Sivers Semiconductors AB and SemiNex Corporation have announced a jointly executed program to develop next-generation indium phosphide light sources aimed at powering AI data center interconnects.

The program, announced on August 13, 2026, is initially valued at approximately USD 3.4 million and targets a class of optical components that both companies describe as increasingly central to the scalability of AI infrastructure.

Program Scope and Technical Targets

The program encompasses three primary development areas: high-power external laser sources for co-packaged optics, high-channel-count distributed feedback laser arrays for wavelength-multiplexed links, and semiconductor optical amplifier gain stages designed to extend optical reach as channel counts scale.

The technical framing of the program reflects a broader shift in how the optical interconnect industry is approaching AI data center architecture. As optics move from the faceplate into the package, a transition referred to as co-packaged optics, power per wavelength, wall-plug efficiency, and thermal stability become increasingly critical variables.

According to the companies, these characteristics increasingly determine how far an architecture can scale, which repositions the light source from a component selection decision to a system-level design problem.

SemiNex, founded in 2003 and headquartered in Greater Boston, Massachusetts, designs and manufactures high-performance InP photonic devices including ultra-high-power optical amplifiers, DFB lasers, gain chips, and external cavity lasers.

Sivers Semiconductors, listed on the Stockholm Stock Exchange under the ticker SIVE, describes itself as a provider of energy-efficient photonics and wireless solutions, with laser and RF beamformer technologies serving customers in AI data centers, satellite communications, defense, and telecom.

Executives Frame Light Sources as Infrastructure Constraint

Both companies positioned the program as a response to what their leadership characterized as a fundamental bottleneck in AI infrastructure scaling.

Vickram Vathulya, Chief Executive Officer of Sivers Semiconductors, said the advanced light sources needed for next-generation optical networking within AI factories are as crucial as the compute, memory, and silicon photonics elements. He cited the partnership with SemiNex as a means of combining innovative technologies to deliver light source solutions at scale to the market.

Ronald Moore, Chief Executive Officer of SemiNex, described the optical layer as the constraint on how far AI infrastructure can scale. Moore stated that the industry requires more power, at more wavelengths, and at higher efficiency, and argued that those gains cannot be achieved by pushing existing devices harder.

He said the required improvements must come from the design and processing of the InP material itself, which he identified as the shared focus of the two engineering teams.

Emerging Architectures Driving Demand

The program is explicitly directed at optical requirements created by co-packaged optics as well as what the announcement describes as emerging scale-up and scale-out network architectures within AI data centers.

These architectural approaches place new demands on optical components, particularly as channel counts increase and interconnect distances vary across different parts of the network fabric.

The inclusion of SOA gain stages in the program reflects the specific challenge of maintaining signal integrity as the number of wavelengths carried on a single link grows.

By integrating amplification capability into the light source assembly, the companies are targeting a design approach that addresses optical reach without requiring separate amplification infrastructure at every stage of the network.

High-channel-count DFB arrays, also included in the program scope, are relevant to wavelength-division multiplexing approaches that allow multiple data streams to be transmitted simultaneously over a single optical fiber at different wavelengths. The ability to produce arrays with a large number of precisely controlled wavelengths is considered a key technical challenge in scaling these systems.

Production Timeline

Customer sampling and early production ramps under the program are targeted for the second half of 2027. The announcement did not provide detail on specific customer engagements or identify end customers for the components being developed.

Company Backgrounds

Sivers Semiconductors is headquartered in Kista, Sweden, and describes its mission as enabling a greener data economy through energy-efficient photonics and wireless technologies.

The company's photonics business is focused on high-precision laser solutions for applications including AI data center interconnects and telecommunications.

SemiNex was founded in 2003 and is based in the Greater Boston area of Massachusetts. The company's business model centers on designing and manufacturing custom InP photonic devices in partnership with technology companies requiring advanced optical solutions for specialized applications. Its product portfolio, as described in the announcement, includes ultra-high-power optical amplifiers, DFB lasers, gain chips, and external cavity lasers.

The two companies did not disclose the history or duration of their prior relationship, though the language used by Sivers Semiconductors CEO Vathulya referred to the collaboration as an existing partnership that the new program builds upon.

Tracking the Next Wave of Semiconductor Infrastructure Across Markets Worldwide

How confident are you that your organization has full visibility into the semiconductor construction and development pipeline unfolding across markets worldwide right now? From advanced fab expansions to new chip packaging facilities, the pace of capital investment in this sector leaves little room for missed intelligence.

The Global Project Tracking (GPT) platform by Blackridge Research is purpose-built to give semiconductor industry professionals a structured, continuously updated view of project activity at every stage of the development lifecycle. Whether you are monitoring competitor facility expansions, evaluating supply chain positioning, or identifying early-stage opportunities, the platform delivers the data depth that strategic decisions demand.

Semiconductor cycles move fast, and the organizations that act on project intelligence earliest consistently hold the strongest position. The GPT platform ensures your team is never working from incomplete or outdated information when it matters most.

• Upcoming Projects

• Tender Notices

• Contract Awards

• Projects Under Construction

• Completed Projects

To see how the Global Project Tracking (GPT) platform by Blackridge Research can sharpen your semiconductor market intelligence strategy, contact our team today to Book a Free Demo.