The artificial intelligence boom is creating a huge demand for computing power, electricity and cooling. Google is now investigating whether some of those challenges could eventually be addressed by moving AI infrastructure into space.
Through Project Suncatcher, Google is studying the possibility of building interconnected satellites equipped with its Tensor Processing Units.
The basic idea is straightforward: put AI computing hardware where sunlight is abundant and use solar power to operate the machines.
Google says satellites in low Earth orbit can receive near-constant sunlight and potentially generate up to eight times more solar power than on Earth.
That could become increasingly important as AI systems require larger amounts of computing infrastructure.
However, space introduces completely different engineering problems. A conventional data center relies heavily on air circulation, cooling systems and large terrestrial power networks. In space, there is no atmosphere to carry heat away.
Google is therefore developing thermal-management systems to transfer heat away from the TPUs and release it into space.
The first orbital prototype will provide an important real-world test of these technologies. Google’s long-term concept involves multiple satellites communicating with each other through high-bandwidth laser links.
The concept remains experimental, and significant engineering work is still required. For now, Suncatcher represents a question about the future of computing rather than a replacement for today’s data centers.
Source: Google Research
आपके लिए सुझाव
