The next bottleneck for AI data centers is connectivity… Salience Labs’ bet on optical switches

Date
Empty
Classification
  1. Investment/Funding
작성자
에디터 L
UK-based Salience Labs has raised a Series B round to mass-produce optical switches for AI data centers and expand its U.S. business. Technology that connects computing equipment is becoming increasingly important in the race to build AI infrastructure.
Salience Labs announced on October 14 that it had raised $125 million in Series B funding. Bullhound Capital led the round, with participation from AMD Ventures, Intel Capital, and others. The company plans to use the funds to scale the supply and mass production of its optical circuit switching technology and expand its U.S. business. The key question is whether its vision for improving connectivity efficiency in AI data centers will translate into increased production and real-world supply.

Connectivity becomes more important as computing equipment proliferates

In AI data centers, multiple computing devices work by exchanging data. Alongside the processing capabilities of the equipment itself, the performance and cost of the networks that move data are also important. Salience Labs is developing optical circuit switching, or OCS, technology to connect GPUs deployed across multiple racks.
OCS is a technology that changes the paths taken by signals transmitted as light. The company aims to integrate it into silicon photonics-based chips to help reduce connectivity bottlenecks and power demands in AI data centers. However, the potential efficiency gains described in this funding announcement should not be taken as proven results across all customer environments.

A 64-port demonstration and plans for next-generation products

The company said it demonstrated a 64-port optical switch at ECOC 2026. It also plans to develop 128-port and 256-port products with the new funding. This is why it is important to distinguish between technology demonstrated publicly and products yet to come.
A demonstration shows that a technology works. Applying it in a real customer's data center still involves challenges such as connecting with existing equipment, ensuring reliable operation, and scaling supply. Going forward, the key will be not the number of ports itself, but what problems the technology solves in customer environments and how broadly it is adopted.

After raising funding, the next steps are mass production and customer acquisition

Salience Labs said it would build up its mass-production capabilities while expanding its presence in Silicon Valley. The funding is intended to support not only technology development but also production and market expansion.
The participation of AMD Ventures and Intel Capital is also notable. However, investment by corporate venture arms does not necessarily mean there is a supply agreement or product integration with their parent companies. Fundraising and customer acquisition are separate business milestones that must be assessed independently. What matters is whether the large funding round ultimately leads to actual deliveries and revenue.

Implications for the startup ecosystem

This case shows that when startups assess the AI market, they need to look beyond models and applications to the infrastructure challenges beneath them. Business opportunities can be found in areas such as data movement, power, and cooling, where the burden grows as facilities scale. Technology companies need to go beyond describing their technology's performance and explain which customer costs it can reduce, how it can be applied to existing infrastructure, and what a pilot project would validate. This is Start Lounge's editorial interpretation of this funding news.
Sources
Salience Labs official announcement · October 14, 2026
https://www.saliencelabs.ai/salience-labs-125-million-series-b-funding/

#Startups #Investment #AI #DataCenters
[Advertisement]
Collaboration proposals / business inquiries: hello@startlounge.kr
IR/pitch deck inquiries: hello@pitchmound.kr
For startups, everything. Start Lounge - Launchpad 「Lunch Pad」
ⓒ 2026. Start lounge. All rights reserved.