Home

About Us

Newsroom

Blog

With close to 18,000 registered attendees and more than 700 exhibiting companies, OFC 2026 was the most widely attended in recent history. More than any single attendance metric, however, OFC 2026 proved particularly exciting — at least in my view — for the breadth and depth with which it covered the emerging photonics ecosystem optimized for AI/ML infrastructure. It also established clearly that the transition to this new ecosystem is now underway.&nbsp;Expanding Optical Connectivity: From Scale-Out to Scale-Up&nbsp;In many respects, the excitement around OFC began the week before the conference, with several important announcements. Among these, the OCI MSA by Meta, Microsoft, OpenAI, and others was probably the most significant. It defined a new generation of 50G NRZ per wavelength interconnects optimized for scale-up clusters, an entirely new optical use case that is currently served exclusively by copper. Optical scale-up is set to expand optical bandwidth from GPU by nearly an order of magnitude, beyond its traditional role in scale-out and scale-across architectures.&nbsp;<img src="/sites/default/files/inline-images/OCI-MSA%20and%20XPO.jpg" data-entity-uuid="655f63f2-d966-4207-b2bb-36ddd8fc0577" data-entity-type="file" alt="OCI-MSA logo and XPO module image" width="701" height="156" loading="lazy">Image1: OCI-MSA | Image 2: XPO moduleIn addition to the OCI MSA, several other optical technologies were introduced as promising solutions for scale-up, including a slow-and-wide interconnect system based on widely available 3D-sensing VCSELs. At the same time, there were important announcements on the scale-out side, including the first demonstrations of multiple optical channels operating at 400G PAM4 in a pluggable form factor, as well as additional momentum behind 200G PAM4 CPO and NPO initiatives such as the Open CPX MSA. High-capacity pluggable modules like the 12.8T XPO also drew attention — a capacity that, just a few years ago, would have been an entire Tomahawk 3 data center switch.&nbsp;Challenges and Path to Deployment&nbsp;Alongside the enthusiasm for these new optical technologies and their surrounding ecosystems, discussions at OFC also highlighted several system- and network-level challenges that must be addressed before widespread deployment can occur. These topics were explored extensively in multiple forums, including an OCP panel (organized by Vladimir Kozlov from LightCounting) and an IEEE panel (which I organized), both held on the same day. As several speakers from major hyperscalers emphasized, total cost of ownership (TCO) will ultimately determine the pace of adoption for this new ecosystem. These system considerations are becoming even more critical with the upcoming transition to next-generation SerDes, nominally referred to as 400G. A range of compelling technology directions is under consideration in this transition, many of which will undoubtedly become topics of our future discussions.&nbsp;Finally, another critical requirement repeatedly emphasized in OFC discussions was the need to scale manufacturing capacity for optical technologies to meet the anticipated increase in demand as the industry shifts from copper to optical scale-up interconnects over the next two to three years.&nbsp;In this sense, even for those of us who experienced the highly dynamic OFC conferences that explored the growing role of optical interconnects during the early internet era 25 years ago — and later in cloud and DCI infrastructure 10-15 years ago — this was an even more exciting moment to be part of OFC.&nbsp;To explore how the trends discussed above translate into real technology, read the next blog in this series: <a href="https://www.lumentum.com/en/blog/enabling-next-phase-ai-optical-infrastructure">Enabling the Next Phase of AI Optical Infrastructure: From Industry Shifts to Technology Implementation</a>.

OFC 2026 Takeaways, Part 1 – Event and Market Trend Overview

The New Photonics Ecosystem for the AI/ML Infrastructure 

AI infrastructure growth follows a reinforcing cycle. Larger models drive the need for more hardware. That hardware demand accelerates technology innovation, particularly in bandwidth density and interconnect efficiency. Innovation improves power efficiency and economics, enabling even larger models to be deployed, which then require more hardware.&nbsp;As this cycle has been running, the inherent limitations of electrical connections are becoming bottlenecks for bandwidth, power, and reach. Optical interconnects are becoming increasingly important to enable future scaling and a continued AI innovation cycling.&nbsp;From Pluggables to Co-Packaged OpticsNot only is a transition occurring from copper-based connections to optic based, but a transition is also occurring within optics architectures. Optical connectivity, for many of the same reasons as the fundamental shift from copper to optic, is transitioning from pluggable solutions towards co-packaged optics (CPO).&nbsp;Traditional pluggable optics place optical engines at the faceplate. While modular and serviceable, they rely on long electrical traces, consuming significant power and limiting bandwidth density due to copper reach constraints.Near-package optics represent an intermediate step, moving optical engines closer to the ASIC to reduce electrical losses while maintaining relatively manageable integration complexity.CPO takes the next leap by integrating optics directly within the package. This approach minimizes electrical I/O, achieves the lowest energy per bit, and enables extreme bandwidth density. However, these benefits come at the price of a significant increase in packaging and integration complexity. For co-packaged solutions, both advanced packaging techniques and heterogenous integration become essential.&nbsp;Heterogeneous integration, at its most basic is the integration of separately manufactured components into a single system, package or even die. These components may have different functions (computer, memory), process nodes, or even material systems (Si, InP) - but they are all brought together to make a single efficient device. In short - the final component is greater than the sum of its parts. Heterogeneous integration makes it possible by leveraging advanced packaging techniques. In the context of CPO, heterogeneous integration is bringing optic components (lasers, PICs, photodetectors, etc.) together with compute components.Why Advanced Packaging Matters Beyond CPOAlthough enabling CPO is a major motivation, heterogeneous integration and advanced packaging have powerful benefits that extend far beyond any single architecture.First, heterogeneous integration enables alternate scaling pathways, similar in spirit to Moore’s Law extensions in the semiconductor industry. As an example, if fab yield concerns make a large die not economically viable, these technologies can ‘stitch’ smaller chiplets and result with an equivalent compute power. Specifically in the regime of optical networking, these techniques can enable optical bandwidth to scale generation over generation without relying solely on device-level improvements.Second, it allows best-process integration. Different photonic technologies excel in different roles—III-V materials for light generation and modulation, and silicon photonics for routing and integration. Advanced packaging makes it possible to combine these technologies into a single, optimized solution.Finally, advanced packaging supports faster innovation cycles. Chiplet-based approaches enable IP reuse, reducing design timelines and validation timelines. This results in both a faster development cycle but also a lower risk one, which is increasingly important as upgrade cycles compress.&nbsp;Advanced Packaging Is Already in ProductionThese concepts are not theoretical. Advanced packaging is already being deployed at scale in both the wider industry and here at Lumentum.One example is Lumentum’s stacked VCSEL-on-driver technology, which leverages 2.5D integration techniques to improve performance, footprint, and manufacturability. VCSELs themselves are the highest volume optical source technology in history, with decades of deployment across enterprise data centers, campus networks, and consumer sensing applications.A second example is <a href="https://investor.lumentum.com/financial-news-releases/news-details/2026/Lumentum-Showcases-Breakthrough-Optical-Scale-Up-Demonstration-at-OFC-2026-Using-VCSEL-Technology/default.aspx">a recent technology demonstration by Lumentum at OFC 2026</a> – a fan out wafer level packaging architecture was used to integrate Lumentum VCSEL and photodiodes with commercially available TIA and driver chips. This integrated device – combining photonic chips with electronic chips in a unified package provides a unique solution for scale-up applications.<img src="/sites/default/files/inline-images/OFC%202026%20-%20VCSEL%20demo.jpg" data-entity-uuid="8f7e6988-f9d3-4917-9388-4c5b723dae88" data-entity-type="file" alt="VCSEL-based scale-up interconnect demo at OFC 2026" width="609" height="417" loading="lazy"> Image 1: VCSEL-based scale-up interconnect demo at OFC 2026<img src="/sites/default/files/inline-images/vortix%20on%20quarter.jpg" data-entity-uuid="be917628-06aa-4888-bd89-6f58ef4374eb" data-entity-type="file" alt="10 Tbps VCSEL+PD optical engine prototype based on 3D FOWLP" width="319" height="257" loading="lazy"> Image 2: 10 Tbps VCSEL+PD optical engine prototype based on 3D FOWLPIn parallel, multiple programs are applying advanced packaging and heterogeneous integration techniques to next-generation photonic platforms, reinforcing that this shift is already well underway.Access additional information on CPO technology on our&nbsp;<a href="https://www.lumentum.com/en/search?q=cpo">website</a>.&nbsp;

