LightCounting回顾Optica高管论坛称AI将长期拉动光互连需求，但正文为4月旧研究摘要

LightCounting releases a research note on highlights from Optica’s Executive Forum
by Daryl Inniss, Vladimir Kozlov and Roy Rubenstein
April 7th, 2026
LightCounting attended the day-long Optica’s Executive Forum. The event is held alongside the OFC conference and exhibition and took place in March in Los Angeles, California. The forum offered a wide-ranging discussion of the optical industry, bringing together perspectives from end-users and suppliers. The event provided a timely glimpse into the hyperscalers’ priorities and requirements, while leading photonics firms’ CEOs and CTOs reflected on the state of the industry and technology roadmaps.
For the keynote, Richard Ho of OpenAI highlighted the extraordinary pace of AI’s progress since the launch of ChatGPT in November 2022, emphasizing optics’ central role in sustaining the relentless scaling needs of AI infrastructure. Indeed, all the sessions pointed to optics being central to AI scaling, or as Coherent's CTO, Julie Eng, put it: optics is now inside the machine.
Executives from Coherent, Lumentum, and Nokia discussed their respective investments from Nvidia. These investments not only will secure critical future components but will help fund parallel developments across multiple technology paths to address the scaling challenges already evident. Challenges extend beyond core technologies. A critical requirement is the ability to scale back-end manufacturing and supply chain processes to meet the surge in demand expected as the industry moves from copper to optical interconnects towards the decade’s end.
The CTO fireside chat session provided insight into the CTOs’ concerns and their thoughts on technologies. The panelists included Coherent's Julie Eng, Matthew Sysak of Lumentum, Gregg Bartlett of GlobalFoundries, and Hong Liu of Google.
Bartlett outlined GlobalFoundries recent acquisitions of the Singapore foundry AMF, and stressed that given the current speed to market demands, the foundry must have multiple capabilities. For now, it is unknown which technologies will win. The theme of being engaged widely was also mentioned by Coherent’s Eng. CPO and NPO optics will have many, diverse solutions, just like we have many types of pluggable transceivers now. It will co-exist with pluggable modules: CPO/NPO will be used for density, pluggables for their flexibility. Multiple technologies must be pursued because each hyperscaler has unique architectural requirements, and the engineering decisions they make are based on different constraints and business models.
An important question addressed by the CEOs and supplier leaders panel is the health of the industry, that is, the supply chain and whether the industry is experiencing a bubble. The panel included three CEOs—Michael Hurlston of Lumentum, Jim Anderson of Coherent, and Matt Murphy of Marvell. The two other panelists were Bill Gartner, Cisco’s Senior VP and general manager, and David Heard, Nokia President, Network Infrastructure and former CEO of Infinera.
One question asked was will Nvidia’s funding to secure supply create allocation pressure for other firms? Cisco’s Gartner said this is a concern and that Cisco has already been actively working to diversify its suppliers. But the reality is that larger players will get priority and smaller firms will likely face restrictions.
The debate shifted to whether such demand will continue and is AI a bubble. The panel gave solid reasons why they believe AI is not a bubble.
Unlike the telecom bubble around 2000 where overcapacity happened quickly, AI has some inherent brakes: demand for chips, memory, optics, and power are all issues constraining growth. Both AI training and inferencing continue to drive demand. There is also an ongoing transition from copper to optics, an industry driver independent of any perceived AI hype. Coherent’s Anderson also highlighted the optics for scale-up opportunity which could result in demand an order of magnitude larger than scale-out.
The CEOs referred to multi-year contracts being placed with their firms and even lasting to the decade’s end. Cisco’s Gartner also pointed to a second wave of AI demand, once enterprises embrace inferencing. That could massively expand demand, he says.
That said, the panelists expect market ups and downs going forward. But the sense is that the AI cycle is durable because demand is real while supply has constraints.
Full text of the research note is available at: https://www.lightcounting.com/login