Podcast Episode
The kit is targeted for broader availability in the second quarter of 2026, with early access already open for select customers.
All optical systems require a laser chip to generate the light beams that carry data, and those chips are made with indium phosphide, a material primarily used in long-distance telecommunications. Supply is not currently sufficient to meet surging demand from AI data centres, which prompted Nvidia to invest two billion dollars each in Lumentum and Coherent, two of the largest laser manufacturers, earlier this month.
Chief executive Matt Crowley said Scintil is in discussions with six or seven companies interested in deploying the technology by 2028. The company plans to showcase live demonstrations, including a sixteen-wavelength configuration, at the Optical Fiber Communication Conference in Los Angeles next week.
Nvidia-Backed Scintil Photonics Starts Testing Laser Chips That Could Replace Copper in AI Servers
March 12, 2026
0:00
4:18
French startup Scintil Photonics, backed by Nvidia, has begun sending its LEAF Light laser chips to customers for testing. The technology integrates multiple lasers onto a single chip and could help replace copper connections inside AI data centres with light-based data transmission, promising faster speeds and lower power consumption.
Laser Chips Move From Lab to Customer Hands
Scintil Photonics, a Grenoble-based startup backed by Nvidia, has announced that it is now providing its laser chips to customers for real-world testing. The company launched its LEAF Light Evaluation Kit, which it describes as the first single-chip dense wavelength division multiplexing laser source to move from laboratory validation into a customer-facing evaluation programme.The kit is targeted for broader availability in the second quarter of 2026, with early access already open for select customers.
Why Light Is Replacing Copper
As AI systems scale from single-rack setups to massive clusters connecting thousands of processors, traditional copper interconnects are hitting their limits in speed, density, and reach. Co-packaged optics, which places optical components directly alongside processor chips, is emerging as the preferred architecture for next-generation AI networks. However, adoption depends on the availability of specialised laser components.All optical systems require a laser chip to generate the light beams that carry data, and those chips are made with indium phosphide, a material primarily used in long-distance telecommunications. Supply is not currently sufficient to meet surging demand from AI data centres, which prompted Nvidia to invest two billion dollars each in Lumentum and Coherent, two of the largest laser manufacturers, earlier this month.
A Different Approach to Manufacturing
Scintil, which raised fifty-eight million dollars in a Series B round with Nvidia participation last year, has developed a way to integrate indium phosphide lasers with other optical components onto a single chip using its proprietary SHIP technology platform. The company partners with Tower Semiconductor for manufacturing and claims its LEAF Light technology delivers a fifty percent power reduction compared with single-wavelength co-packaged optics.Chief executive Matt Crowley said Scintil is in discussions with six or seven companies interested in deploying the technology by 2028. The company plans to showcase live demonstrations, including a sixteen-wavelength configuration, at the Optical Fiber Communication Conference in Los Angeles next week.
Published March 12, 2026 at 12:34pm