Bloomberg Tech→ original

Lumentum is sold out of optics for AI servers through early 2029

Lumentum Holdings, a maker of optical components for AI data centers backed by Nvidia, is "fully sold out" through early 2029. The company's CEO said so, according to Bloomberg. The order horizon runs more than two years from October 2026. Demand comes from tech companies that need faster networks inside their clusters. - Orders are booked through early 2029. - The shortage is shifting from accelerators to optics and lasers. - The company did not disclose volumes, prices or customers in this statement.

AI-processed from Bloomberg Tech; edited by Hamidun News
Lumentum is sold out of optics for AI servers through early 2029
Source: Bloomberg Tech. Collage: Hamidun News.
◐ Listen to article

Lumentum Holdings, a maker of optical components for AI data centers backed by Nvidia, said its products are "fully sold out" through early 2029. The statement comes from the company's CEO and was reported by Bloomberg. From October 2026 to early 2029 is more than two years, and the production plan for that entire period is already spoken for. The reason is simple: tech companies want faster networks inside their AI clusters and are willing to reserve capacity well in advance.

What exactly has been bought up

Lumentum makes lasers and photonic components. These parts convert electrical signals into light and carry data between servers, racks and halls of a data center. According to the company's public materials, its portfolio includes lasers for optical transceivers, laser sources for silicon photonics, and optical switches. Without these parts, accelerators sit in racks but cannot exchange data at the speed they need.

Bloomberg's short excerpt contains no breakdown by product. It is unknown which product lines are sold out completely and which only partly. There is also no data on revenue, prices or buyers. In such cases, "sold out" usually describes signed contracts and agreed volumes, not just money the company has already received.

Why optics is the bottleneck

When clusters grow to tens or hundreds of thousands of accelerators, the limit is no longer just a chip's compute power. It is also how fast data moves between chips, and how much is lost along the way. At high speeds, copper links work over ever shorter distances, so optics keeps moving deeper into the system: from connections between halls to connections inside the rack and closer to the switch itself.

Laser components scale more slowly than server assembly. Substrates made from specialty semiconductors, die yield and long testing cycles all matter here. This is our own assessment, not a quote from a report, but it explains why a backlog of more than two years is possible. New production lines do not come online in a quarter.

Nvidia's stake in Lumentum shows that the accelerator maker treats optics as part of its own supply chain. If optics runs short, so do finished clusters, no matter how many chips come off the line.

Key figures and facts

  • Horizon: orders are booked through early 2029, which is more than two years from the date of the report.
  • Source of the statement: Lumentum's CEO, as relayed by Bloomberg.
  • Buyers: tech companies building AI data centers.
  • Link to Nvidia: Bloomberg's headline describes the company as Nvidia-backed; the size of the investment is not given.
  • Not disclosed: volumes, prices, customer shares, plans to expand capacity.

Practical takeaways

For data center operators, the signal is clear: for projects starting in 2027–2028, optics should go into procurement plans now, not after the project is approved. Alternative suppliers and different interconnect formats reduce the risk but do not remove it entirely if the shortage hits the whole industry.

Investors should look for three things in Lumentum's upcoming reports: revenue growth, capacity investment and contract terms. These determine whether a full order book turns into profit or runs into production limits. For now, this is a statement from the company's executive, not verified figures.

For everyone else, the takeaway is simpler. Over the next two years, the pace of AI infrastructure buildout will depend not only on accelerator supply but also on how many lasers manufacturers can make.

ZK
Hamidun News
AI news without noise. Daily editorial selection from 50+ sources. A product by Zhemal Khamidun, Head of AI at Alpina Digital.
✦ Enterprise AI & Advisory

Executive AI Consulting, Custom Agents & Team Workshops

Work directly with Zhemal Khamidun at hamidun.com — enterprise AI architecture, production RAG & multi-agent systems, and executive training.

Book AI Consultation →Explore NomadNet AI
What do you think?
Loading comments…