- The definition between the telecom and data center infrastructure market is blurring
- AI is pushing optics deeper into the telecom network
- Startups are rethinking everything from silicon photonics and comb lasers to optical I/O and switching
AI is driving optics beyond the long-haul network and deeper into data centers, compute systems and network infrastructure. A new class of startups is betting that silicon photonics, comb lasers, optical I/O and circuit switching can move more data without allowing power consumption and heat to spiral. However, that work could matter beyond AI data centers.
Startups and incumbent vendors are developing photonic technologies that could change the components available to operators and optical transport vendors as the boundary between telecom and data-center infrastructure continues to blur. “AI and superintelligence (SI) workloads are placing pressure on optical connectivity across the full spectrum: bandwidth, latency, power efficiency, thermal management, density, reach, economics, serviceability, and manufacturability,” Frank Yang, research director at Dell’Oro Group. The result is a massive uptick optical startups, innovation, advancements, funding and valuations.
While funding does not prove that every technical optical networking approach will reach commercial scale, it shows that investors expect optical connectivity to become a challenge as AI systems require more bandwidth within racks, between clusters and across data-center networks.
One company, Lumilens, said in August that it was valued at $5.51 billion after raising more than $700 million in a Series C round co-led by Atreides Management, Bain Capital Ventures, Meritech, Seligman Ventures and Spark Capital. Its total funding exceeds $900 million. On Wednesday, Salience Labs announced it had raised $125 million in a Series B led by Bullhound Capital. Salience is also backed by AMD Ventures and Intel Capital, Hudson River Trading and British Business Bank.
5 companies to watch
The five companies below are not necessarily the largest or best funded. They are worth watching because each is tackling a different bottleneck in the emerging optical infrastructure stack, from chip-to-chip links and multi-wavelength light sources to integrated lasers and programmable switching.
Ayar Labs
Ayar Labs is developing optical I/O chiplets and multi-wavelength light sources designed to move data optically between chips. Its technology sits at the center of the shift from electrical interconnects to optics as bandwidth demands, power consumption and heat become harder to manage.
OpenLight
OpenLight is integrating lasers into an open silicon-photonics platform aimed at datacom, telecom, high-performance computing, AI and optical computing. Its relevance lies in whether integrated lasers can simplify optical designs and provide a practical manufacturing path across several fast-growing markets.
iPronics
iPronics is developing programmable silicon-photonics optical circuit switching, including rack-mounted connectivity for AI data-center networks. Unlike a startup focused primarily on faster transceivers, iPronics is addressing network architecture: how operators, hyperscalers and vendors connect AI clusters and distributed data centers as traffic volumes grow.
Comb lasers could simplify the component stack
What is a comb laser?
A comb laser generates multiple evenly spaced wavelengths from a single source, allowing one device to support many optical channels. That could simplify the component stack as networks move toward higher bandwidth and greater wavelength counts, although manufacturing scale, reliability and cost will determine how broadly the technology is deployed.
Quintessent
Quintessent is working on scalable optical interconnect hardware and quantum-dot DWDM comb lasers, essentially using one laser source to provide multiple wavelengths. It raised a $40 million Series A announced in August 2026.
Enlightra
Swiss startup Enlightra is developing comb-laser technology and photonic interconnects. Its approach uses a specialized laser to produce more than 50 optical wavelengths for connections between GPU and TPU clusters, offering a test of whether frequency-comb technology can become a practical building block for next-generation dense wavelength-division multiplexing and optical interconnects.
The number of optical component startups has 'exploded'
Cignal AI's July 2026 tracker said the number of venture-backed optical component startups had “exploded,” with companies working across lasers, optical circuit switches, digital signal processors, silicon photonics and integrated optical engines. The breadth of that field makes technical differentiation important, but commercialization, manufacturing and customer adoption will ultimately determine which approaches will succeed.
Read more about optical networking on Fierce Network
Cisco and Nvidia develop new pluggable optic module for AI data centers
Nvidia’s SVP of networking explains the company's networking stack
Hitachi Vantara sees AI infrastructure cycle echoing the fiber boom