Lumentum Says AI Data Center Optics Demand Could Outrun Supply Into Early 2029
Редакция ЦОД:Главное
Lumentum says demand for certain optical components used in AI data center infrastructure exceeds its available supply, with all of its production capacity already committed to customers. The company expects imbalances for some products to extend into 2027 and 2028, while the broader supply constraint could persist until early 2029.
- Lumentum says all of its available manufacturing capacity for the relevant optical components has been committed to customers.
- For certain unspecified products, the company indicated that about 70% of demand could remain unmet in 2027.
- A 30% supply deficit for other unspecified products could persist through 2028.
- The broader shortage is expected to continue until early 2029, but this is a forecast rather than a confirmed market outcome.
- New optical-component manufacturing capacity requires an estimated three to five years to establish.
- Possible US restrictions on Chinese optical transceivers remain unconfirmed but could further concentrate supply risk.
Lumentum has warned that supply of selected optical components for AI data center infrastructure remains materially below demand, and that capacity additions will not provide a near-term remedy. Speaking in Tokyo on the ninth, Chief Executive Michael Hurlston said the company cannot meet a significant portion of demand for certain products and that its available manufacturing capacity has already been sold to customers.
The company indicated that, for some unspecified products, about 70% of demand could remain unmet in 2027. For other products, a 30% supply shortfall could remain through 2028. The source cited by the analyst expects the overall supply-demand imbalance in optical components to continue until early 2029, although Lumentum has not provided a date for fully resolving the shortage.
The constraint is linked to intensive build-out of AI data centers and their higher compute requirements. High-speed optics are a critical part of the networking layer supporting AI clusters; sustained shortages could therefore delay equipment deliveries or limit the pace at which network infrastructure can be deployed and expanded.
Lumentum is increasing production, but bringing new manufacturing capacity online is a multi-year process. Establishing capacity for companies of this type takes roughly three to five years, according to the source. This long lead time limits the industry’s ability to rebalance supply quickly as AI infrastructure demand rises.
Potential US restrictions on new Chinese optical transceivers could add to supply-chain risk if introduced. The measure remains discussed rather than confirmed. Counterpoint Research has said Lumentum and Coherent may be unable to cover the supply gap that could result from the loss of Chinese producers, including Innolight. If such restrictions were implemented, they could increase supply concentration among the remaining vendors.
For data center operators, hyperscalers and AI infrastructure integrators, the development points to a potential bottleneck outside the usual constraints of power, site delivery and accelerator availability: the availability of high-speed optical interconnects required to deploy and scale network fabrics.
High-speed optical components are critical to AI data center networking. A prolonged supply shortfall could constrain deliveries and scale-out of network infrastructure for new AI clusters even where demand for compute systems and data center capacity remains strong.
- TechNews 科技新報 / КитайскийПервоисточник →
Материал подготовлен редакцией ЦОД: Главное на основе указанных первоисточников.