GlobalFoundries, a company that makes chips for others including AMD and General Motors, previously announced a partnership with Lightmatter. Harris says his company “works with the largest semiconductor companies in the world, as well as the hyperscalers,” referring to the largest cloud companies such as Microsoft, Amazon and Google.
If Lightmatter or another company can reinvent the wiring of massive AI projects, a major bottleneck in the development of smarter algorithms could disappear. The use of more computing power was fundamental to the advances that led to ChatGPT, and many AI researchers see the further scaling up of hardware as crucial to future progress in the field – and to the hope of one day achieving the vaguely specified goal of artificial intelligence reach. general intelligence, or AGI, that is, programs that can match or exceed biological intelligence in every possible way.
Linking a million chips together with light could enable algorithms that go several generations beyond current developments, says Nick Harris, CEO of Lightmatter. “Passage will enable AGI algorithms,” he confidently suggests.
The large data centers needed to train massive AI algorithms typically consist of racks filled with tens of thousands of computers running specialized silicon chips and a spaghetti of mostly electrical connections between them. Maintaining training runs for AI on so many systems – all connected by wires and switches – is a huge technical undertaking. Converting between electronic and optical signals also places fundamental limits on the ability of chips to perform calculations as a whole.
Lightmatter’s approach is designed to simplify the difficult traffic within AI data centers. “Typically you have a bunch of GPUs, and then a layer of switches, and a layer of switches, and a layer of switches, and you have to traverse that tree” to communicate between two GPUs, Harris says. In a data center connected by Passage, Harris says, every GPU would have a high-speed connection to every other chip.
Lightmatter’s work on Passage is an example of how the recent AI boom has inspired companies large and small to try to reinvent the key hardware behind developments like OpenAI’s ChatGPT. Nvidia, the leading provider of GPUs for AI projects, held its annual conference last month, where CEO Jensen Huang unveiled the company’s newest chip for training AI: a GPU called Blackwell. Nvidia will sell the GPU in a “superchip” consisting of two Blackwell GPUs and a conventional CPU processor, all connected using the company’s new high-speed communications technology, called NVLink-C2C.
The chip industry is known for finding ways to get more computing power out of chips without making them bigger, but Nvidia chose to buck that trend. The Blackwell GPUs in the company’s superchip are twice as powerful as their predecessors, but are made by bonding two chips together, meaning they consume much more power. This trade-off, alongside Nvidia’s efforts to glue its chips together with high-speed connections, suggests that upgrades to other key components for AI supercomputers, as suggested by Lightmatter, could become more important.