Ciena (CIEN.US) Performance Strengthens AI Optical Interconnection Theme, Q2 Profit Surges 290%

According to Financial Associated Press, Ciena (CIEN.US), which focuses on high-speed interconnection DCI between different data centers and high-speed optical interconnection within AI data centers, announced quarterly results and fiscal year revenue outlook that far exceeded market expectations. This highlights how the wave of global AI data center construction, led by major US cloud computing service providers, has driven demand for high-speed optical interconnection—demand that has already evolved from a “long-term growth narrative” into a trajectory of concurrent acceleration in AI optical interconnect/optical communications infrastructure revenue, gross margin, and adjusted EPS.

Shortly before Ciena announced its strong earnings, Nvidia CEO Jensen Huang, also known as the “Godfather of AI,” shared the stage with Marvell CEO at the 2026 Computex conference in Taipei, China. He stated that Marvell Technology (Marvell) has the potential to become the next publicly traded company with a market capitalization exceeding $1 trillion, emphasizing Marvell’s critical role in large-scale AI data center optical interconnection infrastructure systems. This endorsement from the most dominant leader in the AI computing power field has greatly strengthened market confidence in Marvell’s long-term growth trajectory.

Matt Murphy, CEO of Marvell Technology, delivered a keynote speech entitled "The Future of AI Scaling Depends on Connectivity" at the 2026 Computex conference, specifically underscoring that with the exponential increase in AI large model-related compute resource demand and massive AI infrastructure clusters, improvements in computational performance have begun to outpace data movement and high-speed DCI interconnection between and within data centers. Thus, the next bottleneck for AI compute scaling will be governed by “connectivity, not single-node performance,” and leading companies in the AI optical interconnect/optical communications infrastructure chain such as Ciena, who “stand in the light” alongside Marvell, are unarguably positioned at the heart of this new infrastructure lifeline chain with AI data center optical interconnection, ultra-high-speed DSP, 1.6Tb/s coherent optical transport, and silicon photonics technology stack.

Ciena’s revenue soars 40%+ and raises guidance for both the quarter and full fiscal year

For the second quarter of fiscal 2026, ending May 2, 2026, Ciena reported total revenue of $1.5707 billion—a substantial YoY increase of 39.5%, significantly higher than $1.1259 billion in the same period of 2025; adjusted EPS was $1.64, up 290% from $0.42 a year earlier. Both of these core metrics beat market expectations, with FactSet consensus showing EPS of $1.46 and revenue of about $1.51 billion, so Ciena actually exceeded expectations by $0.18 per share and about $60 million in revenue. According to Seeking Alpha, revenue topped expectations by about $70 million—both sets of data pointing in the same direction, highlighting revenue and profits that are significantly stronger than the recently upwardly revised forecasts of Wall Street analysts.

Earnings quality also improved significantly. Ciena’s GAAP gross margin for the second quarter rose from 40.2% a year ago to 44.0%, while adjusted gross margin rose from 41.0% to 44.9%; GAAP operating margin increased significantly from 2.9% to 15.1%, and adjusted operating margin rose from 8.2% to 19.5%. GAAP diluted EPS jumped from $0.06 to $1.49, while adjusted EBITDA soared from $116.7 million to $341.8 million. This shows the company is not simply relying on a “data center order boom” to drive revenue, but has entered a phase of strong operating leverage release: the combination of optical interconnectivity + data center DCI + network hardware, along with platform software, product after-sales service mix, and client concentration, has jointly pushed up profit elasticity.

Examining the business structure, Ciena’s growth is highly oriented towards the AI data center high-speed interconnection cycle. In Q2, Networking Platforms revenue was $1.274 billion, accounting for 81.1% of total revenue; among this, Optical Networking, closely tied to data center optical interconnect, hit $1.0998 billion—70.0% of total revenue, and significantly higher YoY from $773.6 million a year ago; Routing and Switching contributed $174.2 million, also up YoY from $92.7 million. Two customers each accounting for over 10% generated a combined 34% of revenue, showing that hyperscale cloud customers and major AI data center projects remain the main driver for this round of growth, but also suggesting customer concentration is a risk variable worth monitoring for valuation.

For the results outlook that investors are focused on, the company raised its outlook again, further confirming that AI optical interconnect demand is not a one-off expansion. Ciena projects Q3 fiscal revenue at $1.625 billion, ±$50 million, higher than Wall Street’s consensus view of $1.5 billion, with an adjusted gross margin near 45%, ±50 basis points, and an adjusted operating margin of 19%–20%; the company also lifted full-year fiscal 2026 revenue guidance to $6.3 billion ± $100 million, with the midpoint implying a YoY increase of about 32%, far exceeding analyst expectations of about $6.08 billion, and foresees an adjusted gross margin of 44.5%–45%, and adjusted operating margin of around 19% ± 50 basis points. This is a clear upshift from the $5.9–$6.3 billion range given in March, reflecting greater management confidence in data center order visibility, delivery pace, and overall supply-demand environment.

