While the transistor-based silicon chip has increased computing power exponentially in recent decades as transistors have grown to the width of several atoms, shrinking them further is a challenge. Not only is it difficult to make something so tiny, but as they get smaller signals can leak out between them. Thus, Moore’s law, according to which the density of transistors on a chip should double every two years and bring down costs, is slowing down, which pushes the industry to seek new solutions to meet the growing needs. in artificial intelligence computing.
According to data firm PitchBook, silicon photonics startups raised more than $750 million last year, doubling from 2020. In 2016, that was around $18 million.
“Artificial intelligence is growing like crazy and taking over large parts of data centers,” Ayar Labs CEO Charles Wuischpard said in an interview with Reuters. “The challenge of data movement and the power consumption in that data movement is a big, big problem.”
The challenge stems from the fact that many large machine learning algorithms may use hundreds or thousands of chips for computation, and there is a bottleneck on the speed of data transmission between chips or servers using current electrical methods. Light has been used for decades to transmit data through fiber optic cables, including undersea cables, but bringing it to the chip level has been difficult because the devices used to create or control the light are not as easy to shrink as transistors.
Brendan Burke, senior emerging technologies analyst at PitchBook, expects silicon photonics to become mainstream hardware in data centers by 2025 and estimates the market to reach $3 billion by then. similar in size to the AI graphics chip market in 2020.
Beyond connecting transistor chips, start-ups using silicon photonics to build quantum computers, supercomputers and chips for self-driving vehicles are also raising significant funds.
PsiQuantum has raised around $665 million so far, although the promise of quantum computers changing the world is still a long way off. Lightmatter, which builds processors using light to accelerate AI workloads in the data center, has raised a total of $113 million and will launch its chips later this year and test them with customers soon after. Luminous Computing, a startup building an AI supercomputer using silicon photonics, backed by Bill Gates, has raised a total of $115 million.
Photonic foundries
Start-ups aren’t the only ones pushing this technology forward. Semiconductor makers are also gearing up to use their silicon chip fabrication technology for photonics.
Amir Faintuch, head of computing and wired infrastructure at GlobalFoundries, said the collaboration with PsiQuantum, Ayar, and Lightmatter has created a silicon photonics fabrication platform that others can use. This platform was launched in March.
Peter Barrett, founder of venture capital firm Playground Global, which has invested in Ayar Labs and PsiQuantum, believes in the long-term prospects of silicon photonics for accelerating computing, but believes the road will be long.
“What the guys at Ayar Labs are doing so well…is they’ve solved the problem of interconnecting data for traditional high-performance computing,” he said. “But it will take time before we have pure digital photonic computing for non-quantum systems.”