among them, addresses 0x6c8... d84f6 reduced their position by 38000 ETH and profited $9.407 million. Currently, the four addresses still hold multiple orders of 82000 ETH, with a floating profit of over $20 million.

Microsoft CEO Satya Nadella shared on-site photos of data centers, and the first batch of NVIDIA Vera Rubins from Microsoft has arrived. Nvidia confirms that Vera Rubin has entered the full-scale production phase. Vera Rubin is a new generation rack level AI computing platform after Blackwell, with a set of NVL72 integrated with 72 Rubin GPUs and 36 Vera CPUs. Nvidia claims that compared to GB200 NVL72, the platform reduces the inference cost per million tokens by about one tenth and the number of GPUs required to train MoE models by a quarter.