Advanced Packaging and Heterogeneous Integration: Reshaping Photonic Architectures for AI

In the high-stakes world of artificial intelligence and cloud computing, where milliseconds matter and petabytes of data flow continuously, there's a critical technology that rarely makes headlines: the Electro-absorption Modulated Laser, or EML.&nbsp;Like arteries pumping lifeblood through the human body, EMLs drive optical pathways, moving information at the speed of light between compute nodes, storage systems, and networks. Without optical connectivity, every GPU cluster would sit idle and isolated, every training run would stall, and the AI revolution would grind to a halt. In AI data centers, where speed is everything, these are the unsung heroes making the impossible possible.Understanding EML TechnologyAn Electro-absorption Modulated Laser (EML) is a photonic integrated circuit that integrates two key components on a single indium phosphide chip: a Distributed Feedback Laser (DFB) and an Electro-Absorption Modulator (EAM). The DFB generates a stable, wavelength-stabilized optical carrier signal. The EAM then rapidly changes its absorption properties in response to electrical data signals, modulating this continuous-wave light at extraordinary speeds. Modern EMLs routinely support 100 Gbps per lane, with next-generation 200 Gbps per lane designs already in production.<img src="/sites/default/files/inline-images/Electro-absorption%20Modulated%20Laser%20EML.jpg" data-entity-uuid="68393beb-3065-4075-acac-b81d250c9f78" data-entity-type="file" alt="electro absorption" width="250" height="141" loading="lazy">The elegance of EML technology lies in its combination of low chirp (minimal frequency variation enabling longer transmission distances), high extinction ratio (clear distinction between signal states), superior signal fidelity, and energy efficiency. EMLs support advanced modulation formats like four-level Pulse Amplitude Modulation (PAM4), which transmits two bits per symbol, effectively doubling the data rate over the signaling rate. The integration of laser and modulator on a single chip reduces packaging complexity, improves thermal behavior, and enhances reliability.&nbsp;The Scale of the Challenge: The EML Demand EquationUnderstanding the technology is one thing, but grasping the scale at which these components must operate is quite another. The global AI data center market, valued at approximately $15 to $17 billion in 2024, is projected to surge to $60 to $165 billion by 20301 with hyperscalers like Microsoft, Amazon, Google, and Meta collectively investing nearly $200 billion in 20242 alone. To put the infrastructure challenge in perspective: AI-driven data center power capacity in the United States is projected to grow from 30 gigawatts in 2025 to over 90 gigawatts by 2030, exceeding California's entire power demand today3. Training workloads now demand power densities of 100 to 200+ kilowatts per rack, with new systems pushing toward one megawatt per rack3.This massive power density is mostly due to an equally massive data throughput requirement, and that's precisely where EMLs become indispensable. Most high-speed optical transceivers, the physical interfaces converting electrical data into optical signals and back again, rely on EML laser transmitters for high-speed applications. In a 400G FR4 transceiver, four 100G EMLs operate in parallel, each handling one wavelength in a standardized Coarse Wavelength Division Multiplexing (CWDM4) grid.&nbsp;The mathematics of scale are clear. Consider a single 51.2 Tbps switch with 64 ports of 800G. Each 800G port can use an optical transceiver containing eight individual 100G EML chips (since 8 x 100G = 800G). Multiply that across all 64 ports, and you get 512 EML chips in just one switch (64 ports x 8 EMLs per port = 512 chips). In a facility with 100 such switches, you need over 51,000 EML chips. Hyperscale AI data centers deploying thousands of switches require millions of EMLs. Industry projections indicate that global demand for 800G+ optical transceiver modules will reach 24 million units in 2025 and surge to nearly 63 million units by 20264, translating to approximately 200 to 500 million EML chips annually.The speed evolution tells a compelling story. The 400G era, deployed around 2020 to 2022, established 100G PAM4 per lane as the standard. The 800G generation doubles throughput through either an 8x100G architecture or a 4x200G architecture using next-generation 200G EMLs, with shipments surging in 2024 and expected to double again in 20254,. The transition to 1.6T is accelerating faster than anticipated5 according to industry analyst LightCounting.Each speed jump demands more from EML technology: higher bandwidth, lower noise, better extinction ratios, tighter thermal management, and enhanced reliability under stress. Without them, scale-out architectures connecting thousands of GPUs would collapse, data center interconnects would fail, storage networking would bottleneck, and the entire AI infrastructure performance can suffer. In essence, EMLs are the physical embodiment of data center arteries, and if they fail or cannot meet demand, the entire system would be considerably constrained.When Demand Surge Meets Quality&nbsp;As data centers transition from 400G to 800G and 1.6T, the margin for error in laser manufacturing vanishes. In this regime, scaling is not merely a matter of increasing volume; it is a matter of maintaining manufacturing excellence at the limits of physics. The "cost of quality" in EML production is exponential. Because these devices are deployed by the millions in synchronized GPU clusters, a single failing EML can cause a "cascading stall." If one laser in an 800G transceiver fails during a multi-week training run involving 50,000 GPUs, the entire computation may need to roll back to the last checkpoint, costing hundreds of thousands of dollars in wasted compute time.The Reliability Challenges of Advanced Photonics:<ul><li data-list-item-id="e40ac8c1a4e96abc78321aa9b4ff2f458">Epitaxial Precision: EMLs require multi-stage epitaxial material growth, where different semiconductor materials are joined with nanometer precision. Any impurity or crystal defect can shorten lifetime.