AI needs more than chips: Why Lumentum and the optics trade are lighting up

Key points

• Optics is becoming an increasingly important part of AI infrastructure. As AI clusters grow, more data needs to move quickly and efficiently between GPUs, servers, racks and data centres.
• Lumentum is emerging as an important company to watch within the theme. AI-related demand, improving operating margins and Nvidia’s $2 billion strategic investment have increased attention on the company.
• The theme extends beyond one stock. Coherent, Zhongji Innolight, Eoptolink, Ciena and others sit across the optical ecosystem, while the recently launched Roundhill Photonics & Optics ETF (LYTE) provides a more diversified way to track the sector.

The AI infrastructure investment cycle remains powerful, but it is also broadening.

GPUs and memory still attract much of the attention, but increasingly large AI data centres also require faster networking, more power, cooling and greater connectivity.

That is putting another part of the AI infrastructure buildout into focus: photonics and optical networking.

AI systems can involve thousands of accelerators working together, creating huge amounts of traffic between chips, racks and data centres. Optical connections transmit data using light and can provide the high bandwidth and energy efficiency increasingly required as AI networks scale. Nvidia itself describes optical interconnects as important to improving the energy efficiency and resilience of large-scale AI networks.

We believe that as AI spending expands from individual chips towards entire AI factories, networking and connectivity could take a larger role in infrastructure spending. That does not necessarily mean every optics company benefits equally, but it makes the sector increasingly relevant when assessing the broader AI value chain.

Companies including Lumentum, Coherent and Applied Optoelectronics provide lasers, optical components and transceivers used to connect data-centre infrastructure.

Why optics is gaining investor attention

1. AI clusters are becoming larger and more connected

The more accelerators an AI system uses, the more data needs to move between them. That is helping drive adoption of faster optical technologies, including the transition towards 1.6T optical modules.

Lumentum said in its latest results that its cloud-module business was advancing 1.6T adoption, alongside growing demand for optical circuit switches and high-power lasers for co-packaged optics.

Our view: The investment significance is that rising AI compute demand does not only require more processors. It can also increase the amount and sophistication of networking equipment required around those processors. The risk is that technology standards evolve quickly, creating winners and losers even within a growing market.

2. Nvidia is putting billions behind the optical supply chain

In March 2026, Nvidia announced separate $2 billion investments in Lumentum and Coherent, alongside multiyear supply and technology agreements (Source: Nvidia press release, 2 March 2026).

For Lumentum, the agreement includes a multibillion-dollar Nvidia purchase commitment and future capacity rights for advanced laser components. Nvidia's agreement with Coherent similarly includes a multibillion-dollar purchase commitment covering advanced laser and optical networking products.

Our view: Nvidia's investments provide a strong signal that optical connectivity is becoming strategically important to future AI architecture. But strategic investment by a major customer should not itself be treated as evidence of future investment returns. Customer concentration, execution and the pace of AI infrastructure spending remain important risks.

3. Lumentum's earnings are starting to show the demand

Lumentum's fiscal fourth-quarter results (Source: Lumentum Holdings Fiscal Q4 and FY2026 results, 11 August 2026) showed how rapidly the business is expanding.

• Revenue reached $1.01 billion, up 109% year on year.
• GAAP operating margin rose to 27.8%, while non-GAAP operating margin reached 36.6%, versus 15.0% a year earlier.
• Management guided for fiscal first-quarter 2027 revenue of $1.225–1.275 billion and non-GAAP operating margin of 39.5–40.5%.

The company also reported a $7.2 billion GAAP net loss in the quarter. This was driven primarily by a $7.8 billion one-off, non-cash loss on debt extinguishment related to the equitisation of certain convertible notes. Non-GAAP net income was $326.3 million.

Growth is being supported by products including 1.6T cloud modules, optical circuit switches and high-power lasers used in emerging co-packaged optics systems.

Our view: The key development is that the optics thesis is increasingly visible in reported revenue and operating margins rather than relying only on expectations around future AI demand. However, current growth rates should not simply be extrapolated: capacity additions, competition, pricing pressure or slower hyperscaler spending could change the earnings trajectory.

4. US-China tensions could reshape the competitive landscape

The Trump administration is drafting restrictions on imports of new models of Chinese data-centre components, including optical transceivers (Source: Reuters, 4-5 August 2026). The measures remain under consideration and are not final.

Chinese companies including Zhongji Innolight and Eoptolink are important suppliers of optical modules to AI and cloud infrastructure.

Our view: Restrictions could potentially shift some demand towards non-Chinese suppliers such as Lumentum and Coherent. But the outcome is not straightforward. Restrictions could also raise equipment costs, tighten supply, disrupt global supply chains or trigger further Chinese retaliation.

Who are the major global optics players?

The optical ecosystem spans lasers, transceivers, networking equipment, fibre and semiconductor components.

Broadcom and Marvell also have exposure through networking and connectivity chips, although optics represents only one part of their broader semiconductor businesses.

Our view: The companies have very different business mixes, geographic exposures and competitive positions. Investors assessing the optics theme therefore need to distinguish between relatively concentrated photonics businesses and diversified companies where optical connectivity is only one earnings driver.

ETF exposure to the theme: Roundhill Photonics & Optics ETF

For investors assessing the sector but seeking exposure across several companies rather than a single stock, a new thematic ETF has also emerged.

The Roundhill Photonics & Optics ETF (LYTE) began trading on 6 August 2026. It is an actively managed ETF focused on companies involved in technologies including optical transceivers, lasers, silicon photonics, optical interconnects and related photonic infrastructure.

Its initial top five holdings are: Lumentum, Coherent, Eoptolink, Zhongji Innolight, and Ciena (Source: Roundhill Photonics & Optics ETF fund information and holdings, data as of 6 August 2026).

Our view: LYTE can be viewed as a global photonics basket rather than a pure US reshoring or China-substitution theme. Its exposure to both US and Chinese companies provides diversification across the optical supply chain, but also means geopolitical restrictions affecting Chinese suppliers could create risks within the portfolio itself.

Investor takeaway

The AI investment story is becoming broader than GPUs alone.

Memory, power, cooling and data-centre infrastructure have already become important parts of the AI value chain. Optical networking is increasingly another layer worth watching as larger AI systems require greater volumes of data to move quickly and efficiently.

But the risks remain significant. Hyperscaler capital expenditure could slow, customer concentration is high, technologies and standards are evolving quickly, US-China restrictions could disrupt supply chains, and high expectations can leave companies vulnerable when results disappoint.

But for investors looking beyond GPUs while still participating in the AI infrastructure buildout, optics is becoming increasingly difficult to ignore.