Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
New era Silicon Valley works on networking, not the kind you find on LinkedIn.
Like the technology industry Turns billions into AI data centerschipmakers both large and small are ramping up innovation around the technology that connects chips to other chips, and server racks to other server racks.
Networking technology has been around since the dawn of the computer, crucially connecting mainframes so they could share data. In the world of semiconductors, networking plays a role at almost every level of the stack — from the interconnections between transistors on the same chip, to the external connections made between boxes or racks of chips.
Chip giants like Nvidia, Broadcom, and Marvell already have goodwill in the networking space. But with the boom in artificial intelligence, some companies are looking for new networking approaches that help them speed up the flow of massive amounts of digital information through data centers. This is where deep tech startups like Lightmatter, Celestial AI, and PsiQuantum come in, which use optical technology to accelerate high-speed computing.
Optical technology, or photonics, is having a coming-of-age moment. The technology was considered “vulnerable, expensive and fairly useful” for 25 years, until the AI boom sparked renewed interest, according to PsiQuantum founder and chief scientific officer Pete Shadbolt. (Shadbolt appeared on a panel last week co-hosted by WIRED.)
Some venture capitalists and institutional investors, hoping to catch the next wave of chip innovation or at least find a suitable acquisition target, are funneling billions to startups like this one that have found new ways to speed up data throughput. They believe that traditional interconnection technology, which relies on electrons, simply cannot keep up with the growing need for high-bandwidth AI workloads.
“If you look back historically, network coverage was really boring, because they were switching packets of bits,” says Ben Bajarin, a veteran technology analyst who is CEO of research firm Creative Strategies. “Now, because of AI, it has to move fairly robust workloads, which is why you see innovation centered around speed.”
Large chip power
Bajarin and others give Nvidia credit for being prescient about the importance of networking when it made two major technology acquisitions years ago. In 2020, Nvidia spent nearly $7 billion to acquire Israeli company Mellanox Technologies, which makes high-speed networking solutions for servers and data centers. Shortly after, Nvidia bought Cumulus Networks to operate it Linux-based software system for computer networks. This was a turning point for Nvidia, which was rightly betting that its GPU and parallel computing capabilities would become more powerful when bundled with other GPUs and placed in data centers.
While Nvidia dominates vertically integrated GPU lineups, Broadcom has become a major player in custom chip accelerators and high-speed networking technology. The $1.7 trillion company works closely with Google, Meta, and most recently OpenAI, on data center chips. It is also at the forefront of silicon photonics. Last month, Reuters reported this Broadcom is preparing a new networking chip Called Thor Ultra, it is designed to provide “a critical link between the AI system and the rest of the data center.”
On its earnings call last week, semiconductor design giant ARM announced plans to acquire networking company DreamBig for $265 million. DreamBig makes AI chips — small modular circuits designed to be pieced together into larger chip systems — in partnership with Samsung. The startup has “interesting intellectual property… (which) is very fundamental to scaling and scaling networks,” ARM CEO Rene Haas said on the earnings call. (This means connecting components and sending data up and down one chipset, as well as connecting racks of chips to other racks.)
light on
Lightmatter CEO, Nick Harris He pointed out The amount of computing power required by AI is now doubling every three months, which is much faster than… Moore’s law He dictates. Computer chips are getting bigger and bigger. “When you’re at the cutting edge of the largest chips you can make, all the performance after that comes from connecting the chips together,” Harris says.
His company’s approach is cutting-edge and does not rely on traditional networking technology. Lightmatter builds the silicon photonics that connects the chips together. She claims to make The world’s fastest optical engine for AI chipswhich is essentially a 3D stack of silicon connected by light-based interconnection technology. The startup has raised more than $500 million over the past two years from investors like GV and T. Rowe Price. Last year, its value reached $4.4 billion.