</li><li data-list-item-id="e029f270800f20eb79b34c6e9f8f1e44d">Thermal Management at Scale: At 200G per lane, the heat density on the InP chip increases significantly. Control is critical to prevent wavelength shift and extinction-ratio degradation.</li><li data-list-item-id="ef52bf77dc20d6f0a001bb97f87aa4ad0">Signal Integrity (PAM4): Unlike traditional, two-level "on-off" signals, PAM4 uses four distinct voltage levels to encode data. This requires the EML to be incredibly linear. Even minor manufacturing variations, tolerable at lower speeds, create "noise" that prevents the DSP (Digital Signal Processor) from recovering data at 1.6T speeds.</li><li data-list-item-id="e4095cd2ba7a9d6614238a5b9b1b886bf">ESD Sensitivity: Integrated EMLs are more ESD sensitive than discrete components. Robust cleanroom protocols and packaging reliability are non-negotiable.</li></ul>In the AI era, quality is no longer a premium feature; it is an existential requirement.&nbsp;Cisco Systems captures this reality as follows: “each hour of [GPU] downtime due to optics failures could cause a loss of from $2,000 to $12,0006. In other words, quality is not a nice-to-have spec sheet parameter - it is a budget line item.Lumentum's Edge - Scaling with QualityWhen quality becomes existential and volumes reach millions of units annually, the differentiator is straightforward: who can scale optical lanes without trading reliability for volume. Lumentum's edge comes from high-volume indium phosphide-based chip manufacturing, complete vertical integration, and a multi-site footprint designed to support volume ramps while protecting field reliability.Unlike companies who outsource epitaxial growth, Lumentum controls the entire lifecycle from initial wafer fabrication through final chip packaging. The company operates wafer fabs in San Jose, CA (USA), Caswell (UK), and Sagamihara and Takao (Japan), with flagship assembly and test in Thailand, providing geographic diversification that ensures supply continuity during periods where 800G and 1.6T components face projected supply gaps approaching 40-60%7.Lumentum’s experience validates and enables execution at scale.&nbsp;<ul><li data-list-item-id="ec5f452e0c963de1b230d3522b1647c1b">Over 125 million optical components shipped in the last five years.&nbsp;</li><li data-list-item-id="e86b9c8ee833e0a3fbdb0e7b75b4666b4">Number one quality score for EMLs from top-tier customers for two consecutive years.&nbsp;</li><li data-list-item-id="e3d85e9a66f9ef22fd5678cafe0b27a78">Lumentum is currently shipping 200 Gbps per lane EMLs at scale, the foundational component enabling the latest 800G &amp; 1.6T transceivers.&nbsp;</li><li data-list-item-id="e08bf41ac85a7a7665036119665f6b16b">Recent quarters saw record EML revenue, with production capacity expanding 40%8 year over year to meet hyperscaler demand.</li><li data-list-item-id="e169a806da784d524fbf9d5cd6cc072a8">The GaAs platform alone has shipped over 2 billion VCSEL arrays with negligible RMAs, a level of manufacturing maturity now applied to complex InP EMLs powering AI clusters.</li></ul>Furthermore, stress testing in our Telcordia GR-468 qualification validates field reliability. Industry awards from Lightwave Innovation Reviews9 and ECOC recognize innovation10, but our deeper validation comes from deployment across data center networks globally.&nbsp;In a market where bandwidth scaling and uptime are inseparable, the advantage isn't merely having a fast EML. It's having a repeatable, customer-audited, high-volume InP manufacturing operation that can ship at scale while holding the reliability line that hyperscale AI demands.ConclusionIn the high-stakes race for AI supremacy, speed may be the goal, but quality and reliability are an absolute requirement. As the data center "arteries" face unprecedented pressure from the transition to 800G and 1.6T and a global supply crunch, there is no room for a weak link.&nbsp;Lumentum doesn’t just build EMLs; it builds certainty. By pairing our telecom-grade pedigree with in-house manufacturing, we ensure that these unsung heroes perform flawlessly under stress. When network performance depends on microns of epitaxial material, it matters who builds it - Lumentum can be a partner that meets industry scale without compromising the quality and consistency your links can rely on.Explore our <a href="https://www.lumentum.com/en/products/data-center/modulated-lasers/emls">EML solutions</a> moving AI data at light speed.<h6>References</h6><h6>1 - Precedence Research. (2025). "AI Data Centers Market Size Expected to Reach USD 165.73 Billion by 2034."&nbsp;</h6><h6>2 - MarketsandMarkets. (2025). "AI Data Center Market Size, Share and Industry Report, 2025 To 2030."&nbsp;</h6><h6>3 - McKinsey &amp; Company. (2025). "The next big shifts in AI workloads and hyperscaler strategies."</h6><h6>4 - LightCounting September 2025 Ethernet Optics Report</h6><h6>5 - LightCounting Q4 2025 Market Report.</h6><h6>6 - https://www.cisco.com/c/en/us/products/collateral/interfaces-modules/transceiver-modules/critical-high-quality-ai-networks-wp.html</h6><h6>7 - https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/opportunities-in-networking-optics-boosting-supply-for-data-centers</h6><h6>8&nbsp;- https://www.semiconductor-today.com/news_items/2025/mar/lumentum-030325.shtml</h6><h6>9&nbsp;- https://investor.lumentum.com/financial-news-releases/news-details/2021/Lumentum-Honored-by-2021-Lightwave-Innovation-Reviews-2021-3-24/default.aspx</h6><h6>10&nbsp;- https://investor.lumentum.com/financial-news-releases/news-details/2022/Lumentum-Wins-ECOC-2022-Data-Center-InnovationBest-Product-Award-for-its-High-Performance-200G-PAM4-EMLs/default.aspx</h6>

