Lumentum Joins Nasdaq-100 as AI Optical Interconnect Era Begins

Before the market opens on May 18, 2026, Lumentum Holdings (NASDAQ: LITE) will replace CoStar Group as a member of the Nasdaq-100 Index. This inclusion in the index solidifies what the market has been speculating on for months: optical interconnects are no longer a niche supply chain segment; they are core AI infrastructure. After rising 345% in 2025, Lumentum has now gone up 145% so far this year, making it one of the best performers in the global AI hardware theme.

Michael Hurlston, the CEO of the company, said that within two quarters, all of the available space until the end of 2028 will be booked. Directly, Nvidia has promised to pay $2 billion to secure most of Lumentum's production output.

Why Optical Interconnects Have Become the AI Bottleneck Nobody Priced

For the past three years, GPUs, HBM, and data centre power have been at the centre of the AI infrastructure story. A lot of people didn't pay much attention to optical interconnects, the technology that lets data move at the speed of light between thousands of GPUs. That is quickly changing, and Lumentum's addition to the Nasdaq-100 is the first sign that the market as a whole understands the change.

There is no doubt about the basic physics. When tens of thousands of GPUs or TPUs work together in a modern AI supercluster, traditional copper lines and electronic switching can't handle the high bandwidth needs without losing a lot of energy and heat. Optical interconnect technologies, such as co-packaged optics (CPO), silicon photonic switches, and optical circuit switching (OCS), use light signals instead of electrical ones. This makes bandwidth density, energy economy, and latency much better. Both Nvidia's GPU clusters and Google's TPU clusters need this, so Lumentum doesn't care about the GPU vs. ASIC debate that has been going on in AI hardware research.

BNP Paribas recently found that optical interconnects and memory chips are the two technology areas with the biggest supply-demand gaps in the world. These are also the two areas where European institutional investors want to put their money into AI infrastructure the most. Lumentum thinks that the compound annual growth rate (CAGR) of OCS shipments from 2025 to 2028 will be over 150%, with the goal of making $1 billion in annualised OCS income by 2027. Through 2030, AI data hubs are expected to need 85% more InP than they do now.

What Makes Lumentum Structurally Irreplaceable

In the supply chain for optical interconnects, lumentum is the hardest thing to copy. Broadcom, which focuses on switching ASICs, and Coherent, which works with a wide range of optical technologies, are platform-oriented players. Lumentum, on the other hand, is better at the parts that are harder to scale: InP lasers, EML (electro-absorption modulated lasers), UHP (ultra-high-power lasers), external laser sources, and OCS are all areas with tough technical challenges, low yields, and limited industrial capacity around the world.

Both present and next-generation AI network architectures can't work without these parts. The Spectrum-X and Quantum-X silicon photonic network switches from Nvidia use Lumentum's high-performance lasers and optical parts to make the switches more power-efficient and increase network bandwidth. Google has used a lot of Lumentum's R300/R64 OCS products in its Jupiter AI data centre network to support TPU systems. These products use MEMS optical circuits to make direct optical links between endpoints, skipping all electrical switching in the middle.

The numbers for expanding capacity back up the industry's promise: by the end of fiscal 2026, EML capacity will have grown by more than 50% compared to 2025; and about 40% of the InP capacity growth plan has already been moved forward. These commitments require a lot of money, take years to build, and can't be quickly copied by rivals.

What Nasdaq-100 Inclusion Actually Changes

In addition to the value signal, the index inclusion has three other clear effects. First, passive funds and exchange-traded funds (ETFs) that track the Nasdaq-100 handle more than $600 billion in assets around the world. Requiring institutions to buy Lumentum shares that track this index will greatly increase its institutional ownership and trading liquidity. Second, the addition changes Lumentum's value from a standard optical communications cyclical to an AI computing infrastructure growth stock. This changes the earnings multiples that institutional analysts use. Third, it gives customers confidence in the supply chain: Lumentum is no longer a niche parts seller; it is now a named member of the index that tracks the world's biggest tech companies.

This week, Corning and Nvidia revealed a partnership that will increase the capacity of optical fibres in the U.S. by 10 times. This supports the idea that AI capital expenditure is growing from the GPU layer to the whole optical connectivity stack. Because Lumentum is a component in that stack, it gains from every dollar spent on data centre optical infrastructure, no matter which hyperscaler or AI architecture does it.

The fact that Lumentum joined the Nasdaq-100 on May 18 is not a sentiment event but a structural trigger. Orders placed through 2028, Nvidia's promise to build 2 billion GPUs, OCS income of $1 billion by 2027, and an 85% expected success rate. Most semiconductor businesses would be jealous of the InP CAGR through 2030 because it shows how demand is likely to change over that time. The stock's 145% year-to-date gain and 345% return through 2025 mean that a lot of value has already been built into the price. The risk is that capacity growth will happen at a time when demand is highest. Hurlston's comment that the company is "increasingly falling behind demand" is not a sign of trouble, though. It's just a description of a market where supply is the problem, not sales. That is the place you want to be in AI technology.