In the past hour, total liquidations across the network reached $19.78 million, with $2.39 million from long positions, with $17.39 million from short positions. BTC liquidation $746,445.69, ETH liquidation $4.56 million.

According to Bijie.com, CECPort announced that it expects its net profit attributable to shareholders of the listed company for the first half of 2026 to be between 500 million and 530 million yuan, representing a 176.43%-193.01% increase compared to 181 million yuan in the same period last year. Net profit excluding non-recurring gains and losses is also expected to be between 500 million and 530 million yuan, representing a 186.25%-203.42% increase compared to 175 million yuan in the same period last year. The performance growth is mainly due to strong demand in fields such as artificial intelligence, data centers, and advanced computing, as well as the continued rise in prices of core products such as memory chips, which has driven a significant increase in operating revenue.

According to CoinWorld, the UK-based AI Security Institute points out a significant blind spot in current mainstream AI agent testing: the fixed computing power cap assessment method severely underestimates the true capabilities and iteration speed of models. The research team tested the performance of several cutting-edge large-scale models in benchmarks such as cybersecurity, software engineering, and mathematics, finding that agent performance continuously improves with increasing computing power during testing. In network attack and defense tests, when the computing power budget increased from 2.5 million tokens to 50 million tokens, the upper limit of the complexity of the tasks that the agent could overcome surged from 2 hours to 14 hours. The new model's utilization efficiency under sufficient budget was significantly higher than the old model, and the evolution trend of cutting-edge capabilities was approximately 60% steeper than in low-computing-power tests. However, in fields lacking immediate feedback, such as healthcare, increasing computing power does not improve agent performance, and low-budget assessments may lead decision-makers to underestimate the actual risks of AI agents.

According to CoinWorld, Deutsche Bank analysts believe that Meta plans to sell its idle AI computing power externally. By the end of 2027, Meta's AI-related computing power is projected to reach 8 to 11.5 GW, of which approximately 1.2 to 2.7 GW will be available for sale. Assuming 75% of the capacity is sold and the annualized revenue per GW is between $1 million and $1.5 million, Meta could generate between $9 million and $30 million in additional revenue in 2027, with a baseline scenario of approximately $17.5 million.

According to Odaily Planet Daily, SolanaFloor posted on the X platform that Solana has added over $540 million in RWA value in the past 7 days, bringing its total RWA value to a record high of $3.62 billion.