Jensen Huang stood on a dusty lot in Sherman, Texas, this week and declared that the next industrial revolution would be built not on steel and steam, but on lasers and artificial intelligence. The Nvidia CEO was in town to mark a pivotal moment: Coherent Corp., a major photonics supplier, broke ground on an expanded indium phosphide semiconductor manufacturing facility, fueled in part by a massive $2 billion partnership with Nvidia that Huang said would help satisfy the world’s insatiable appetite for AI infrastructure. The move cements Texas as a critical node in the rapidly evolving AI supply chain — but also reignites urgent questions about how the state, and the nation, will power these energy-hungry factories of the future.
The event, held at Coherent’s existing Sherman campus, drew local officials, industry executives, and a swarm of press who came to hear Huang outline his vision of “AI factories” — dedicated facilities that churn out not physical goods, but the computational intelligence driving everything from autonomous vehicles to drug discovery. “This is not a metaphor,” Huang told the crowd, gesturing toward the construction site where new cleanrooms will soon rise. “We are building the factories where AI itself gets manufactured.” While Huang has used the term before, his Texas appearance connected the concept to tangible hardware: the optical transceivers and components that stitch together GPUs inside massive data centers, all using Coherent’s indium phosphide lasers.
A $2 Billion Handshake for the Next-Gen AI Backbone
The financial details of the Nvidia-Coherent deal remain partly under wraps, but sources familiar with the agreement confirmed that Nvidia has committed to an upfront prepayment of roughly $2 billion to secure a multi-year supply of photonics components. That capital gave Coherent the confidence to push ahead with the Sherman expansion, which will create more than 600 new high-tech manufacturing jobs in Grayson County and double the facility’s output capacity for wafer-level optics. The expanded site is expected to begin volume production in early 2026, just as Nvidia’s next-generation GPU architectures — codenamed Rubin and beyond — demand exponentially more networking bandwidth.
Indium phosphide (InP) plays a starring role because it enables high-efficiency lasers and photodiodes that transmit data at terabit-per-second speeds over fiber optic cables. Unlike silicon-based photonics, InP offers a direct bandgap, making it ideal for emission and detection of light at telecommunication wavelengths. As AI clusters grow to 100,000 GPUs or more, the network fabric becomes the bottleneck; every GPU needs low-latency, high-bandwidth connections to its neighbors. Coherent’s transceivers, which will use the Sherman-produced InP chips, are designed to scale to 1.6 terabits per second per lane — a leap that Nvidia’s Spectrum-X and InfiniBand switches can exploit.
“AI is essentially an extreme-networking problem disguised as a semiconductor problem,” said Dr. Julie Sheridan, a photonics analyst at Lightwave Insights, noting the industry’s pivot toward co-packaged optics. “Nvidia isn’t just securing supply; it’s making a strategic bet that optical innovation will keep pace with compute, and Coherent is one of the few companies that can deliver at scale.”
Texas Becomes the Unlikely Photonics Hub
Sherman, a quietly ambitious city about 65 miles north of Dallas, is rapidly transforming into a semiconductor and photonics powerhouse. Texas Instruments is building a $30 billion chip complex nearby, and GlobalWafers has a massive silicon epiwafer plant under construction. Now Coherent’s expansion adds a specialized photonics dimension that sets the region apart. The company already operates a significant gallium arsenide (GaAs) fab in Sherman, so the InP line deepens its expertise in compound semiconductors — materials essential for radio frequency and optical devices.
The facility will produce ultra-precision epitaxial wafers, laser chips, and integrated optical engines in cleanroom environments that rival those of advanced logic fabs. Coherent expects to leverage existing talent from the University of Texas at Dallas and Texas A&M, which have strong photonics and electrical engineering programs, to staff the new roles. The jobs span process engineers, epitaxy specialists, vacuum technicians, and optoelectronic packaging designers — positions that command salaries well above the local median and reinforce Texas’s claim as a destination for high-tech manufacturing reshoring.
But the growth comes with headwinds. The Sherman area, like much of Texas, faces mounting challenges in providing reliable electricity and water to industrial users. The state’s grid, famously strained during Winter Storm Uri in 2021, has been racing to accommodate new semiconductor fabs, which consume staggering amounts of power and ultrapure water. A single large fab can use as much electricity as a small city, and the addition of Coherent’s new line — while not as power-intensive as a logic fab — still adds a meaningful new load to the Electric Reliability Council of Texas (ERCOT) grid.
The Power Constraint Nobody Wants to Talk About
Huang didn’t shy away from the power issue during his appearance, calling it “the fundamental challenge of our generation.” He framed the problem in terms that resonated with the Texas crowd: abundant, low-cost energy was once the state’s best calling card, and it needs to be again. “If we want to build these AI factories, we need to build them where the power is — or we need to generate the power ourselves,” Huang said. He suggested that some future data centers could be colocated with dedicated renewable or nuclear power sources, citing recent Nvidia investments in energy-aware scheduling and liquid-cooling technologies that slash facility-level PUE (power usage effectiveness).
