O Que É SHIH

According to Beating AI news, SpaceXAI will deploy NVIDIA's Vera CPU to run the next generation of AI Agents. This CPU is not responsible for replacing GPUs in large model calculations; instead, it primarily handles tasks between model calls, including tool management, code execution, data processing, task orchestration, and simulation. SpaceXAI will also continue to expand Grok's AI infrastructure using Vera Rubin. As AI Agents become more complex, the workload on CPUs increases. After model generation tool calls, tasks such as executing code and querying data are typically assigned to the CPU. If the CPU runs too slowly, the GPU has to wait for results. Vera is NVIDIA's CPU specifically designed for this purpose, featuring 88 NVIDIA-developed Olympus CPU cores and a memory bandwidth of up to 1.2TB/s. NVIDIA claims that for tasks such as Agents, reinforcement learning, and data processing, Vera can complete tasks up to 1.8 times faster than x86 CPUs. NVIDIA's hardware allocation for Agents is becoming increasingly clear: the Rubin GPU is responsible for large model computations, the Groq LPU accelerates token generation, and the Vera CPU manages tool coordination and code execution. SpaceXAI also plans to take Vera Rubin into space, with the first generation of Starmind AI satellites based on the optimized Vera Rubin NVL72.

According to Beating AI news, SpaceXAI will deploy NVIDIA's Vera CPU to run the next generation of AI Agents. This CPU is not intended to replace GPUs for large model calculations; instead, it will handle tasks between model calls, including tool coordination, code execution, data processing, task orchestration, and simulation. SpaceXAI will also continue to expand Grok's AI infrastructure using Vera Rubin. As AI Agents become more complex, the workload on CPUs increases. After model generation tools are invoked, tasks such as executing code and querying data are typically assigned to the CPU. If the CPU runs too slowly, the GPU has to wait for results. Vera is NVIDIA's CPU specifically designed for this purpose, featuring 88 NVIDIA-developed Olympus CPU cores and a memory bandwidth of up to 1.2TB/s. NVIDIA claims that for tasks involving Agents, reinforcement learning, and data processing, Vera can complete tasks up to 1.8 times faster than x86 CPUs. The hardware roles prepared by NVIDIA for Agents are becoming increasingly clear: Rubin GPUs are responsible for large model computations, Groq LPUs accelerate token generation, and Vera CPUs handle tool coordination and code execution. SpaceXAI is also preparing to take Vera Rubin into space, with the first generation of Starmind AI satellites based on the optimized Vera Rubin NVL72.

On August 25, according to Farside data, the total net inflow for Bitcoin spot ETFs in the US yesterday was $337.6 million. Among them, BlackRock IBIT had a net inflow of $208.9 million, Fidelity FBTC had a net inflow of $104.6 million, Bitwise BITB had a net inflow of $3 million, Grayscale BTC had a net inflow of $16.4 million, HODL had a net inflow of $3.3 million, and MSBT had a net inflow of $1.4 million. The total net inflow for Ethereum spot ETFs was $115.6 million. Among them, BlackRock ETHA had a net inflow of $90.9 million, Fidelity FETH had a net inflow of $6.8 million, ETHW had a net inflow of $900,000, ETHV had a net inflow of $4.5 million, and ETH had a net inflow of $12.5 million.