IN Brief:
- IQE will manufacture the $14m multi-year production order at its Newport wafer facility.
- The components will support AI data-centre applications, including high-performance storage and data communications.
- Related development covers indium phosphide, silicon photonics, and gallium arsenide VCSEL technologies.
IQE has secured a $14m multi-year production order for compound-semiconductor wafers used in artificial intelligence and data-centre systems, with manufacturing scheduled to take place at its Newport facility in Wales.
The order covers components for high-performance storage and data communications, both of which are under growing pressure as AI accelerators exchange larger datasets with memory, storage, network interfaces, and neighbouring processors. IQE is also undertaking related development work in indium phosphide optical communications, silicon photonics, and gallium arsenide vertical-cavity surface-emitting laser technology.
Although processor performance remains central to AI systems, overall throughput is increasingly limited by the energy and delay associated with moving data. Electrical connections become progressively harder to operate efficiently as bandwidth and distance rise, encouraging greater use of optical links within and between data-centre racks.
Indium phosphide supports high-speed lasers, modulators, detectors, and amplifiers at wavelengths used for optical communications, while gallium arsenide VCSELs serve shorter-reach links and sensing applications. Silicon photonics provides a route to integrating waveguides and optical functions through manufacturing processes derived from established semiconductor production.
These technologies are converging as equipment manufacturers move optical interfaces closer to switches, processors, and accelerators. Pluggable transceiver modules remain widely used, but co-packaged and near-packaged optics can shorten high-speed electrical paths, reducing signal loss and the power consumed by retimers and interface circuitry.
Placing optical engines close to high-power silicon also creates new challenges. Lasers, modulators, fibre coupling, control electronics, and thermal structures must coexist beside processors dissipating substantial heat, while the complete package must remain manufacturable, testable, and serviceable at data-centre scale.
Compound-semiconductor wafers sit near the beginning of that chain. Uniform epitaxial layers, controlled doping, low defect density, and repeatable thickness are essential because variations introduced during wafer growth can reduce yield or alter optical performance after multiple downstream fabrication stages.
The Newport order follows an indium phosphide supply agreement between IQE and Tower Semiconductor, extending the company’s involvement in optical communications as data-centre architectures shift towards higher-bandwidth links.
Multi-year production commitments can improve factory utilisation and planning, but compound-semiconductor manufacturing remains exposed to abrupt changes in customer demand. Specialist epitaxy equipment and process teams carry high fixed costs, leaving profitability sensitive to yield, product mix, and the proportion of installed capacity that remains in regular use.
Material supply introduces another variable. Indium, gallium, and related inputs are produced through geographically concentrated supply chains and are exposed to export controls, price volatility, and competing demand from displays, communications, defence, and power electronics.
Qualification can restrict the ability to respond quickly when supply conditions change. Once an epitaxial structure has been validated for a laser, detector, or optical module, moving to another wafer source may require fresh device fabrication, reliability testing, and system-level assessment, even when the alternative material appears nominally equivalent.
Europe retains strong research and manufacturing capability in compound semiconductors and integrated photonics, although production remains fragmented across specialist suppliers, pilot lines, foundries, and packaging houses. A TNO and ASML photonic-chip pilot line is intended to reduce the gap between laboratory processes and repeatable industrial output.
That transition requires common design rules, process controls, packaging methods, and test procedures. Photonic devices are more sensitive than conventional digital chips to alignment, surface quality, coupling geometry, and material variation, while their performance often depends on the complete optical and mechanical assembly.
Newport’s role carries strategic weight because the UK retains an established cluster of compound-semiconductor research, epitaxy, device fabrication, and packaging expertise. Preserving that position will require sustained production volumes alongside research funding, since process knowledge weakens when facilities operate mainly through short development programmes.
IQE expects the AI infrastructure cycle to create further demand for optical and storage-related components. The $14m award provides a defined production programme, while the associated indium phosphide, silicon-photonics, and VCSEL work could broaden its scope if customers progress from device development into larger deployments.
