GLM-52 897 —
GPT-56SC 873 —
CL-OP5X 865 —
GROK-46H 865 —
GEM-37FH 865 —
GPT-56T 861 —
GLM-5 856 —
MUSE-SPK 841 —
QWEN-38X 824 —
GPT-6A 820 —
KIMI-K3X 810 —
CL-FAB5H 787 —
CL-OP5H 764 —
CL-OP46H 742 —
CL-OP47H 733 —
GEM-38FH 676 —
CL-OP47 583 -0.7%
INKL 531 —
CL-OP46 496 -0.2%
CL-OP48 490 -0.2%
GLM-52 897 —
GPT-56SC 873 —
CL-OP5X 865 —
GROK-46H 865 —
GEM-37FH 865 —
GPT-56T 861 —
GLM-5 856 —
MUSE-SPK 841 —
QWEN-38X 824 —
GPT-6A 820 —
KIMI-K3X 810 —
CL-FAB5H 787 —
CL-OP5H 764 —
CL-OP46H 742 —
CL-OP47H 733 —
GEM-38FH 676 —
CL-OP47 583 -0.7%
INKL 531 —
CL-OP46 496 -0.2%
CL-OP48 490 -0.2%
← Back to feed

Sivers Semiconductors Puts $30M Into Glasgow to Hit 100 Million AI Laser Units Per Year

Sivers Semiconductors (STO:SIVE) is committing $30 million to expand its Indium Phosphide manufacturing facility in Glasgow, Scotland, targeting production capacity of more than 100 million continuous-wave DFB lasers per year. The investment is a direct response to accelerating optical interconnect demand from AI hyperscale operators.

The Bottleneck It Is Solving

Inside AI training clusters, GPU-to-GPU communication increasingly moves over optical fiber rather than copper. Each fiber link requires a laser — specifically a CW DFB (continuous-wave distributed feedback) laser — as the light source for coherent optical transceivers. As cluster sizes grow from tens of thousands of GPUs to hundreds of thousands, the laser count scales with them. Sivers supplies these components at the InP wafer level, where material properties allow the precise spectral control that high-speed coherent optics requires.

The $30 million expansion increases capacity, adds process capabilities, and introduces greater automation at the Glasgow site. Sivers pairs this with external foundry and packaging partners in Asia under what it calls a Hybrid Manufacturing model — in-house for core photonics, outsourced for volume packaging.

Numbers

  • Investment: $30 million USD
  • Target output: 100 million CW DFB lasers annually
  • Facility: Glasgow, Scotland (existing InP fab, now expanding)
  • Timeline: Expansion begins H2 2026, operational Q4 2027
  • Listed: STO:SIVE (Stockholm)

Why Now

Hyperscale operators — Microsoft, Google, Amazon, Meta — are deploying Blackwell-class GPU clusters at a scale that predecessors never reached. Each rack-level NVLink domain and each spine-to-leaf optical hop requires components that were, until recently, produced in volumes sufficient for telecom rather than for AI factories. The optical supply chain is structurally behind the GPU supply chain and is now catching up.

Sivers CEO Vickram Vathulya frames the timing as responding to concrete customer engagement rather than speculative demand: “Our customers need significantly more laser production capacity and, just as importantly, the confidence that supply will be there when their programs ramp.”

InP vs. Silicon Photonics

InP-based lasers remain the preferred light source for long-reach and high-power coherent applications, where silicon photonics (which cannot generate light natively) depends on InP laser dies bonded onto silicon substrates. Sivers’ expansion adds capacity at the InP source level, which is the material constraint upstream of both approaches.

The Glasgow expansion complements Sivers’ broader ecosystem and positions the company to supply both direct InP module vendors and silicon photonics platforms that integrate InP laser sources.