Research|Optics: CIOE Takeaway - 2026 Orders Extend into 2027, Supply Determines Who Delivers First

The clearest takeaway from this year’s CIOE is that 1.6T orders and production schedules now extend into 2027. Demand is growing rapidly, and the availability of DSPs, lasers and PCBs will determine the pace of deliveries and vendor market share. Zhongji InnoLight (InnoLight), Eoptolink, Coherent and Lumentum have all entered volume production of 1.6T products, supported by the strongest customer bases and most secure supply chains. For second-tier vendors such as Cambridge Industries Group (CIG), Linktel and Source Photonics, growth will come first from higher 800G shipments, followed by the introduction of 1.6T products. Adoption of near-packaged optics (NPO) is broadening, creating additional demand for high-power light sources, fiber arrays and connectors. Demand is also taking shape for 2028: Rubin Ultra, TPU v9 and Chinese supernodes already incorporate optical interconnects into their designs.

Held in Shenzhen on 9–11 September, the show recorded nearly 190,000 visits over three days, up 32% year on year and a record for CIOE. The halls were packed, visitors queued to see samples at popular booths, and many exhibitors posted recruitment notices. One exhibitor said that, having booked late, it could secure only a small booth.

1. Order visibility extends into 2027 and supply determines the pace of shipments

Vendors at the show consistently reported strong orders. The leading suppliers agreed that 1.6T shipments have begun to ramp in 2026 and that demand will rise sharply in 2027. Market estimates for 2027 demand run to the high tens of millions of units or more, and production schedules already extend into that year. Common refrains were that orders had grown much faster than expected this year and that suppliers could not accept all the orders coming in. InnoLight expects continued rapid growth in both 1.6T and 800G orders in 2027. At Eoptolink, 800G now accounts for the bulk of shipments, while 1.6T volumes rose substantially quarter on quarter in Q2. Yuanjie Semiconductor Technology reported approximately 1,200% year-on-year growth in first-half net profit attributable to shareholders. Long-term supply agreements and capacity investments are progressing in parallel: Corning signed an agreement with Verizon covering more than 80 million fiber-miles through 2032, while Far East Smarter Energy is investing nearly RMB 2b in an optical fiber preform project for AI data centers, with annual capacity of 1,800 tonnes.

Research firms continue to raise their demand forecasts. At the forum, LightCounting raised its forecast for 2026 Ethernet optical transceiver sales growth to more than 70% and its 2031 market estimate to $74b, with another update due in late September. Yole projected a total optical transceiver market of $112.3b in 2031, while Cignal AI reported that Q1 datacom optical component revenue doubled year on year to $7.7b. These figures cover different markets and periods: annual Ethernet transceiver sales, annual sales of all optical transceivers, and quarterly datacom optical component revenue. They therefore provide separate indicators of industry strength rather than directly comparable market estimates. Upstream procurement reinforces that picture. NVIDIA, Google and Meta are reserving capacity directly with laser manufacturers; Lumentum is securing higher prices on new orders and has signed multiyear agreements. InnoLight’s and Eoptolink’s prepayments rose to RMB 1.488b and RMB 682m, respectively, in Q1, reflecting a greater commitment to securing supply. Everbright Photonics’ chairman expects this cycle to last at least three to five years, and potentially a decade. We share his view that this will be a sustained cycle.

2. Tight 3nm DSP supply will keep the leaders ahead in 1.6T in 2027

Almost every transceiver vendor at the show claimed 1.6T capability, but actual shipments reveal three distinct groups. InnoLight, Eoptolink, Coherent and Lumentum have all entered volume production. Among second-tier vendors, CIG, Linktel and Source Photonics are advancing relatively quickly: they have secured upstream components, are supplying 800G products in volume to major Western customers, and are qualifying and introducing 1.6T products. CIG has annualized 1.6T capacity of 3.5 million units. These vendors should see strong growth in 2027, with potential earnings upside at CIG and Linktel. Most other vendors are still qualifying their products, and 800G remains the bulk of their shipments. The Chinese installed base of 1.6T switches remains small, and transceiver deliveries follow the rollout schedules of server and switch manufacturers. Yole does not expect 1.6T to become the industry’s main driver of shipment growth until 2027.

DSP availability is the key constraint on 1.6T. Leading 1.6T transceivers primarily use Broadcom’s Sian3 and Marvell’s Ara, both manufactured on a 3nm process. Ara is already shipping in volume; Ara X, Ara T, Petra and Aquila M began sampling in 26Q1. Both companies manufacture their DSPs on TSMC’s 3nm process, competing with AI accelerators for wafer capacity. Transceiver output therefore depends first on TSMC’s wafer allocation to optical DSPs, and then on how Broadcom and Marvell allocate finished DSPs among customers. Customer programs and DSP procurement are closely coordinated, giving leading transceiver vendors more reliable access to supply. That should keep them in the lead in 2027.

Several developments could ease DSP supply constraints. TSMC’s 3nm capacity expansion may help, and alternatives to Broadcom and Marvell are emerging. MaxLinear’s Rushmore 1.6T DSP is manufactured by Samsung, avoiding dependence on TSMC’s 3nm capacity. If it qualifies with leading customers and enters volume production in 2027, it would give second-tier transceiver vendors a route to 1.6T that does not depend on Broadcom’s or Marvell’s allocations. Credo’s 1.6T linear receive optics (LRO) solution still uses its 3nm Bluebird DSP, but reduces the extent of retiming and lowers power consumption. In short-reach applications, it can reduce demand for fully featured DSPs. Linear pluggable optics (LPO) is already proven at 800G, and Eoptolink confirmed progress with customers. Marvell said that fully retimed optics (FRO) currently accounts for roughly 95% of 800G products and LPO for 5%; a larger LPO share would reduce DSP demand. NVIDIA’s in-house 1.6T DSP reduces its reliance on merchant DSP allocations, although wafer capacity and packaging remain constraints. For second-tier vendors, 2027 will most likely bring continued growth in 800G shipments and a gradual introduction of 1.6T. The 800G market itself is still expanding, leaving room for these vendors to grow.

3. 400G per lane has been demonstrated but the route to volume production remains open