How microscopic semiconductor devices move AI data at light speed through modern data centers

EMLs: The Unsung Heroes Powering AI&#039;s Digital Arteries

AI data centers are scaling at an unprecedented pace, but the constraints shaping that growth are changing. While GPUs and accelerators continue to advance rapidly, power delivery, power efficiency, and heat dissipation increasingly define how far and how fast AI clusters can scale.In absolute terms, compute still dominates total system power. Network infrastructure typically represents on the order of 10–20% of an AI system’s power budget. However, as clusters grow from tens to hundreds of thousands of accelerators, interconnect power and efficiency become disproportionately important to system architecture and scalability.At large scale, the challenge shifts from how fast we can compute to how efficiently we can move data. Electrical I/O losses, high-speed SERDES, retimers, and long copper traces all consume significant power, and—critically—their power consumption does not scale linearly with bandwidth. As lane speeds move from 100G to 200G and aggregate bandwidth pushes toward 1.6T and beyond, these effects compound.The result is that data movement increasingly constrains system design, not because it consumes the most power overall, but because it concentrates power and heat in ways that limit reach, density, and topology flexibility.Optics are becoming a power-management tool, not just a bandwidth tool. Traditional electrical scaling starts to break down at very high speeds. Pushing more data over copper requires higher voltage swings, aggressive equalization, and additional signal conditioning—all of which translate directly into power and heat.Optical interconnects change that equation. By shifting high-speed data transmission into the optical domain earlier in the signal path, optics dramatically reduce electrical losses. That’s why optics are no longer just about reach—they’re now one of the most effective ways to lower system-level power per bit.The key idea behind many emerging connectivity architectures is simple: shorten electrical paths, simplify DSP processing, and let optics do more of the work.<h3>Linear Optics: Cutting DSP Power Where It Counts</h3>One of the most practical ways to reduce power today is through linear optical architectures. Transmit/Linear Receive Optics (TRO), a specific implementation of linear receive architectures, keep DSP processing on the transmit side but simplify or eliminate DSP functionality on the receive side—when channel conditions allow (see Figure 1 below).<img src="/sites/default/files/inline-images/blog-compute-chrisitan-figure1v2.jpg" data-entity-uuid="889cc178-98eb-4ef5-a6e0-91434c255ede" data-entity-type="file" alt="Comparison of FRO, TRO, and LPO transceiver module architectures" width="550" height="634" loading="lazy">Figure 1 – Comparison of FRO, TRO, and LPO transceiver modules architecturesFor short, well-controlled links inside AI clusters, this approach works extremely well. Connections between NICs and first-stage switches are a good example. In these cases, removing one of the two DSPs found in Fully Retimed Optics (FRO) designs can cut optical module power by roughly 30–40%, depending on reach and channel conditions, while also reducing latency.Multiply that savings across thousands of links, and the impact becomes very real at the data center level.Linear Pluggable Optics (LPO) take power reduction a step further by removing DSPs from the optical module entirely. Instead, all signal processing is handled by the host ASIC or NIC, allowing the module itself to operate as a largely analog device.In short-reach environments with carefully managed electrical channels, LPO modules can reduce optical module power by more than 50% compared to traditional DSP-based designs. They also lower latency and simplify thermal design.The tradeoff is that LPO demands tighter system-level co-design. Host electronics, board layout, and signal integrity all need to be dialed in. For hyperscalers, that’s often a reasonable trade-off.<h3>Moving Optics Closer to the ASIC</h3>As data rates climb past 1.6T, even optimized pluggable optics start to hit limits. That’s where Near-Packaged Optics (NPO) and Co-Packaged Optics (CPO) come in.Both approaches reduce power by shortening the electrical distance between the switch ASIC and the optical engine. NPO places optics on the same board as the ASIC, while CPO integrates them directly into the ASIC package. The closer the optics, the less electrical loss—and the lower the power.CPO offers the biggest potential gains, but it also introduces new challenges around heat and serviceability. That’s where architecture matters as much as integration.One of the biggest challenges with CPO is putting lasers right next to the hottest silicon in the system. External Laser Source (ELS) architectures solve this by separating the laser from the optical engine. High-power continuous-wave lasers are placed at the system faceplate in pluggable form factors such as ELSFP, while light is delivered via fiber to optical engines near the ASIC. This improves thermal management, allows lasers to be shared across multiple optical engines, and makes the system far more serviceable.<img src="/sites/default/files/inline-images/blog-compute-chrisitan-figure2v2.jpg" data-entity-uuid="1432a1f5-d14e-49c4-8159-2e58d3bb9209" data-entity-type="file" alt="CPO Architecture Using ELS" width="750" height="549" loading="lazy"> Figure 2 – CPO Architecture Using ELSFigure 2 shows a possible configuration of a system implementing CPO-based optical connectivity. The host IC (e.g., a GPU or Ethernet switch) is co-packaged with silicon photonics-based optical engines that perform the optical Tx and Rx functions. The light source used to enable such Tx function comes from a UHP laser inside pluggable ELSFP modules in the front plate of the system chassis, simplifying thermal management and replaceability.ELS turns CPO from an elegant idea into something that can actually scale in real data centers.<h3>Extending Optical Efficiency Beyond the Data Center</h3>While much of the focus in AI clusters is on short-reach interconnects, power efficiency also matters for campus-scale and inter–data center connectivity. “Coherent Lite” architectures have been developed to bridge the gap between direct-detect PAM4 links and traditional long-haul coherent systems.By eliminating wavelength tunability, simplifying modulation formats, and reducing DSP complexity, coherent-lite solutions deliver higher capacity per wavelength than direct-detect approaches and lower power and cost than full coherent systems. These links are well suited for reaches from a few kilometers up to tens of kilometers, supporting scalable AI fabrics that extend beyond a single facility.<h3>What This Means for Lumentum</h3>Across all these approaches—linear optics, pluggable and co-packaged designs, and campus-scale interconnects—the laser is a foundational enabler. Power efficiency, signal quality, and long-term reliability all start with the light source.Lumentum brings decades of leadership in indium phosphide laser technology, delivering the narrow linewidth, low noise, high output power, and reliability required for next-generation AI interconnects. Our <a href="https://www.lumentum.com/en/products/external-laser-source-els-module-ultra-high-power-laser">Ultra High Power (UHP) continuous-wave lasers</a> are designed specifically to support shared-light and external-laser architectures, enabling scalable, power-efficient optical fabrics.Recent demonstrations of 1.6T-class linear optical transceivers and ELS-based solutions show how advances in laser efficiency, packaging, and manufacturing scale are moving these architectures from concept to deployment.Looking ahead, as AI infrastructure continues to grow, power per bit will matter more than peak bandwidth. Linear optics, co-packaged architectures, external laser sources, and Coherent Lite solutions are all part of a broader shift toward tighter electrical–optical integration and more efficient data movement.The industry is already deploying 1.6T systems, and the path to 3.2T is taking shape. The winners won’t just be the ones who build faster chips—they’ll be the ones who move data with the least power, at scale.That’s where advanced photonics, and the lasers at their core, will make the difference.

AI data centers are scaling at an unprecedented pace, but the constraints shaping that growth are changing. While GPUs and accelerators continue to advance rapidly, power delivery, power efficiency, and heat dissipation increasingly define how far and how fast AI clusters can scale.

