supercomputing

SpaceX plans to build two supercomputing parks in Starbase, Texas, and Cochise County, Arizona, with a total investment of over $74 billion. Among them, the Starbase project is expected to invest over $24 billion, while the Arizona park is projected to invest about $50 billion. SpaceX's AI business contributed more than half of the company's revenue in the third quarter.The Arizona park still needs to obtain approval from Cochise County and is expected to create about 5,500 construction jobs and 1,000 permanent jobs, using dry cooling and not planning to use industrial groundwater. A petition with 2,400 signatures has called for officials to deny the issuance of permits, and a public hearing will be held locally in December. The Starbase facility will generate its own electricity and will not connect to the public power grid, with the relevant planning covering approximately 444 acres, including a gas power plant and supercomputing facilities; the project has already created over 3,000 construction jobs and is expected to support more than 500 permanent jobs after it becomes operational.

SpaceX plans to invest over 24 billion dollars in the Rio Grande Valley for the construction of new supercomputing infrastructure.

According to IT Home, the national supercomputing internet has officially launched the DeepSeek-V4-Flash-0731 model API calling service and model files.This version has undergone extensive post-training, significantly enhancing the agent's instruction-following capabilities, with performance in multiple benchmark tests comparable to the strongest closed-source models. Enterprises and developers can access it with one click without the need for environment configuration.The platform's AI community has gathered over 1,700 open-source large models and has become the first national resource pool of super-intelligent integrated computing power at the level of 100,000 cards.

The China National Supercomputing Internet has announced the launch of a new round of token giveaways. This event is open to all users of the platform, offering a limited-time free distribution of up to 30 million tokens per individual to lower the experience threshold for research-specific agents like "lobster" SClaw.In addition, supercomputing internet users will enjoy a special renewal price of 0.1 yuan per million tokens, which will be extended until April 6.

The National Supercomputing Internet announced that its OpenClaw service has integrated with Feishu and WeChat Work. Users can flexibly choose from various model API calling services based on task scale and access them through the Feishu and WeChat Work clients.Currently, the Supercomputing Internet's OpenClaw has pre-configured various large models (such as MinMax-M2.1, MinMax-M2.5, Qwen-235B, etc.) to provide high-performance, low-cost inference computing power.

ChainCatcher news, the Hong Kong listed company International Commercial Settlement Holdings Limited (00147.HK) has signed an agreement through its joint venture Keen Golden Limited with Bitmain Technologies Georgia Limited and Tecnoenergia Capital Paraguay S.A. to purchase a total of 1,863 HASH supercomputing servers, with a total price of approximately 10.87 million USD. The buyer can choose to pay in cash or with Bitcoin collateral, and the expected delivery time is by the end of September, with a one-year warranty for the servers.

ChainCatcher news, according to the "Supercomputing Internet" public account, the DeepSeek-R1, V3, Coder and other series models have been successively launched on the national supercomputing internet platform.The national supercomputing internet platform has officially launched the 1.5B, 7B, 8B, and 14B versions of the DeepSeek-R1 model, and will soon update the 32B, 70B, and other versions.

ChainCatcher news, Google (GOOG.O) CEO Sundar Pichai stated that its newly released quantum computing chip Willow is an important step in building useful quantum computers, with practical applications in drug discovery, fusion energy, battery design, and more. Its breakthrough can exponentially reduce errors as we use more qubits, solving a challenge that has persisted in the field for 30 years.In benchmark tests, Willow completed a "standard benchmark computation" in less than five minutes, while even the fastest supercomputers today would take "10 to the power of 25" years to complete this computation ------ a number far exceeding the age of the universe.