Michael Hurlston – CEO, Lumentum

Why Lasers Are the Invisible Infrastructure of the AI Economy

The compute layer of a modern AI data center gets most of the attention. Nvidia GPUs, Google TPUs, AMD accelerators, and the companies behind them dominate the headlines. Storage has become a major story, with companies like Micron and Sandisk capturing investor attention. But none of it works without connectivity. Compute needs to talk to storage. Storage needs to talk to switching. Switching needs to talk back to compute. For years, that connectivity ran on copper Ethernet cables, the same basic technology that runs out of the wall in homes and offices. As data center speeds have climbed toward 1.6 terabits per second and beyond, the resistive nature of copper has run out of runway. The physics simply do not support it anymore.

Fiber optics solve that problem, and lasers are the component that makes fiber optics work. Lumentum manufactures the laser chips that sit at the heart of fiber optic transceivers, enabling data to travel at the speeds modern AI infrastructure demands. The material that makes those laser chips possible is indium phosphide, a semiconductor compound that is far less widely produced than silicon. When Nvidia looked at the trajectory of data center buildout and the conversion from copper to optical connectivity, it identified a supply risk and acted on it, investing two billion dollars in Lumentum to secure access to indium phosphide-based laser chips over the next four to five years. Lumentum used a portion of that investment to fund a new fabrication facility in Greensboro, North Carolina, giving the company domestic production capacity for a material that currently comes primarily from Japan and China.

From Telecom to Data Center: The Reboot That Changed Everything

Lumentum’s core technology was not invented for data centers. It was built for the telecom industry, powering the fiber optic cables that carry internet traffic across continents and under oceans. For two decades, the company served telecom operators and cable providers, helping wire the United States and connect the global internet backbone. That business hit a significant reset in 2021 and 2022 as the telecom capital expenditure cycle contracted sharply. The board recognized that the company had the right technology at exactly the wrong moment in the telecom cycle, and that the data center opportunity was emerging faster than anticipated. They brought in Hurlston to lead the transformation.

The reboot required changes across every dimension of the business. The people who built Lumentum into a successful telecom supplier were experienced at managing production runs in the hundreds and low thousands per quarter. The data center opportunity was demanding hundreds of thousands, potentially millions, and in some cases hundreds of millions of units per quarter. That is not an incremental scaling challenge. It is a fundamentally different operational model, and it required bringing in new leadership with experience at that scale while preserving the engineering talent that had invented the products on which the growth was being built. “The products that we have were invented before I got here,” Hurlston said. “Credit to the prior CEO. The issue was whether we had the right DNA in the company to handle the growth curve that was coming.”

Building a Skyscraper on Quicksand: The Platform Challenge

The operational infrastructure Lumentum had built for a telecom-scale business was not designed to support data center-scale volumes. IT systems, manufacturing processes, supply chain management, and forecasting capabilities all needed to be rebuilt or significantly upgraded while the company was simultaneously trying to meet exploding customer demand. Hurlston describes it as building a skyscraper on quicksand, getting the foundation solid enough to support the structure being erected on top of it, without the luxury of pausing operations to do so.

One of the most significant structural advantages Lumentum has in navigating this challenge is vertical integration. Unlike Nvidia, Broadcom, and most of the major semiconductor companies, Lumentum owns its own fabrication facilities and factories. It does not outsource manufacturing to TSMC or similar foundries. That means the company controls its own production destiny, which is a strategic advantage in a supply-constrained market, and a significant operational challenge when scaling rapidly. Finding experienced leaders who know how to run semiconductor fabs and high volume factories is not easy, and building that management layer has been one of Hurlston’s primary focuses since joining. On top of that, the company is using AI internally to improve forecasting accuracy, eating its own dog food in terms of applying intelligent analytics to the long-cycle investment decisions that a three-year manufacturing buildout requires.

The Pure Play Bet on Data Centers

One of the most consequential strategic decisions Hurlston has made at Lumentum is to make the company more focused rather than more diversified. Industrial lasers, used in factories for cutting, welding, and machining, had historically been part of Lumentum’s portfolio. Rather than expanding in that direction as industrial automation grows, the company has deliberately de-emphasized those product lines and redirected the engineering resources and capital toward the data center opportunity. The reasoning is straightforward: the data center conversion from copper to optical is the largest near-term market opportunity Lumentum has ever faced, and splitting focus dilutes the ability to capture it.

That is a meaningful risk for a public company. Concentrating on a single market trend means the company is exposed if that trend slows, pauses, or reverses. Hurlston is direct about the risk and direct about why he believes it is the right call. The data analytics and AI-assisted forecasting the company now uses support the conclusion that data center demand will continue growing through the end of the decade and beyond. The conversion from electrical to optical connectivity inside data centers is still in its early stages. And the timeline for the new North Carolina fab to reach full production aligns with when that conversion is expected to accelerate most significantly. “We’re betting that in 2028 this music won’t have stopped,” he said. “And we’re going to continue to invest until the analytics tells us there is a turn in the market.”

The Catastrophic Failure That Built a CEO

Before Lumentum, before Broadcom, there was a startup that went bankrupt. In the late 1990s, after a string of early career successes including IPOs and a company sale, Hurlston co-founded a semiconductor company targeting the transition from analog to digital over-the-air television. The FCC delayed the regulatory transition that would have driven adoption of the company’s products. Facing a timeline that had stretched beyond what the business could sustain, the team pivoted to broadband Ethernet over wireless, a completely different market with different customers, different development cycles, and different competitive dynamics. The pivot did not work. The company went bankrupt while friends and peers in the industry were making money during the dot-com boom.

The failure was complete and humbling. Hurlston describes it as doing the exact opposite of everything the business school textbooks recommend, in real time, with his own company. What came out of it was a decision to go to Broadcom, then a much smaller company than it is today, specifically to learn how a high-growth technology company was actually built and run at scale. That apprenticeship became the foundation for everything that followed. “I certainly wouldn’t be here today if that had been successful,” he said. “It was a massive lesson learned, no doubt about it.” The experience of building something that failed completely, and then rebuilding from it, is the same muscle Lumentum is exercising now as it transforms from a telecom supplier into a data center infrastructure powerhouse.