Why Power, Not Compute, Limits AI Scaling

I am thrilled to share the launch of our newly modernized Lumentum company website. It’s cleaner, faster, and designed to make it easier than ever to explore the products powering the future of AI infrastructure and data center connectivity. This new digital experience reflects not only our technological innovations but also our ongoing commitment to serving customers with cutting-edge photonics solutions.As we unveil this new online presence, I would like to take a moment to reflect on our ten-year journey — a decade defined by transformation, innovation, and growth. Since our beginnings as part of JDS Uniphase (JDSU), Lumentum has evolved into a global leader in optical communications, industrial lasers, and 3D sensing, enabling the world’s most advanced networks and devices. While there is not enough space here to capture every milestone, I have attempted to highlight a few in the timeline below that remind me of the progress we have made. These events stand as a testament to the dedication and vision of an entire team, many that came before us.Looking ahead, one thing is certain: the future is photonic, and Lumentum is at the forefront — illuminating the networks of tomorrow and the technologies that will define the next decade and beyond.We invite you to explore the new site. There is much more to come.<h2>A Decade of Lumentum Milestones</h2><h3>2015 — A New Beginning</h3>Lumentum was officially launched following the spin-off from JDSU, establishing a focused company dedicated to optical communications and commercial lasers. From day one, our purpose was clear: to accelerate the future through photonics innovation.<h3>2016 — Powering Next-Generation Networks</h3>Within our first full year, we strengthened our leadership in optical components for telecom and datacom. Lumentum’s products helped major network providers deploy higher-speed transmission systems — paving the way for today’s 100G and 400G infrastructure.<h3>2017 — Pioneering 3D Sensing</h3>Lumentum’s VCSEL (vertical-cavity surface-emitting laser) technology entered volume production for 3D sensing in consumer electronics, powering the first generation of devices with facial recognition and depth sensing. This breakthrough positioned Lumentum as a trusted partner in the emerging world of spatial computing.<h3>2018 — Scaling Manufacturing Footprint</h3>We expanded our manufacturing footprint in advanced laser manufacturing and wafer-fab capabilities through the acquisition of Oclaro. This move strengthened Lumentum’s market position and underscored our role as a key enabler of the world’s growing demand for bandwidth, connectivity, and intelligent optical solutions.<h3>2019 — Strengthening Our Portfolio</h3>The company expanded its product portfolio to meet rising demand for 100G and 400G optical components and coherent solutions powering hyperscale data centers and 5G networks.<h3>2020 — Resilience in a Challenging Year</h3>As the global pandemic reshaped supply chains and workforces, Lumentum remained operational and resilient. Our teams ensured uninterrupted delivery of critical optical and laser technologies that kept data centers, networks, and medical systems communicating — demonstrating the essential nature of photonics.<h3>2021 — Leading in High-Speed Data Center Components</h3>Lumentum accelerated production of its industry-leading 100G EMLs for hyperscale data centers by leveraging its high-volume InP capabilities, enabling higher bit rates with lower power.<h3>2022 — Expanding in Tunable and Coherent Components</h3>Through the acquisition of NeoPhotonics, we added advanced coherent components and InP-enabled solutions for ultra-high bandwidth and low-power applications to our portfolio.<h3>2023 — Expansion into Data Center Market</h3>A landmark year marked by the acquisition of Cloud Light which expanded our position in AI/ML-driven connectivity with cutting-edge optical transceivers and active optical cables.<h3>2024 — Developing the Next Generation</h3>We made significant progress in advancing our Optical Circuit Switch (OCS) platform by refining our MEMS-based architecture to meet the extreme density and performance requirements of large data centers operators. These enhancements set the stage for future large-scale deployment of Lumentum’s OCS solution.<h3>2025 — A Decade of Light, and a Bright Future Ahead</h3>As we celebrate our tenth anniversary, we stand at the forefront of the next era of photonics — where light enables intelligence, connectivity, and sustainability. From AI to data center infrastructure to laser and sensing, Lumentum continues to lead with innovation, integrity, and purpose.

Lumentum: Illuminating the Networks of Tomorrow

Optical circuit switching (OCS) is drawing intense interest when it comes to improving data center networking (DCN) in AI hyperscale environments, where efficient workload processing hinges upon massive parallelism, real-time data movement, and ultra-low latency networking. OCS was a leading topic of conversation at the recent ECOC Conference and Exhibition where we announced our new <a href="https://www.lumentum.com/en/products/64x64-optical-circuit-switch-ocs">OCS R64 platform</a>, the second member of the OCS family after the R300 which we introduced earlier this year at OFC in San Francisco.<style>
.small-image {
max-width: 500px;
margin-right:auto;
}
</style><div class="small-image"><img src="/sites/default/files/2025-11/r300ocs.png" alt="blog image" width="1600" height="1185" loading="lazy"></div>OCS R300OCS is seen as a replacement for traditional packet-based switching fabrics, such as DCN spine switches, which act as a central backbone connecting all leaf switches, which in turn connect to servers and storage. Replacing spine switches with an OCS would reduce significantly both power consumption and networking latency, which are top requirements to address the scale and performance demands of AI and ML workloads.<img src="/sites/default/files/2026-01/OCS-diagram.jpg" alt="blog image" width="704" height="446" loading="lazy">We recently presented our views at an <a href="https://cignal.ai/2025/07/optical-circuit-switching-ocs-for-ai-scaling-and-datacenter-automation/" target="blank">OCS webinar</a> hosted by the industry analyst firm Cignal AI. The webinar focused on AI scaling and data center automation / increased efficiency and covered how technologies such as MEMS, silicon photonics, robotic fiber switching and digital liquid crystal can reduce operational complexity, increase speeds, lower power consumption and enable larger networks.A frank discussion also took place on the ongoing deployment challenges in various environments and the market trajectory for these applications. One can conclude from this webinar that this sector is growing and evolving rapidly and that well-funded technology platforms have a chance of gaining significant market share. What follows is a closer examination of a MEMS-based OCS and its benefits in terms of scalability, latency reduction, and energy efficiency in hyperscale data center applications.A MEMS OCS features an array of tiny movable mirrors embedded on a silicon substrate to direct optical signals between input and output fibers. Each input port corresponds to an optical fiber carrying high-capacity wavelength channels and each output port connects to another fiber leading to a computing node or another switch. This is part of the “all optical” model that reduces power consumption and latency as the signal is not converted into the electrical domain, as is the case with conventional packet / spine switches.As a result, the R300 has demonstrated 98% reduction in switch latency vs. an OEO packet switch solution. Furthermore, our 3D-MEMS architecture incorporates mirrors that tilt continuously along two axes, allowing precise alignment of any input beam to any output fiber across a large free-space switching fabric. Our single-crystal silicon MEMS design also eliminates “hinge fatigue” and enables hundreds of millions of operational cycles with no failures.This capability is embodied in the R300, which currently supports up to 300x300 ports with a typical insertion loss of 1.5 dB or less. Furthermore, the high switch radix of the R300 enables a flat network architecture that can scale to support the massive numbers of GPUs (100,000+) needed for hyperscale AI deployments. And a note about the importance of low-loss performance –this is a critical feature, as it is one of the key technical hurdles to the introduction of OCS technology into hyperscale environments. This is seen in the requirement to support additional OCS loss within transceiver link budgets – around 4dB for FR4/DR4 and 6dB for LR4/ER4.For MEMS systems in general, the control plane is instrumental in establishing a circuit by computing the required tilt angles and activating the mirrors accordingly. Once set, the optical path remains stable and creates a dedicated, high-bandwidth channel until the configuration is changed.The control software for the R300 is based on the popular, open-source SONiC OS and augments our proprietary MEMS design, which does not need a “closed loop” control mechanism. SONiC is also beneficial, in terms of “seamless integration,” in that it provides an interface to integrate the OCS into existing datacenter network level control planes.Operating costs are reduced further via the R300’s ultra-wideband capability, i.e., seamless access across the O-, C- and L-bands and its bit-rate transparency. In fact, the “bit rate” element is another key advantage of MEMS systems, as a MEMS OCS does not impose a bandwidth limit beyond the optical fiber itself. Whether the link operates at 400 Gbps, 800 Gbps, or future 1.6 Tbps line rates, the switch remains protocol- and rate-transparent. This makes it inherently future-proof and capable of scaling with next-generation optical transceivers.<h3>Closing Thoughts and Next Steps</h3>Training large AI models involves hundreds or thousands of graphical processing units (GPUs) or tensor processing units (TPUs) working in tightly coupled clusters. These accelerators require rapid exchange of data, gradient updates, and parameter synchronization. The resulting east-west traffic patterns within the data center are characterized by high volume, burstiness, and the need for low-latency communication.Traditional electronic packet switches are now struggling to handle these traffic demands, as they can suffer from buffering and queuing delays and the power consumption factor makes large fabrics unsustainable.A MEMS-based OCS, on the other hand, complements or overlays the packet-switched network by providing direct, reconfigurable optical paths between clusters of compute nodes. For example, the OCS can dynamically create dedicated optical circuits interconnecting the hundreds of GPU nodes assigned to a large training job. This eliminates intermediate hops through multiple electronic switches, significantly reducing both latency and congestion. Additionally, with the new AI training applications being implemented, the traffic is generally considered to be very predictable, which is beneficial for OCS.As for power consumption, unlike electronic switches that consume power continuously in processing and forwarding packets, a MEMS switch consumes energy primarily during reconfiguration of mirrors. Once set, the mirrors hold their position with minimal static power draw. This leads to orders-of-magnitude lower energy consumption per switched bit.Looking down the road, opportunities for future OCS evolution include the implementation of multi-hop optical switching, faster switching – from the milliseconds of today to the microseconds for optical burst switching, and very high radix fabrics, as 3D MEMS switches can scale to very large port counts (e.g., 1024×1024 or beyond) in a compact footprint.For more information, visit our <a href="/en/products/data-center/optical-circuit-switches">Optical Circuit Switches page.</a>

