AAOI Is Losing Sales to Capacity, Not Demand

AAOI reported record Q2 revenue of $191.9M, up 86% year over year. The more important line in the release was not the growth rate, but the reason the company cannot sell more today. Management said demand for 800G and 1.6T products is strong, while factory capacity is limiting current shipments. AAOI is not waiting for customers to place orders. It is shipping as many units as its plants can make.

The factory is the near-term limit

The original post puts current monthly capacity near 200,000 units. Management's targets are above 650,000 units per month for 800G and 1.6T combined by the end of 2026, and above 930,000 by the end of 2027. Those are not demand estimates. They are the volume targets needed to turn orders into revenue. That changes how the next quarter should be read. Revenue can rise while demand still outruns output. If shipments and production lines expand together, the constraint can become operating leverage. If they do not, a strong order book will remain a promise rather than reported sales.

What the numbers add

The cross-checked earnings transcript adds that data-center revenue was $107.7M, or 56% of total revenue, while 100G weakness was linked to a customer's switch supply issue. The original demand-versus-output picture therefore has a second layer: product-generation mix and customer procurement can move in the opposite direction.

CPO changes the clock, not the bottleneck

AAOI plans to ramp ELSFP for CPO toward roughly 400,000 units per month in 2028. It also expects an InP-related facility to come online in early 2027. This is not a claim that CPO will immediately replace pluggable optical transceivers. It is a claim that both paths can grow while AAOI works through laser and transceiver availability. In-house laser manufacturing can protect AAOI from shortages that affect other suppliers. It does not make the revenue automatic. Hyperscaler qualification for 1.6T products, yield, and actual shipments still have to arrive in sequence. The near-term test is whether the qualification timeline described as a matter of weeks becomes customer volume.

What the next earnings print must show

The score is not whether management repeats that demand is strong. It is whether the above-650,000 monthly target for the end of 2026 and above-930,000 target for the end of 2027 appear in shipments and revenue. The ELSFP ramp toward 2028 is a separate checkpoint. If the targets hold and shipments rise with data-center revenue, the production bottleneck was a growth lever. If the targets are cut or qualification and expansion slip, strong orders alone will not explain the results.