Ciena’s earnings report also noted management emphasized that AI data center expansion is putting high-speed connectivity at the bottleneck position in infrastructure, placing special focus on how cloud computing vendors are building data centers at scale for AI, driving perpetually increasing demand for “high-speed optical connection within, around, and between data centers.”

From a mid- to long-term perspective, as AI compute infrastructure becomes increasingly clustered, distributed, and cross-data-center coordinated, optical interconnection upgrades from being a mere "communications equipment cycle product" to a hard infrastructure bottleneck asset for AI compute.

It’s noteworthy that while Ciena, Marvell, Lumentum, and Coherent all belong to the AI optical interconnect/optical communications infrastructure supply chain, they occupy different segments. Marvell is more focused on base-level semiconductors: optical DSP, coherent-lite DSP, PAM4 DSP, switching chips, SerDes, custom AI ASIC, and silicon photonics platforms; Lumentum and Coherent are more into optical components/optical modules/lasers/transceivers; Ciena, by contrast, is more platform-centric for complete data center optical systems and coherent optical transmission, with core competencies in WaveLogic coherent optical engines, 800G/1.6T coherent transport, optical networking equipment, routing/switching, network automation, and integrated operator/cloud services networking solutions, rather than just selling optical chips or key lasers.

So, Ciena is more of a “system-level AI optical interconnect and DCI” player, with a traditional strength in high-speed interconnection DCI between data centers, metro/long-haul/operator backbones, and WAN connections for cloud providers—connecting multiple data centers, campuses, regional clouds, and telecom networks with high-speed coherent optical systems. However, Ciena is also expanding from “DCI between data centers” to “high-speed optical interconnect within AI data centers”. Its acquisition of Nubis is meant to expand its inside-the-data-center strategy to handle AI workloads, and industry trends are pushing optoelectronic chains from between rooms, buildings, and campuses ever closer to racks, switching chips, and even advanced packaging systems.

1.6T DCI, CPO, and high-speed optical interconnect within data centers ignite a new wave of investment in the compute infrastructure chain

Within the AI optical interconnect infrastructure supply chain, Ciena is best positioned as a “system-level beneficiary” in WAN + inter-data center DCI + high-speed internal and peripheral interconnection, while Marvell is more the “chip/module/switching chip/optical DSP/ASIC/silicon photonics” semiconductor layer. Ciena has also made it clear that its long-term strategy is both WAN and high-speed internal/external data center connectivity, and through Nubis, it is moving nearer to the AI cluster interior step-in CPO/NPO and high-speed copper interconnect: Nubis technology includes CPO/NPO optical modules with up to 6.4Tb/s full-duplex bandwidth, as well as active copper cable solutions supporting 200Gb/s per channel and up to 4 meters, aimed directly at solving the energy, density, and latency constraints in AI cluster scale-up and scale-out.

Ciena’s strong earnings and the recent sharp revaluation of Marvell in the markets are driven by the same logic: the bottleneck in AI infrastructure is shifting from just GPU compute expansion to connectivity, switching, optical modules, DSP, SerDes, CPO, DCI, and network power consumption. As AI training and agent workloads drive GPU clusters from single cabinets to multi-cabinets, multiple data centers, and even across regions as full AI factories, data traffic will explode across a three-tiered structure—chips/cabinet interiors need low-power high-speed interconnect, data center interiors need high-radix switches and optoelectronics synergy, and inter-data center networks require 1.6T-level coherent-lite, DCI interconnections, long-haul coherent optics transmission, and even CPO as part of the AI optical interconnect infrastructure stack.

Morgan Stanley’s latest forecast sharply illustrates that as AI agent adoption booms globally, AI infrastructure investment is pivoting from a “single-point computing competition centered on AI GPU” to “AI agent-driven full-stack compute systems,” and the supply chain bottleneck is already moving from “large-scale GPU/ASIC purchases” to “aiming to solve the entire AI data center delivery flow chain—covering data center power infrastructure, liquid cooling, data center CPU, DRAM/NAND/HBM, optical comms/interconnect, high-performance Ethernet or DCI high-speed interconnect, transformer, gas turbines, etc.—with storage and optical interconnect now representing the biggest bottlenecks for compute systems in data centers.

Jensen Huang’s judgment on the development trend of AI compute power is also seen as an important foundation for the market’s growing optimism about Marvell and other leading AI optical interconnect players. He suggested that AI commercialization has entered the era of “AI agent-dominated Useful AI,” meaning Token generation now drives not just research and experimentation, but brings real profitability. As AI workloads are decomposed and distributed, the need for “massive interconnectivity” has surpassed the possibility of mere single-node upgrades—system-level scale-out has become central to design philosophy.

In subsequent keynotes, Nvidia founder and CEO Jensen Huang’s conversation with Marvell Technology’s CEO profoundly expounded on how the entire AI compute supply chain is upgrading to “connectivity” as the core bottleneck, and revealed why Huang is so bullish on Marvell—choosing to pursue deep cooperation and even executing a $2 billion strategic investment in the company.