Optical circuit switching (OCS) is drawing intense interest when it comes to improving data center networking (DCN) in AI hyperscale environments, where efficient workload processing hinges upon massive parallelism, real-time data movement, and ultra-low latency networking. OCS was a leading topic of conversation at the recent ECOC Conference and Exhibition where we announced our new OCS R64 platform, the second member of the OCS family after the <a href="https://www.lumentum.com/en/media-room/news-releases/lumentum-optical-circuit-switch-improve-next-generation-ai-data-center">R300</a> which we introduced earlier this year at OFC in San Francisco.

Optical Circuit Switches in AI Hyperscale Data Centers – The MEMS Advantage

As demand for artificial intelligence, ubiquitous cloud computing, and other emerging data-intensive applications continues to surge, the need for larger and more energy-efficient data centers has become increasingly critical. Meeting this demand requires a leap in infrastructure— and Lumentum is at the forefront of this leap with its various next-generation photonic technologies and products.<h3>The Need for Next-Generation Data Centers</h3>Traditional data centers are struggling to keep up with the exponential growth in data traffic and compute workloads. AI/ML training and inferencing already require massive bandwidth density and low-latency interconnects, and those requirements are expected to grow exponentially. At the same time, total power consumption and scalability are becoming critical concerns. Lumentum, a global leader in optical networking and photonic solutions, is addressing these challenges with innovations designed to transform the AI infrastructure ecosystem.<h3>Indium-Phosphide Photonics Solutions</h3>At the core of Lumentum’s innovation lies our advanced indium phosphide (InP) platform, enabling ultra-high-speed optical links—up to 200 Gbps per lane today and expected to scale to higher rates in the near future. Lumentum’s energy-efficient InP technology underpins pluggable optical transceivers operating at 400 Gbps, 800 Gbps, 1.6 Tbps, and beyond, across back-end and front-end intra-data center, as well as data center interconnect (DCI) coherent applications. These solutions are essential for scaling today’s AI infrastructure that powers everything from large language models to real-time data analytics. Looking ahead, one of the most exciting developments on the horizon is the move toward co-packaged optics (CPO). The concept of CPO is to integrate optical connectivity in the same package as the host IC (GPU, CPU, Ethernet switch or any other ASIC). This integration brings the ability to support very high bandwidth density links (e.g., multiple Terabit per second per fiber) with very low power consumption over extended distances, enabling larger AI compute clusters spanning multiple equipment racks. Although its initial applications will be for AI scale-out networks, this innovation is key to supporting the next generation of AI accelerator scale-up, given that today’s copper connectivity and traditional pluggable optics are expected to fall short due to their limited reach and high power consumption, respectively. Lumentum is well positioned to be a leader in the emerging CPO ecosystem, with our <a href="https://www.lumentum.com/en/products/industrial-laser/ultrashort-pulsed-lasers">ultra-high power (UHP) laser</a> suitable for external laser source (ELS) pluggable modules, &nbsp;as well as other more fully integrated multi-Terabit per second connectivity solutions on our roadmap.<img src="/sites/default/files/2026-01/AIblog-diagram.jpg" data-entity-uuid="2dccd1b6-9170-4a77-a266-46789fef23a1" data-entity-type="file" alt="UHP laser for CPO applications" width="819" height="257" loading="lazy"> Figure 1 – UHP laser for CPO applicationsFigure 1 shows a possible configuration of a system implementing CPO-based optical connectivity. The host IC (e.g., a GPU or CPU) is co-packaged with a silicon photonics (SiPh)-based optical engine that performs the optical Tx and Rx functions. The light source used to provide such Tx function comes from a UHP laser inside a pluggable ELS module in the front plate of the system chassis, simplifying thermal management and replaceability.<h3>Optical Circuit Switch (OCS): A Game-Changer for AI Infrastructure</h3>Lumentum’s <a href="https://www.lumentum.com/en/products/300x300-optical-circuit-switch-ocs">R300 Optical Circuit Switch (OCS)</a> is another breakthrough product designed for the AI ecosystem. Built on our industry-leading MEMS technology, the R300 enables dynamic, low-latency optical switching between thousands of endpoints. Unlike traditional Ethernet-based architectures, which are power-hungry and less scalable, OCS offers a more energy efficient and flexible alternative. The R300 OCS shown in Figure 2 supports up to 300x300 ports with a typical insertion loss of 1.5 dB, and can scale to support 100,000+ GPU clusters, making it ideal for hyperscale AI deployments. It also reduces power consumption by up to 65% compared to conventional switching fabrics—an essential advantage as data centers strive to meet sustainability goals.<img src="/sites/default/files/2026-01/blog2-OCS.jpg" data-entity-uuid="2dccd1b6-9170-4a77-a266-46789fef23a1" data-entity-type="file" alt="R300 Optical Circuit Switch (OCS)" width="819" height="492" loading="lazy">Figure 2 – Lumentum R300 OCS<h3>Real-World Impact and Industry Momentum</h3>Lumentum’s innovations are already showing their impact across the industry. At OFC 2025, we showcased a <a href="https://investor.lumentum.com/financial-news-releases/news-details/2025/Lumentum-to-Showcase-Cutting-Edge-Optical-Innovations-for-Cloud-AI-and-Networking-Infrastructure-at-OFC-2025/default.aspx" target="blank">live demonstration</a> of our InP electro-absorption modulated lasers (EMLs) running at 448 Gbps, paving the way for 3.2 Tbps optical connectivity, as well as our R300 OCS product – both of which drew strong interest from hyperscale cloud service providers and others in the AI ecosystem. These technologies are not just theoretical—they are expected to be adopted soon in real-world deployments to enable the next generation of data centers. Market analysts predict very rapid growth in demand for optical interconnects and circuit switching as AI continues to scale (see Figures 3 and 4). Lumentum is well-positioned to advance this transformation with its robust portfolio and deep expertise in photonics, based on decades of innovation and market leadership.<img src="/sites/default/files/2026-01/blog2-figure3.jpg" data-entity-uuid="2dccd1b6-9170-4a77-a266-46789fef23a1" data-entity-type="file" alt="Live demonstration" width="819" height="458" loading="lazy">Figure 3 – Impact of AI Applications on optical transceiver market volume*<img src="/sites/default/files/2026-01/blog2-figure4.jpg" alt="Live demonstration" width="819" height="458" loading="lazy">Figure 4 – Expected market growth for optical circuit switching**<h3>Conclusion</h3>As the digital world accelerates its transformation, data centers must evolve to keep pace. Lumentum’s Cloud and Networking solutions—ranging from high-speed lasers, photonic chips, and pluggable transceivers, to scalable optical switching and CPO—provide a path to a more energy-efficient and sustainable future.To learn more about how Lumentum is shaping the future of AI infrastructure, visit <a href="https://www.lumentum.com/en">www.lumentum.com.</a>*Source: LightCounting, “Optics for Cloud Data Centers”, July 2025 **Source: LightCounting, “Ethernet, InfiniBand and Optical Switches for Cloud Data Centers”, April 2025

As demand for artificial intelligence, ubiquitous cloud computing, and other emerging data-intensive applications continues to surge, the need for larger and more energy-efficient data centers has become increasingly critical. Meeting this demand requires a leap in infrastructure— and Lumentum is at the forefront of this leap with its various next-generation photonic technologies and products.

