Lumentum (LITE.US) orders are close to full through 2028! Can optical communication take over from the storage sector and become the next "super windfall" for AI computing power?

Optical module giant Lumentum (LITE.US) announced that orders may be filled until 2028, and NVIDIA Corporation has invested $2 billion to secure production capacity. Various signs indicate that optical components are replicating the surge logic of HBM and are set to become the next "storage-grade" windfall in AI computing infrastructure. Lumentum CEO Michael Hurlston said that demand for its optical components from top US tech companies is accelerating and is expected to fill its order book by 2028. Hurlston said in an interview in Tokyo on Friday, "The capital expenditures of US hyperscale cloud service providers are enormous and seemingly endless," "We are increasingly falling behind in meeting demand. It is expected that our production capacity will be fully booked by the end of 2028 within two quarters." Hurlston's forecast extends the company's previous disclosure that its production capacity was already sold out by the end of 2027. This highlights market expectations: even amid disruptions in the oil market and global economy due to the Iran war, demand for AI data center equipment will remain resilient. The company, together with its competitor Coherent, received a $2 billion investment from NVIDIA Corporation last month. Lumentum supplies advanced indium phosphide devices aimed at achieving high-speed data transmission. As optoelectronic technology appears more in cloud computing clusters, its stock on the Nasdaq has risen by more than 1500% over the past year. Optical company stocks have soared after years of decline To meet the surging demand, Lumentum has increased its production capacity at its core factory in downtown Tokyo by 12 times over the past two years. Currently, the company plans to invest at least $100 million in the factory area and adjacent facilities. The CEO said, "We may see the investment amount rising to $250 million," "You cannot imagine the pressure faced by the team to expand production. All US hyperscale cloud service providers want to know how many units we can produce this quarter, next quarter, and the quarter after, as this is truly the bottleneck of the entire industry." As advanced AI hardware requires faster transmission speeds, silicon photonics technology that uses light to transmit data is expected to replace copper-based interconnects in data center clusters. Optical networks also help systems reduce heating and energy consumption. Japan has companies with expertise in indium phosphide, fiber optic materials, and silicon photonics technology, including Fujikura, Sumitomo Electric, and Hamamatsu Photonics. Lumentum gained some technology from Japan through its acquisition of Oclaro Inc. in 2018, which had inherited optical research results from Hitachi. Today, its Sagamihara factory is one of the world's most advanced indium phosphide device production bases, providing stable and high-quality output. The capacity to increase production space at the Sagamihara factory and nearby Takao facilities is limited, so Lumentum is looking for a new location in Japan to expand production. The company's goal is to find an old electronic manufacturing facility that can be reused, which would shorten the time needed to bring new production capacity online. Hurlston said the company had considered purchasing such a site in Japan last year. The CEO pointed out that moving production out of Japan is not feasible because the manufacture of indium phosphide requires precise control of variables such as temperature, humidity, mixing speed, and water quality. A rich supplier network built around this material is also crucial, and the company has signed a seven-year agreement with a Japanese supplier described by Hurlston as "very critical." Despite investing heavily in factories, long-term supplier contracts, and hiring more engineers, Lumentum expects its backlog of orders to continue to grow. Hurlston said, "This momentum cannot last forever, it is not realistic, but for at least five years, this cycle seems quite sustainable," "When we talk about being sold out, we mean irreversible agreements. This is a big deal." Paradigm shift: From "computing-centric" to "connectivity-centric" To understand why optical components are so important, one must first look at the evolution of AI cluster sizes. Early AI tasks may have only required single-card or single-machine operations, but today's models with trillions of parameters require tens of thousands of GPUs working together. In this context, "on-site computing" is no longer the only challenge, "off-site communication" has become fatal. Traditional copper cable interconnects suffer greatly from high-speed transmission and produce astonishing amounts of heat. As computing clusters scale from thousands to hundreds of thousands of cards, optical communication is no longer an optional "accessory" but the only "artery." The role of optical components (lasers, optical chips, optical modules, etc.) is to convert electronic signals into light signals instantaneously and achieve lossless, low-latency transmission in a complex computing network. Just as HBM was developed to solve the "memory wall" problem, optical components are developed to break the "communication wall." When the computing speed of NVIDIA Corporation's chips increases several times every two years, if the connection bandwidth does not keep up, expensive GPUs will be in a "waiting" state. This elevation in strategic position is the core logic behind optical components becoming a windfall. If upgrading from HBM to HBM3e was the fuel for the windfall, then the evolution of optical modules from 800G to 1.6T is the "accelerator pump" for the optical sector. As AI chip power consumption exceeds 1000W, the traditional plug-and-play method for optical modules is facing thermal and efficiency limits. Silicon photonics technology can directly integrate the optoelectronic conversion link in chip packaging, which is not just product iteration but a revolution in underlying architecture. This allows optical component manufacturers to move from traditional "peripheral suppliers" to deeply integrated into the "semiconductor core circle." Giant companies' "physical options" Currently, US hyperscale cloud service providers are engaged in a "capacity arms race." Lumentum's orders are filled until 2028 and are all "irrevocable agreements." This scenario of "pay first, get the goods later, grab the cargo space" is reminiscent of when Micro and Samsung's HBM production capacity was secured in advance by NVIDIA Corporation. When a component shifts from being "on-demand purchased" to "strategic reserve," its valuation premium undergoes a qualitative change. One of the most notable recent developments is NVIDIA Corporation's strategic investment in Lumentum and Coherent. NVIDIA Corporation, as the "shovel seller" in the age of AI, has a strong signal regarding its moves in the supply chain. Through the investment, NVIDIA Corporation is essentially buying "insurance" for its GPU clusters. This action conveys a clear message: without light, there is no computing power. This endorsement from top downstream customers provides the optical components sector with a high level of "certainty." In the capital market, certainty is often more attractive than mere growth. Optical components have evolved from being an edge communication accessory to an essential underlying pillar in the AI computing empire. For investors and industry observers, if storage was the first wave of AI computing, optical components are likely to take over, dominating the next cycle of the "super windfall."