Texas is adding new generating capacity at a record pace — mostly solar, wind, and natural gas peaker plants — but transmission bottlenecks and permitting delays mean that large industrial users often face multi-year waits for grid hookups. Coherent worked with Oncor, the local transmission provider, to secure a dedicated substation upgrade, but the process took over two years of negotiation. Critics warn that without sweeping policy reforms, the state’s ambition to become a chipmaking hub could outstrip its grid’s ability to deliver electrons.
“Photonics is actually less power-hungry than electronic switching for the same bandwidth, so in a sense, technology like Coherent’s helps,” said Dr. Alan Hu, an energy systems researcher at the University of Texas. “But the sheer scale of AI means any efficiency gains get swallowed by demand growth. We’re going to need every tool — from GaN power supplies to even more advanced cooling — to keep the lights on.”
Riding the AI Demand Wave
The timing of the expansion is no accident. Nvidia’s data-center revenue topped $30 billion in its latest fiscal quarter, driven almost entirely by AI deployments. The company has guided for sequential growth every quarter, and its supply chain partners — from TSMC to memory makers to Coherent — are scrambling to keep up. Coherent’s existing Sherman operations were already running at near-full capacity before the expansion, and lead times for advanced optical modules had stretched to over six months.
By investing directly in manufacturing capacity, Nvidia is applying the same playbook it used with TSMC decades ago: upfront cash to lock in supply in exchange for dedicated lines. The $2 billion prepayment is not a loan but a guaranteed purchase commitment, meaning Coherent carries minimal risk. Huang said the model could be extended to other critical components, including chip substrates and advanced packaging, where capacity is equally tight. “We’re going to see more of these direct investments,” he predicted. “The AI supply chain is too important to leave to chance.”
Jobs, Training, and the Human Factor
Beyond the headline numbers, the Coherent expansion highlights a workforce dilemma. Semiconductor and photonics manufacturing demands highly specialized skills, and the local pipeline, while growing, can’t yet fill 600 new positions overnight. Coherent has pledged to partner with Grayson College and local workforce boards to create tailored training programs in semiconductor equipment maintenance, epitaxial growth, and cleanroom protocols. Some programs will offer paid apprenticeships, with graduates guaranteed interviews for full-time roles.
Jensen Huang spent part of his visit touring the existing plant and speaking with workers, several of whom had been with Coherent for over a decade. In a brief Q&A with employees, he emphasized that the AI boom is not just about engineers in Silicon Valley but about “the hands that build the machines.” He drew parallels to the automotive industry’s transformation in the 20th century, suggesting that towns like Sherman could become the Detroit of the AI age — a metaphor that resonated locally, given Sherman’s manufacturing heritage in everything from cotton to oilfield equipment.
What It Means for Windows and the Broader Ecosystem
For the Windows enthusiast community, the news is more than geopolitics and macroeconomics. The optical components coming out of Sherman will eventually find their way into the networking hardware that connects Nvidia GPUs in Azure, AWS, and on-premises data centers — the very infrastructure that powers Copilot, Windows Studio Effects, and the growing array of AI features integrated into Windows 11 and beyond. Faster, more reliable networking translates to lower latency for cloud AI services and better scaling for enterprise workloads that require synchronized GPU clusters.
Microsoft, a major Nvidia customer, has been aggressively expanding its data-center footprint to support AI, including a rumored “Stargate” supercomputer project that could cost over $100 billion by 2028. Any supply disruption in photonics could ripple into delayed deployment of AI capabilities on Windows platforms. Thus, Nvidia’s investment in domestic manufacturing also serves as a hedge against geopolitical tensions that could threaten Asian supply lines, where the majority of advanced photonics components are currently fabricated.
Looking Ahead: The AI Factory Realized
As Coherent’s construction crews begin pouring foundations this summer, the Sherman site represents a microcosm of the AI revolution’s physical footprint. It’s a place where lasers and laypeople converge, where a $2 billion bet from a trillion-dollar company can jump-start a regional economy, and where the abstract concept of “AI factories” takes concrete form. Huang, ever the visionary, ended his remarks by declaring that the facility would one day be viewed as a turning point. “When you look back in 10 years, you’ll say, ‘That’s where it all started,’” he said. “That’s where we stopped thinking about AI as a software thing and started building it like the industrial giants did — with steel, glass, and light.”
Yet sitting in the Texas sun, the reality of power lines, water pipes, and electrical substations reminded everyone that the digital future has an analog backbone — one that requires careful planning, massive investment, and a delicate balance between ambition and infrastructure. For now, Sherman is all in, and so, it seems, is Nvidia.