Ciena Says AI Demand Is Outrunning Supply as Its Backlog Tops 8.5 Billion Dollars and Heads Past 10 Billion

The AI buildout is no longer just a GPU story, and Ciena's latest quarter is the latest evidence. The optical networking equipment maker reported fiscal third-quarter revenue of 1.67 billion dollars, up 37 percent from 1.22 billion dollars a year earlier, with adjusted earnings per share of 2.11 dollars against a 1.73 dollar consensus. Adjusted gross margin expanded 450 basis points year over year to 46.4 percent. The results, first covered by trade outlets including Fierce Network and Light Reading and picked up by the Wall Street Journal in an interview published Saturday, show the bandwidth layer of AI infrastructure straining as hard as the compute layer.

The driver is AI traffic between and inside data centers. Cloud providers accounted for 53 percent of quarterly revenue, up 82 percent year over year, and chief executive Gary Smith said roughly half of Ciena's business now comes directly from hyperscalers rather than telecom carriers. "AI continues to drive compounding waves of network investment," Smith said in the earnings release.

The clearest signal is the backlog. Ciena ended the quarter with 8.5 billion dollars of orders still to ship, and Smith said on the earnings call that the company is "currently projecting to exit fiscal 2026 with over 10 billion dollars in backlog." Customers have extended their commitments through 2029, and the current backlog already stretches into fiscal 2028. Orders, according to chief strategy officer David Rothenstein, grew about 50 percent from 2024 to 2025 and another 50 percent from 2025 to 2026 — faster than supply can come online.

What holds shipments back is parts, not demand. Management said industry-wide component shortages are unlikely to ease before fiscal 2028, so Ciena has signed supply agreements locking in key components through 2029, and warned that operating cash flow will dip in the current quarter as those prepayments land. Pluggable optics revenue more than doubled year over year on the WaveLogic 6 Nano ramp, and revenue from co-packaged optics is expected to begin in 2027. The bottlenecks run through pump lasers, ceramic packaging and memory — the same components choking everyone else in the AI supply chain.

Guidance reflects both the boom and its constraints. Ciena raised its fiscal 2026 revenue outlook to 6.42 billion dollars, guided fiscal fourth-quarter revenue to 1.75 billion dollars plus or minus 50 million, and offered a preliminary fiscal 2027 view of at least 30 percent revenue growth.

The concentration risk is real: two customers each above ten percent of revenue together accounted for 41.7 percent of quarterly sales, so a pause by either would bite. That is the backdrop to Smith's Wall Street Journal interview, in which the 2001 telecom crash veteran described himself as "suitably paranoid, but optimistic" — arguing this cycle is grounded in real demand from a handful of investment-grade buyers, while insisting investment stay disciplined in case the wave crests.

For the broader AI trade, Ciena is the latest datapoint that the bottleneck is moving down the stack. The GPU roadmap gets the headlines, but the pipes between racks, buildings, and campuses are now booked years ahead, with supply-demand balance not expected before 2028. If the hyperscalers keep spending, the optical layer keeps the pricing power; if the AI capex cycle wobbles, the backlog length itself becomes the risk everyone watches.