This article covers a strategic collaboration between Sivers Semiconductors AB and Jabil. Together, they’re working on a 1.6T linear receive optical (LRO) transceiver module.
Sivers brings its high-performance distributed feedback (DFB) lasers to the table. Jabil adds engineering, supply chain, and manufacturing muscle.
The goal? Deliver a pluggable, energy-efficient optical interconnect for hyperscale AI data centers. The piece also touches on the broader market and why this tech matters for AI and future data-center design.
Overview of the Sivers–Jabil collaboration
Next-generation hyperscale environments need higher throughput and lower power per bit. The 1.6T LRO module aims to boost energy efficiency on the receive path but still play nicely with current optical interconnect setups.
Sivers’ DFB laser tech pairs with Jabil’s end-to-end manufacturing know-how. The hope is to get a high-performance, low-power, pluggable solution out there faster than the usual drawn-out development cycles.
Technical highlights of the 1.6T LRO module
- 1.6T linear receive optical (LRO) transceiver module for high-throughput, energy-efficient optical interconnects
- Uses Sivers’ DFB lasers for stable, high-bandwidth light sources
- Built as a pluggable module to slot into a range of data-center interconnect designs
- Joint effort taps into Jabil’s engineering, supply chain, and manufacturing strengths for scaling up
- Geared toward hyperscale AI data centers with better power efficiency per bit
- Puts the spotlight on receive-side linear optics to cut energy use at crazy data rates
Industry context and market outlook
The timing here isn’t random. The market is hungry for high-speed, energy-efficient optical interconnects—especially for AI workloads.
Research shows a big shift toward 800G and faster transceivers in the pluggables space. That’s all thanks to the relentless demand for AI training and inference at scale.
Market projections and implications
- LightCounting thinks 800G and higher-speed transceivers will make up about 80% of the pluggables market by 2030
- The same research expects the global pluggables market to hit 225 million units shipped in 2030
- All signs point to a premium on energy-efficient, high-throughput interconnects for hyperscale
Why this matters for hyperscale AI data centers
In hyperscale environments, energy efficiency isn’t just nice to have—it shapes total cost of ownership and thermal management. That, in turn, impacts cooling needs and data-center reliability.
The 1.6T LRO approach tries to cut per-bit power on the receive side. When you roll that out across thousands of server and switch nodes, the benefits start stacking up.
By offering a high-throughput, low-power optical interconnect in a pluggable package, this collaboration aims to support AI training and inference at scale—without sending energy bills through the roof.
Opportunities and alignment with industry needs
- Potential for smoother integration into existing pluggables ecosystems, making deployment less of a headache for big facilities
- Supply-chain resilience gets a boost from Jabil’s strengths in sourcing and large-scale manufacturing
- Lines up neatly with AI infrastructure demands: high throughput and tough energy efficiency targets
Next steps and where to learn more
As development moves closer to a market-ready, pluggable 1.6T LRO module, industry folks will be watching closely. People are curious to see if Sivers’ laser tech and Jabil’s manufacturing skills can actually deliver real savings for AI data centers.
More information is up on Sivers Semiconductors’ website. They’ll probably share updates about the partnership and the technology roadmap there for anyone in the optics and AI space who wants to keep tabs.
Here is the source article for this story: Sivers and Jabil collaborate on 1.6T optical module