Enabling the Future of AI: Innovation for Next-Generation Data Centers

The network diagrams below help to provide an overview of where the key Lumentum Cloud and Networking products are typically deployed in an AI/data center or telecommunications network. The Lumentum products underpinning these applications can be organized into four primary photonics functions. Here, we elaborate on how those industry-leading products allow for a world of high speed, interconnected optics that transmit, receive, amplify, and switch.<figure role="group" class="caption">
<img alt="AI-DataCenter-diagram" data-entity-type="file" data-entity-uuid="5765d963-6f7c-4138-b265-1363e1f8b900" height="461" src="/sites/default/files/inline-images/AI-DataCenter-diagram.jpg" width="820" loading="lazy">
<figcaption>AI and Data Center Network</figcaption>
</figure>
<h3>Transmit</h3>All optical communications are initiated by devices that send or transmit data. These are composed of lasers and modulators, and sometimes these functions are combined into a single chip.<ul><li>Electro-absorption modulated lasers (EMLs)</li></ul>EMLs integrate a diode laser and an electro-absorption modulator onto a single chip, which results in efficient modulation of the laser’s light output and makes it suitable for high-speed (up to 400G and above) data transmission.<ul><li>Silicon photonics modulators</li></ul>Built on a similar CMOS process to traditional electronic components, silicon photonics modulators enable high levels of integration and are driving the recent industry advances in co-packaged and near-packaged optics.<ul><li>Continuous wave (CW) lasers&nbsp;</li></ul>CW lasers are built using indium phosphide (InP) processes and generate high-power unmodulated light. They are primarily used as a light source to power silicon photonics modulators. Lumentum also offers our industry leading ultra-high power (UHP) laser in this category based on our high-reliability pump laser that is capable of providing over 400 mW of power.<ul><li>Coherent modulators</li></ul>Lumentum builds a range of coherent modulators based on our high-speed, low-drive voltage InP platform.<ul><li>Tunable lasers</li></ul>Tunable lasers are used as CW light sources for DWDM applications in metro and long-haul networks. They are capable of tuning across a wide range of wavelengths. Lumentum’s tunable lasers lead the industry in performance and shipments, based on our external cavity laser (ECL) technology.&nbsp;<ul><li>Vertical cavity surface emitting lasers (VCSELs) and directly modulated distributed-feedback lasers (DFBs)</li></ul>VCSELs are the lowest cost, lowest power consumption lasers for communications, however, they are only suitable for short reaches (up to 100 m). Directly modulated DFBs operate in a compact and economical footprint for lower data rates.<h3>Receive</h3>Lumentum develops photodiodes for a host of different applications, including high-speed photodiodes for data center and telecommunications applications and for tap/monitor applications. Lumentum’s transceivers include transmitters and receivers in line with the transmit and receive technologies above.<figure role="group" class="caption">
<img alt="Telecommunications Network" data-entity-type="file" data-entity-uuid="4f237e42-7ad1-4cdc-b727-79e4bdb859a1" height="459" src="/sites/default/files/inline-images/Telecom-Network-diagram.jpg" width="820" loading="lazy">
<figcaption>Telecommunications Network</figcaption>
</figure>
<h3>Amplify</h3>Amplification is at the heart of every telecom network. We have an extremely broad-based selection of amplifiers from 980 nm pumps to Raman amplification (which enhances the strength of optical signals via the stimulated Raman scattering effect within an optical fiber) to submarine pumps, and even amplifiers that are used in satellite applications. Here, reliability is key, as a single amplifier is typically loaded with tens of Terabits of traffic, and nowhere is reliability more critical than in the submarine part of the network where Lumentum is a clear market leader. Lumentum also has semiconductor optical amplifier (SOA) technology. These mini amplifiers are built in InP and can be integrated with our chips, including lasers and SiPho chips, and are increasingly believed to be essential elements of next-generation transmitters and receivers.<h3>Switch</h3>We have multiple products that switch, including:<ul><li>Wavelength selective switch (WSS) products that switch individual wavelengths at nodes in metro-area and long-haul networks. These are built on decades of knowledge in micro-electro-mechanical systems&nbsp;(MEMS) and liquid crystal on silicon (LCoS) based technologies. WSS technology has a long history (debuting in the early 2000s), and our MEMS technology proved vital in its development. WSS acts as a flexible and programmable switch in ROADMs by dynamically routing different wavelength channels to various output ports without the latency created by electrical processing.</li><li>The optical circuit switch (OCS) is an exciting new product for us. Targeting data center applications, our OCS is built on a single-crystal silicon MEMS design that has been proven in the field with hundreds of millions of operational cycles and in more than 150,000 WSSs deployed to date in telecom networks. The OCS can eliminate an entire layer of switching in the data center, with significant reductions in power consumption and cost, while simplifying the upgrade cycle for future speeds. It is wavelength agnostic, bit-rate transparent, and provides very low insertion loss.</li></ul>We hope you enjoyed this inaugural post of our company blog and that this overview has inspired you to learn more about our “transmit, receive, amplify, and switch” portfolio.&nbsp;For more details, please visit the <a href="https://www.lumentum.com/en/products">Products </a>page of our website.

Lumentum is a company that is built on 40+ years of heritage as a pioneer and ground breaker in optics and photonics. At the root of this heritage has always been our products. We focus on reliability, and we focus on performance.

Transmit. Receive. Amplify. Switch.

