Sivers Semiconductors and SemiNex launch a $3.4 million InP light-source program for AI data center interconnects
Sivers Semiconductors and SemiNex announced a $3.4M InP light-source program on 13 August 2026, targeting co-packaged optics and wavelength-multiplexed links for AI data centers.

Sivers Semiconductors AB and SemiNex Corporation announced a $3.4 million program on 13 August 2026 to develop next-generation indium phosphide (InP) light sources for AI data center interconnects, with customer sampling and initial production ramps targeted for the second half of 2027[1].
The two companies bring complementary capabilities. Sivers, headquartered in Kista, Sweden, supplies photonics and wireless technologies across AI data centers, satellite communications, defense, and telecom. SemiNex, based in Danvers, Massachusetts, designs and manufactures high-power InP laser diodes, DFB lasers, semiconductor optical amplifiers (SOAs), and external cavity lasers[1].
What the program covers
The collaboration is structured around three device categories, each addressing a specific bottleneck in next-generation optical connectivity[1]:
- High-power external laser sources for co-packaged optics (CPO)
- High-channel-count DFB laser arrays for wavelength-multiplexed links
- SOA gain stages designed to extend optical reach as channel counts increase
The common thread is InP material engineering. As SemiNex CEO Ronald Moore put it, the gains the industry needs "come out of the design and processing of the InP itself"[1] - not from pushing existing devices harder.
Why CPO changes the light-source equation
Co-packaged optics moves the optical engine off the system faceplate and into the ASIC package, cutting electrical path length and reducing power per bit. As optics move from the faceplate into the package, power per wavelength, wall-plug efficiency, and thermal stability increasingly determine how far an architecture can scale, making the light source a system-level design problem rather than a component selection.
That shift puts pressure on the laser specifically. Photonic integrated circuits are highly temperature-sensitive and introduce new forms of thermal coupling when integrated alongside high-power logic dies, and dynamic AI workloads can induce fast, localized temperature excursions that directly impact optical performance. Higher wall-plug efficiency and tighter thermal stability in the InP source directly reduce the burden on the rest of the package.
The market backdrop is substantial. According to LightCounting, sales of lasers and photonic integrated circuits for optical transceivers are expected to grow from $2.4 billion in 2023 to $5.9 billion by 2029, fueled largely by AI data centers[1].
What each side contributes
Sivers Semiconductors CEO Vickram Vathulya framed the light source as a peer to compute and memory: "The advanced light sources needed for next generation optical networking within AI factories are as crucial as the compute, memory and silicon photonics elements"[1]. SemiNex's Moore was direct about the constraint: "The optical layer is becoming the constraint on how far AI infrastructure can scale. The industry needs more power, at more wavelengths, at higher efficiency"[1].
The program pairs SemiNex's InP device expertise with Sivers' photonics integration and production capabilities, with the stated goal of delivering light-source solutions at scale rather than as engineering samples.
The timeline to watch is the second half of 2027, when customer sampling and early production ramps are planned. Whether the program can demonstrate the wall-plug efficiency and thermal stability that CPO architectures demand at volume will determine how quickly this collaboration translates into designed-in wins at hyperscale operators[1].
Written by Electronics Insider's automated desk from the sources above and published automatically. How we work.
Related
Design & EDAETH Zürich and NYU show JEDEC test loops miss worst-case HBM2 power by 23.6%
Researchers at ETH Zürich and NYU tested 36 real HBM2 chips and found JEDEC measurement loops understate peak power by 23.6%, with read ops and data-bus toggling as the main drivers.
15 Oct 2026
Design & EDAFlowAccel's Joel Jorgensen argues chip teams must design their workflows before AI can help them hit tape-out
Joel Jorgensen of FlowAccel argues that AI tools alone won't save a tape-out schedule - teams must first design the cross-team workflows that let those tools compound into real gains.
15 Oct 2026
SemiconductorsETH Zürich and NYU find JEDEC test loops underestimate HBM2 power by up to 35.4%
Researchers at ETH Zürich and NYU tested 36 real HBM2 chips and found JEDEC-specified current loops can underestimate peak power by up to 35.4%, with implications for AI accelerator thermal design.
15 Oct 2026