Coherent will quadruple InP wafer output in Sherman, funded by $650M and AI interconnect rush

Coherent plans to quadruple indium phosphide, or InP, wafer output at its Sherman, Texas, fab. The move is anchored by an eye-watering $650 million investment into the facility, with the company saying it will effectively double the factory’s footprint and quadruple InP wafer output as AI-driven optical interconnect demand rises.

This is not a vague “AI will grow” bet. It follows Nvidia CEO Jensen Huang’s public push for optics in AI systems, including his Computex comments that the technology would make Marvell the next trillion dollar company. Coherent, which is Nvidia-backed, is acting like supply chains can’t afford to be late when rack-scale systems demand more than copper can efficiently deliver.

Here is what Coherent actually makes, and why InP matters. The company operates eight wafer fabs across the US that produce semiconductors used in laser light sources and optical modules. InP semis are commonly employed in lasers, photodetectors, and modulators, which are core building blocks for optical interconnects. In other words, this is not a side quest vendor. It sits upstream of the optical components that help move data between accelerators and across the broader rack-scale AI stack.

The investment package is a mix of strategic capital and public support. Coherent plans to plow $650 million into its Sherman plant, backed by $20 million in funding from the Texas Semiconductor Innovation Fund and the Sherman Economic Development Corporation, and up to $50 million in CHIPS and Science Act funding. Separately, Nvidia invested $2 billion in Coherent in March to bolster its production capacity, underscoring that this is a priority supply-chain relationship, not a casual partnership.

That $2 billion is an important signal for execs who track where “AI spending” becomes “real manufacturing.” Nvidia cannot control every bottleneck in the semiconductor ecosystem, but it can influence which upstream players get enough capital to scale quickly. The source is blunt about the motivation: supply chains must be ready to meet optical interconnect demand when it materializes, and Coherent is one of the suppliers built for that moment.

Now zoom out to the market dynamics that make this kind of capacity expansion urgent. As rack scale AI systems grow from a few dozen accelerators to hundreds or thousands, copper is no longer sufficient and optics are required to reach that scale. That shift changes the economics of buildout. It typically increases the number of optical components in a given system over time, which means optical interconnect suppliers are suddenly operating under tighter scaling timelines than traditional datacenter component cycles.

If you’re a CFO, board member, or operator in a neighboring layer, the second-order implication is straightforward: optical interconnects are moving from “emerging architecture” to “infrastructure planning variable.” Capacity decisions like Coherent’s expansion create lead times, workforce requirements, and qualification timelines that can lock in who can ship at the right volumes. Coherent says the Sherman expansion will create about 1,000 new jobs, roughly 550 directly related to advanced manufacturing, engineering, and technical roles. The company has not shared a timeline for when the expansion will be completed, which is a reminder that scaling is as much about execution risk as it is about funding.

This is also part of a broader Nvidia playbook of bankrolling optics and silicon photonics ecosystems. This spring, Nvidia invested $2 billion in Lumentum, which produces optical products used in datacenters including pluggable transceivers, optical circuit switches, and laser modules. Less than a month later, Nvidia plowed another $2 billion into Marvell, in part to accelerate its silicon photonics roadmap. Taken together, the strategy looks like simultaneous investment across optical components, photonics-enabling semiconductor fabrication, and the silicon photonics roadmaps that connect the dots.

So what should decision-makers take from Coherent’s Sherman ramp? If AI interconnect demand is going to spike, someone has to be ready to manufacture the optical building blocks at volume. Coherent is positioning itself as that someone with a 4x output plan, a $650 million investment, and both private strategic backing and public funding. For peers tracking AI hardware supply chains, the message is clear: when optics move from “nice to have” to “must have,” capacity becomes the competitive moat.