OFC 2026 made one point unmistakably clear: AI infrastructure is no longer driven solely by higher bandwidth. It is pushing the entire optical ecosystem to solve for scaling in bandwidth, power, thermal efficiency, reliability, manufacturability, and supply-chain resilience.&nbsp;That shift was visible across the exhibition and conference. 1.6T optics and 200G/lane technologies are moving from feasibility into deployment, co-packaged optics (CPO) is advancing from concept toward industrialization, new MSAs are emerging around scale-up architectures, and 400G/lane is becoming the next major development target. At the same time, enabling technologies such as external laser sources, VCSEL-based scale-up optics, and optical circuit switching (OCS) are gaining strategic importance.&nbsp;&nbsp;Against that backdrop, OFC 2026 was an important opportunity to show how Lumentum’s technology portfolio maps directly to these market transitions.&nbsp;Next-Generation Optical Connectivity: The Path to 3.2T&nbsp;A good example is the industry’s move to 1.6T pluggable modules. This generation is now clearly taking shape around 200G/lane PAM4, with multiple implementations emerging across the hyperscale ecosystem. At our booth, Lumentum demonstrated two live NVIDIA 1.6T 2xDR4 OSFP transceiver modules built on different optical architectures, but both enabled by Lumentum lasers: one incorporating our Ultra High-Power (UHP) laser as the CW light source for an NVIDIA silicon photonics engine, and the other using our 200G EMLs in an 8x200G architecture. The modules were shown in optical loopback with real-time BER results, highlighting both signal integrity and practical module operation.<img src="/sites/default/files/inline-images/UHP%20NVIDIA%20demo%20v2.jpeg" data-entity-uuid="0e565f2d-5218-421b-840a-5ef2ebbad9cd" data-entity-type="file" alt="Nvidia 1.6T 2xDR4 transceiver module using Lumentum UHP lasers" width="599" height="258" loading="lazy">Image: NVIDIA 1.6T 2xDR4 transceiver module using Lumentum UHP lasersThose demonstrations mattered for more than performance alone. They showed that the path to 1.6T is not defined by a single device architecture. Customers increasingly need multiple high-performance, high-reliability technology options to support different module designs, integration strategies, and supply-chain requirements. This aligns well with Lumentum’s depth in indium phosphide laser technology across both external laser and EML-based architectures, backed by manufacturing scale.&nbsp;In parallel, OFC 2026 also highlighted early progress toward the next scaling step: 400G per lane. While most demonstrations remained at the component level, Lumentum showed a live 4x400G EML-based pluggable module, illustrating a practical path to future 3.2T pluggable modules based on 8x400G architectures. This demonstration reinforces that indium phosphide-based EML technology can scale to higher data rates while maintaining the signal integrity and reliability required for data center deployments. It also highlights the next system challenge: not just achieving 400G per lane at the device level, but closing power, thermal, and integration constraints at the module level.&nbsp;<img src="/sites/default/files/inline-images/Eye%20Diagram%20-%201.6T%20DR4%20OSFP%20Module%20with%20400G%20Differential%20EML.jpg" data-entity-uuid="22042dac-8481-42f3-a88a-83645e1eeb9f" data-entity-type="file" alt="6 DWDM wavelengths @200GHz supported by Lumentum’s UHP laser" width="599" height="328" loading="lazy">Image: Optical transmitter eye pattern from Lumentum 400G differential EML&nbsp;New Optical Architectures for AI Network Scaling&nbsp;OFC also made clear that CPO is moving from concept to deployment readiness. The discussion is no longer just about whether CPO can work. It is about how to industrialize it, how to standardize the ecosystem around it, and how to support multiple architectural approaches for scale-up networks. New initiatives such as the OCI, Open CPX, and XPO MSAs all reflect that the industry is trying to move from isolated demonstrations to deployable system architectures.&nbsp;&nbsp;Lumentum’s OFC demonstrations reflected that same transition. In external laser sources, we showed that our first generation of UHP-based ELSFPs is now being extended in two important directions. First, we demonstrated an 800 mW Super High-Power (SHP) laser, designed to support higher optical power fan-out to a larger number of optical engines. Second, we demonstrated 16-channel DWDM UHP lasers, addressing future fiber-constrained architectures where higher wavelength density and centralized laser delivery become increasingly important.&nbsp;&nbsp;<img src="/sites/default/files/inline-images/16%20DWDM%20wavelengths%20supporyed%20by%20UHP%20laser.jpg" data-entity-uuid="791cdbcd-8fe8-4197-97c4-6f9f2bc569bb" data-entity-type="file" alt="16 DWDM wavelengths @200GHz supported by Lumentum’s UHP laser" width="599" height="335" loading="lazy">Image: 16 DWDM wavelengths @200GHz supported by Lumentum’s UHP laserAnother important theme at OFC was the rise of alternative scale-up architectures. While much of the industry focus remains on silicon photonics and micro-ring-modulator-based CPO, there is growing interest in VCSEL-based approaches for so-called “slow-and-wide” scale-up fabrics, where lower-speed parallel optical channels can improve overall energy efficiency by reducing dependence on very high-speed electrical SerDes. Here again, Lumentum showed relevant leadership, with a live VCSEL-based optical scale-up architecture demonstration that highlighted a different path to solving the same system problem: how to move very high bandwidth efficiently at rack and cluster scale. Industry analysts specifically noted Lumentum’s VCSEL-based CPO demonstration as an alternative to more supply-constrained indium phosphide approaches in some scale-up scenarios.&nbsp;&nbsp;<img src="/sites/default/files/inline-images/FOWLP%20on%20coin.jpg" data-entity-uuid="9a2d20cc-e175-4666-9f63-f7d08a7534bb" data-entity-type="file" alt="Lumentum 10 Tbps VCSEL+PD optical engine prototype based on 3D FOWLP" width="349" height="281" loading="lazy">Image: Lumentum 10 Tbps VCSEL+PD optical engine prototype based on 3D FOWLPScaling AI Fabrics and System-Level Implications&nbsp;Beyond pluggable optics and CPO, OFC 2026 also reinforced the growing role of OCS in AI fabrics. As AI clusters scale, static network topologies become less efficient. OCS offers a path to more dynamic, energy-efficient connectivity and is gaining momentum as AI fabrics scale. Lumentum’s position in this market was visible not only through our own portfolio, but also through ecosystem validation, including Marvell’s live booth demo featuring our R300 OCS.&nbsp;Taken together, these demonstrations showed how Lumentum is positioned across multiple layers of the emerging AI optical stack. We are not participating in just one transition. We are addressing several of them at once: 1.6T pluggable modules with 200G/lane EMLs and UHP lasers, the emerging 400G/lane module ecosystem, next-generation external laser sources for CPO and NPO, VCSEL-based scale-up optics, and OCS for more dynamic AI fabrics.&nbsp;The direction is clear: AI infrastructure is no longer defined by a single optical architecture or a single scaling vector. It is evolving into a set of parallel approaches—pluggable, near-packaged, and co-packaged—each optimized for different constraints in power, density, reach, and serviceability.&nbsp;Enabling this environment requires more than individual device innovation. It requires the ability to deliver high-performance optical technologies across multiple architectures, at scale, and with the reliability and manufacturability demanded by hyperscale deployments.&nbsp;This is where the next phase of optical innovation will be defined—and where the role of photonics continues to expand from connectivity to system-level enablement.&nbsp;&nbsp;<a href="https://www.lumentum.com/en/events/ofc-2026">Visit our show recap</a> to see all Lumentum technology demos from OFC.

OFC 2026 Takeaways, Part 2 – From Industry Shifts to Technology Implementation

Blog

Insights and perspectives from the frontlines of photonics innovation.

Enabling the Next Phase of AI Optical Infrastructure

The New Photonics Ecosystem for the AI/ML Infrastructure

Advanced Packaging and Heterogeneous Integration: Reshaping Photonic Architectures for AI

EMLs: The Unsung Heroes Powering AI's Digital Arteries

Why Power, Not Compute, Limits AI Scaling

Lumentum: Illuminating the Networks of Tomorrow

Optical Circuit Switches in AI Hyperscale Data Centers – The MEMS Advantage