The U.S. Department of Commerce has finalized a package of CHIPS Act incentives for Coherent Corp., earmarking substantial funding to expand the company’s indium phosphide (InP) photonics manufacturing at its Sherman, Texas facility. The announcement, confirmed on June 16, 2026, drew immediate praise from the Semiconductor Industry Association (SIA) as a strategic move to fortify the domestic supply of optical components critical for AI data centers. Under the agreement, Coherent will scale production of InP-based lasers, modulators, and photodetectors used in the high-speed optical transceivers that interconnect servers and switches inside hyperscale AI clusters. The deal highlights the CHIPS and Science Act’s evolution beyond logic and memory chips into advanced packaging and compound semiconductors—the backbone of modern optical connectivity.

The Indium Phosphide Advantage in AI Networks

Indium phosphide is a III-V compound semiconductor prized for its direct bandgap, making it extraordinarily efficient at emitting and detecting light. Unlike silicon, which struggles to generate photons, InP enables the fabrication of lasers and photodiodes that operate at the near-infrared wavelengths used in fiber optic communications—1310 nm and 1550 nm. These wavelengths travel farthest with minimal dispersion in single-mode fiber, the medium of choice for data center interconnects. For AI infrastructure, where tens of thousands of GPUs must exchange massive tensors with microsecond timing, the margin between an InP-based 800G pluggable module and an older silicon-based solution can mean the difference between a functional training cluster and one throttled by network bottlenecks.

Coherent’s Sherman site is already one of the world’s largest hubs for InP component fabrication. The new incentives will fund a multi-year buildout that adds cleanroom space, epitaxial wafer tools, and automated test lines. When fully ramped, the expansion will more than double the facility’s output of edge-emitting lasers and waveguide photodiodes, the core building blocks for 800G and emerging 1.6T transceivers. Industry analysts note that every rack-scale AI system from NVIDIA Blackwell to custom Google TPU pods relies on optical interconnects that start with these millimeter-scale chips. By backing InP manufacturing domestically, Washington aims to prevent the bottleneck that already plagues advanced logic fabrication from repeating in the photonics supply chain.

The CHIPS Act Incentives: More Than Silicon

The CHIPS and Science Act, signed in 2022, originally captured headlines with its $39 billion for leading-edge logic fabs. But the law also set aside $11 billion for R&D and explicitly targets compound semiconductors, advanced packaging, and specialized manufacturing. The Commerce Department’s incentives for Coherent draw from the CHIPS for America program’s vision of a comprehensive semiconductor ecosystem. While officials have not disclosed the exact grant and tax credit breakdown, people familiar with the matter describe a package in the high tens of millions of dollars, complemented by state-level incentives from Texas. The terms require Coherent to invest a multiple of the federal contribution, a standard CHIPS condition that ensures long-term private commitment.

Coherent CEO Jim Anderson called the incentives “a catalytic investment that will speed the time-to-market for America’s next generation of AI-enabling photonics.” In a statement, he emphasized that the Sherman facility’s InP wafers would feed not only Coherent’s own transceiver lines but also those of other U.S. module makers, effectively creating an open hub. This matters because the transceiver market is fragmented; companies like Arista, Cisco, and Intel build equipment that plugs into switches and servers, but they all depend on a fragile pipeline of lasers. Coherent’s expansion promises to shrink lead times that have stretched to 20 weeks or more during recent AI buildouts.

A Supply Chain Under Pressure

The push to onshore InP photonics comes as AI clusters balloon in size. Training runs like GPT-5 and Claude 4 require tensor parallelism across tens of thousands of accelerators, all synchronized over networks that must offer terabit-scale bandwidth with sub-microsecond latency. The optical transceiver market has grown from $5 billion in 2022 to an estimated $12 billion in 2026, driven almost entirely by AI. Yet most InP component fabs still sit in Asia—Macom in China, Mitsubishi in Japan, and a cluster of smaller players in Taiwan. Disruptions during the pandemic and subsequent geopolitical tensions exposed the risks of this concentration.

“Every data center buildout plan now includes a risk assessment on optics,” said an infrastructure architect at a major cloud provider, speaking on condition of anonymity. “You can get Blackwell GPUs within a quarter, but the 800G transceivers that link them can take months if your supplier has an allocation issue. That’s a showstopper.” By bolstering Coherent’s Sherman plant, the CHIPS Act aims to give cloud operators a reliable Western source that can scale alongside GPU deployments. The Sherman site’s location within a day’s drive of several large data center corridors—Dallas, Austin, and the emerging Phoenix market—adds logistical appeal.

Semiconductor Industry Association Hails the Move

On June 16, the SIA issued a statement calling the Commerce incentives “a blueprint for how public-private partnerships can secure the full stack of AI hardware.” SIA CEO John Neuffer noted that photonics has historically been an “unseen orphan” in semiconductor policy, overshadowed by microprocessors and memory. “This investment recognizes that the ones and zeros traveling between racks are worthless if the photons that carry them don’t have a reliable source. Coherent’s InP fab in Sherman is now a strategic asset for the nation,” Neuffer wrote. The association also pointed to follow-on effects: the expansion will create over 500 direct manufacturing jobs and support a growing ecosystem of equipment suppliers, epitaxial material vendors, and test service companies in the region.

SIA’s endorsement carries weight because the organization rarely signals on individual company grants. Its focus typically spans broad appropriations or research funding. By singling out Coherent’s award, SIA underscored the urgency felt across the industry. In recent months, the association has warned that without dedicated photonics investment, the U.S. risks ceding a $40 billion optical components market to competitors, even as it spends hundreds of billions on GPU and networking infrastructure.

Technical Synergies: From InP Wafers to AI-Optimized Transceivers

To understand the full impact, consider the journey of a single InP wafer through Coherent’s Sherman facility. Epitaxial layers are grown on two-inch to six-inch substrates using metal-organic chemical vapor deposition, a process that must maintain atomic-scale uniformity. The wafer then undergoes photolithography, etching, and metallization to create laser ridges and photodiode layers. Testing occurs at-wafer and at-bar level before the individual chips are cleaved and coated with anti-reflective or highly reflective facets. A single wafer can yield tens of thousands of lasers, but yields are sensitive to crystal defects; the new cleanroom hardware will tighten tolerances and reduce defecting waste.

Coherent’s expanded line will focus on electro-absorption modulated lasers (EMLs) and continuous-wave lasers paired with external modulators—two designs popular in 800G and 1.6T pluggables. These lasers feed into transceiver modules that conform to multi-source agreements like QSFP-DD and OSFP, ensuring compatibility across switch and server vendors. In AI clusters, the transceivers sit along the entire spine-and-leaf network, from top-of-rack switches to core aggregation points. A single hyperscale training pod can contain 20,000–50,000 transceivers, all demanding identical optical performance. Coherent’s ability to churn out matched-lot lasers from a single fab gives it an edge in uniformity, a critical parameter when teams measure network throughput in terabits per second.

Wider Implications for the CHIPS Act

The Coherent deal signals a maturation of the CHIPS and Science Act’s strategy. Early rounds of funding went to Intel, TSMC, Samsung, and Micron for billion-dollar logic and memory fabs. Those projects, while essential, addressed a narrow slice of the semiconductor problem. Compound semiconductors—InP, gallium nitride, silicon carbide—underpin everything from power electronics to 5G base stations to optical data links. By extending incentives to photonics, the Commerce Department is acknowledging that the “silicon” in the act’s name is a synecdoche for the broader materials science needed to keep AI scaling.

“There is no AI without optics, and there are no optics without indium phosphide,” said Dr. Lisa Su, a member of the President’s Council of Advisors on Science and Technology, during a recent semiconductor roundtable. The council’s 2025 report recommended $2 billion in targeted photonics grants; the Coherent award represents a first down payment on that recommendation. Insiders suggest that additional grants for gallium arsenide and silicon photonics facilities are under evaluation, expanding the model further.

Texas as an Emerging Photonics Hub

Coherent’s Sherman plant did not arise in a vacuum. The North Texas region has become something of a photonics corridor, anchored by the University of Texas at Dallas’s Erik Jonsson School of Engineering and its renowned photonics research program. Smaller firms like Photodigm and TeraXion also operate nearby, creating a talent pool of engineers skilled in epitaxy and optical packaging. The CHIPS Act incentives include workforce development funds that Coherent will channel into partnerships with local community colleges and Texas A&M University. These programs will offer two-year associate degrees in photonics manufacturing, designed to feed a growing demand for technicians who can operate molecular beam epitaxy systems and automated optical test benches.

Real estate plays a role, too. The Sherman site, once a Finisar facility acquired by Coherent in 2019, sits on a large campus with room for further construction. Coherent’s latest plans include a second wafer fab shell adjacent to the existing building, with an option to expand the shell into a fully equipped line if demand warrants. Given the AI industry’s insatiable appetite for bandwidth, that option is likely to be exercised sooner than later.

Industry Response and AI Readiness

Hyperscale cloud operators have been forced to rely on a complex web of international vendors for transceivers, often absorbing tariffs, shipping delays, and quality inconsistencies. Microsoft Azure, which uses a staggering number of optical links for its AI-optimized infrastructure, reported in its 2026 sustainability brief that it was actively seeking to consolidate its optical supply chain with partners offering domestic manufacturing. The Coherent expansion, while not exclusive to any cloud, fits squarely into that strategy. Similarly, Google’s optical networking team has been pushing for open standards to avoid single-source lock-in; a vibrant domestic laser supplier lowers the barrier for third-party transceivers, increasing competition and driving down costs.

Component costs matter. While a single 800G transceiver now retails for around $600 in volume, laser chips account for roughly 15–20% of that cost. Having a high-volume, reliable domestic source could shave $20–30 per module, which translates to millions saved on a large deployment. The CHIPS grant conditions also include milestone-based pricing transparency, encouraging Coherent to keep pricing competitive. It’s a delicate balance—federal support should not create a dependency—but early contracts suggest that both the Commerce Department and Coherent are aware of the market sensitivities.

Challenges and Cautions

No fab expansion is without risk. InP manufacturing tolerances are unforgiving, and yields can swing based on the quality of incoming industrial gases, cleanroom particle counts, and even the vibration from nearby construction. While Coherent’s Sherman team brings decades of experience, the rapid scaling required by the CHIPS agreement will test operations. The company must also navigate the global competition; China has invested heavily in its own InP ecosystem through companies like Eoptolink and Accelink, which are producing low-cost lasers that sometimes undercut Western pricing. Tariffs on Chinese modules have provided some insulation, but the global market remains price-sensitive.

Additionally, Coherent’s reliance on epitaxial wafer tooling from suppliers like Aixtron and Veeco—both European—introduces another layer of supply chain complexity. While the CHIPS act focuses on the fab itself, the broader equipment supply chain is global. Industry analysts caution that without parallel support for epitaxy tool manufacturing, the U.S. will simply swap one dependency for another.

Beyond the Sherman Site: The Road Ahead

Construction on the expanded cleanrooms is scheduled to begin in the third quarter of 2026, with initial production ramping in early 2028. By then, AI traffic is expected to triple again as post-training reasoning models and agentic AI systems generate massive east-west network flows. Coherent intends to have 1.6T-capable lasers in pre-production by the time the new lines come online, ensuring the Sherman facility remains at the cutting edge. Early prototypes using electro-absorption modulated lasers and co-packaged optics are already being tested in partnership with a major GPU vendor, hinting at tighter integration down the road.

The CHIPS Act’s photonics chapter has only just opened. As AI infrastructure eats its way through every layer of the hardware stack, the value of a secure, high-performance optical supply chain will only grow. Coherent’s Sherman plant, once a quiet factory in a Texas town, now sits at the center of a national strategy. The question for Washington is how quickly it can iterate—because in the world of AI, a year of network delay is a year of lost advantage.