1.6T OSFP transceiver targets hyperscale AI data centers with integrated laser

Gemtek introduced the AiPhoton 1.6T OSFP transceiver for hyperscale AI data centers, built in collaboration with NewPhotonics. The module is based on NewPhotonics’ NPG10204 DR8 photonic integrated circuit (PIC) with a transmitter-on-chip design.

Gemtek describes the AiPhoton as a 1.6 Tbps DR8 OSFP module aimed at scale-out interconnect deployments. The design uses an octal-channel architecture with 224 Gbps per lane and targets 200G-per-channel single-mode PAM4 (IMDD). NewPhotonics’ PIC integrates the laser and modulator in a multi-channel monolithic device, with Gemtek pointing to low power and improved signal integrity as key attributes for these link speeds.

A notable engineering detail is the packaging and assembly approach. The companies say the PIC uses a highly integrated flip-chip method intended to eliminate complex optical assembly and wire-bonding steps. If that holds up in production, it could matter as much as the line rate: manufacturability and yield maturity are often what separate a “1.6T” spec sheet from optics you can actually deploy at volume without unpleasant reliability surprises.

“With the introduction of AiPhoton we are well-positioned to meet the hyperscaler shift into high-volume demand for 1.6T pluggables,” said James Lee, President of Gemtek, adding that automated manufacturing combined with the NewPhotonics PIC is intended to support “fast, reliable time-to-market production.” NewPhotonics SVP and GM of Optical Connectivity Doron Tal said the companies are combining “laser-integrated PIC” technology with “high-volume manufacturing capabilities” to deliver “power-efficient 1.6T connectivity for AI infrastructure,” and described it as “a new class of 200G-per-lane pluggables designed to meet the market’s urgent need for 1.6T optics at hyperscale.”

Source: Gemtek

Get Data Center Engineering News In Your Inbox:

Popular Posts:

mike-nolan-PPST
How to burn-in test megawatt power equipment without drawing a megawatt from the grid
Xendee
Bring your own power: Why data center microgrids are shifting from backup posture to strategy
Why your data center needs 135% cooling capacity- Munters
Why your data center needs 135% cooling capacity
Elvis-Leka,-New-Product-Development-Engineer-—-Parker,-Sporlan-Division
From air to two-phase liquid: how rack cooling options compare on density and risk
Screenshot
Five AI data centers to reach 1 GW power capacity in 2026, new analysis shows

Share Your Data Center Engineering News

Do you have a new product announcement, webinar, whitepaper, or article topic? 

Get Data Center Engineering News In Your Inbox: