Lessengers has unveiled a 3.2 Tbit/s near-packaged-optics platform that uses a glass packaging core, through-glass vias and polymer optical waveguides to place high-bandwidth optical engines close to a network-switch ASIC.
The architecture is being demonstrated at ECOC 2026 in Málaga. It combines the company’s Direct Optical Wiring technology with a glass substrate intended to carry both electrical connections and optical paths inside a compact, socketable assembly.
Near-packaged optics shortens the electrical connection between the switch silicon and optical engine compared with a conventional front-panel transceiver. This becomes increasingly important as electrical lane rates rise and copper traces consume more power while providing less usable reach.
Lessengers routes light through planar structures in the glass core and through three-dimensional polymer waveguides formed by its DOW process. The company says the process does not require the multi-channel lenses and active optical alignment commonly used to connect photonic devices to fibre arrays.
Why use glass?
Glass substrates can provide fine interconnect routing, electrical insulation and dimensional stability. Their coefficient of thermal expansion can also be more closely matched to silicon than that of some organic packaging materials, potentially reducing movement and warpage as the assembly heats and cools.
The new platform incorporates vertical through-glass vias for electrical connections. Lessengers says the optical junctions can be hermetically sealed, making the module suitable for immersion-cooled systems without an additional protective optical package.
The company presents the architecture as a path to 51.2T and 102.4T switch systems. A 3.2T engine can aggregate multiple high-speed lanes while remaining replaceable, addressing one of the serviceability concerns associated with permanently attached on-board optics.
Lessengers has previously demonstrated socketable 800G NPO engines.
The new development increases the aggregate bandwidth and changes the packaging base to a glass core, but the company has not disclosed the physical size, lane configuration or electrical interface of the 3.2T implementation.
Measured data still required
Lessengers claims power savings of up to 50%, but it has not identified the comparison architecture or published the test conditions. The announcement also omits total module power, energy per bit, optical insertion loss, bit-error rate, reach and thermal-resistance figures.
Manufacturing questions remain equally important. No data has been released for through-glass-via yield, polymer-waveguide uniformity, socket durability, thermal cycling, moisture resistance or production capacity. The company has also not given an engineering-sample or volume-availability date.
The platform nevertheless presents a technically differentiated approach to scaling optical connectivity. Its commercial value will depend on whether the glass and polymer assembly can retain optical alignment, low loss and acceptable yield across repeated thermal cycles and high-volume production.
