马斯克：3年内，太空会是部署AI最便宜的地方

Core Viewpoint - Elon Musk predicts that within 36 months, space will become the cheapest place to deploy artificial intelligence, primarily due to the limitations of power supply on Earth compared to the exponential growth of chip production [1][7]. Group 1: Space Data Centers - Musk emphasizes that the core reason for moving data centers to space is the inability of power supply to keep up with the exponential growth of chip production, leading to a potential crisis where large clusters cannot be powered by the end of this year [1][7]. - He suggests that solar panels in space can generate power at approximately five times the efficiency of those on Earth, eliminating the need for costly battery systems to store energy for nighttime use [3][9]. - The deployment of solar panels in space is seen as more efficient and less bureaucratic than expanding solar farms on Earth, which face complex approval processes [1][7]. Group 2: Economic Viability - Musk asserts that deploying AI in space will be the most cost-effective option, with the transition expected to occur within 30 to 36 months [3][9]. - He highlights that the maintenance of GPUs sent to space can be managed by resolving initial faults on Earth before deployment, ensuring reliability after initial testing [3][9]. - The cost of solar panels sent to space is projected to be 5-10 times cheaper than their Earth counterparts due to the absence of weather-related constraints and the need for heavy support structures [10]. Group 3: Power Supply Challenges - Musk expresses concerns about the current bottlenecks in power generation, particularly regarding the availability of gas turbines and the high tariffs on imported solar panels in the U.S. [4][10]. - He notes that operating data centers consumes significant amounts of electricity, with cooling alone increasing power consumption by 40% in his Memphis data center [11]. - The need for approximately 1 gigawatt of power capacity is highlighted to support 330,000 GB300 servers, indicating the scale of energy requirements for data centers [11]. Group 4: Future Vision - Musk envisions that before AI can be deployed in space, energy supply is the primary limitation, while after deployment, the focus will shift to chip availability [12]. - He indicates that TeraFab may need to produce not only logic chips but also storage and packaging components to support its vision [12].