By Andee | 28 April 2026 | 0 Comments
Why is the demand for optical modules so high this year (2026)?
The demand for optical modules surged this year (2026), primarily driven by the explosive growth of AI computing clusters, bandwidth upgrades, the shift from copper to fiber optic networks, and increased capital expenditure by cloud providers. This transformed them from "optional components" into "essential arteries," with speeds jumping from 400G to 800G/1.6T. The usage per AI server is 5-10 times that of traditional systems.
I. Core Reason: Explosive Growth in AI Computing Power (The Biggest Engine)
· Dual Explosion in Training and Inference: Large model training requires tens of thousands of interconnected GPUs, exchanging terabytes of parameters per second; the inference side has an even higher demand density, 3-5 times that of the training side.
• Dramatic Increase in Single-Unit Usage:
* Traditional Servers: 1–2 100G/400G units
* AI Servers (H100/B100): 8–12 800G units
* NVIDIA GB200: 162 1.6T units
• Continuously Increasing Ratio: GPU: Optical Module ratio from 1:2 → 1:5 → 1:8 (ASIC cluster), optical module growth rate > GPU growth rate.
• Widespread Adoption of 10,000-GPU Clusters: Large-scale deployments by Meta, Google, and NVIDIA have driven 800G/1.6T orders to be booked until the end of the year.
II. Bandwidth Upgrade: 400G → 800G → 1.6T, a Two-Tier Leap in Three Years
• 800G Becomes Mainstream: Demand for 40 million+ units in 2026, Goldman Sachs raised its forecast to 33.5 million units (+58%), with Meta alone demanding 10-12 million units.
• 1.6T Year's Demand Surge: Driven by NVIDIA's GB200/300, demand is expected to reach 15-20 million units, a year-on-year increase of **600%**, with Google's order revised upwards to 15 million units.
• The Irreversible Shift from Copper to Fiber: Copper cables (DACs) have bandwidth ≤400G, distance ≤5m, and high power consumption; 800G+ necessitates optical modules, with LPO/CPO further replacing traditional pluggable modules.
III. Cloud Makers' Capital Expenditure Soars
• The world's four major cloud makers (Meta/Google/Amazon) + NVIDIA will have capital expenditures exceeding $400 billion in 2026, with AI infrastructure accounting for >60%.
• Domestic Follow-up: ByteDance, Alibaba, and Tencent are accelerating their intelligent computing centers, with millions of 800G/1.6T units procured.
• Regional Growth: Computing power projects in the Middle East (Saudi Humain) and Southeast Asia are being implemented, adding 1.6T orders. IV. Market Data (Intuitive Perspective)
· Overall Size: The global optical module market will reach $28.75 billion in 2026 (+25%), with AI accounting for 62.6%.
· High-Speed Market Share: 800G/1.6T will account for over 60% of shipments, with total demand exceeding 100 million units.
· Resilient Prices: 800G average price is $800–$1200, 1.6T is $2000–$3000, with slow price reductions and long delivery times due to supply shortages.
V. Why This Year (2026)?
· AI is transitioning from "experimentation" to "large-scale deployment": 2023–2025 will see investment in training, with a surge in inference demand in 2026, resulting in higher demand density.
· Mature Technology + Capacity Expansion: Improved yield rates for silicon photonics/LPO, ramp-up of 800G capacity, and commercialization of 1.6T will result in supply barely keeping up with demand.
• Order Pre-positioning: Customers are locking in 2026-2027 production capacity in advance, leading to shortages, price increases, and long lead times.
VI. Impact on You (Procurement/Selection)
• Low-end (1G/10G/40G): Stable demand, lower prices, and ample inventory.
• Mid-to-high-end (100G/400G): Steady growth, lead time 2-4 weeks.
• High-speed (800G/1.6T): Severe supply shortage, lead time 8-16 weeks, firm prices, priority given to large customers.
In short: The AI computing power revolution + bandwidth necessity + cloud companies investing heavily + fiber-to-the-home (FTTH) replacement of copper—these four factors combined have transformed optical modules from "small components" into "strategic necessities," resulting in explosive demand growth this year.
What is the market size for optical modules?
I. Global Market Size (2026)
· LightCounting / Most Industry Reports: $28.75-30 billion
o YoY: +25%～30%
· Goldman Sachs (Optimistic): $43-52 billion
· 2025 Base: Approximately $23 billion
II. Shipments (2026)
· Total Speeds: Approximately 420-450 million units
· High-Speed Mainstream:
o 800G: 34-40 million units
o 1.6T: 14-25 million units (only 5 million units in 2025)
III. Market Structure (Where is the Money Going?)
· AI Data Centers: $18-26 billion, accounting for 60%+
· Cloud Computing (Non-AI): Approximately 24%
· Telecommunications (5G/Transmission): Approximately 13%, basically stable
IV. China Market
· Size: Approximately $56-65 billion =RMB 100 million (≈80-95 billion USD)
· Global market share: 20%-25%
· Production capacity: Chinese manufacturers account for approximately 70% of the global market (almost monopolistic in the low-to-mid-range market, rapid growth in high-end 800G/1.6T)
What factors mainly influence the growth of the optical module market?
The core drivers of the optical module market growth can be summarized as: the essential demand for AI computing power, accelerated technological iteration, expanded capital expenditure of cloud manufacturers, the irreversible shift of fiber optics to copper, and policy support and emerging application scenarios. These five factors resonate, propelling the industry into a high-growth cycle of "both volume and price increases."
I. Explosive Growth in AI Computing Power (Core Engine, Contributing 60%+ to Incremental Growth)
· Dual-engine drive of training and inference: Large model parameters reach trillions, training clusters expand from thousands of cards to hundreds of thousands of cards, exchanging petabytes of data per second; optical interconnect is the only feasible solution.
•Surge in Single-Unit Usage:Traditional servers use 1-2 optical modules; AI servers (H100/B100) use 8-12 800G modules; NVIDIA GB200 servers reach 162 1.6T modules, with the GPU:optical module ratio increasing to 1:8-1:12.
•Cluster Scalability: Meta, Google, and NVIDIA are widely adopting 10,000-card-level clusters. AI server shipments are projected to reach 3.5 million units in 2026 (+40%).
II. Accelerated Technology Iteration (Core Factors: Simultaneous Increase in Quantity and Price)
Shortened Speed Upgrade Cycle: The iteration cycle from 400G to 800G to 1.6T has shortened from 3-4 years to 2 years. 3.2T is expected to be commercially available in 2027-2028.
High-Speed Penetration:
800G: Demand is projected at 40-50 million units in 2026, becoming mainstream.
o1.6T: Shipments of 15-25 million units in 2026 (+600%), with NVIDIA and Google as core buyers.
· Cost and power consumption optimization: 800G reduces unit bit cost by 30% and power consumption by 25% compared to 400G; 1.6T/CPO further reduces power consumption by 30%-50%.
III. Global cloud vendors' capital expenditure surges (direct driver of demand)
· North American top five cloud vendors + NVIDIA: Capital expenditure exceeding $400 billion in 2026, with AI infrastructure accounting for >60%; Google's expenditure of $175-185 billion (doubling), Meta purchasing 10-12 million 800G units.
· Domestic follow-up: ByteDance, Alibaba, and Tencent's capital expenditure of 163.7 billion yuan in 2026 (+92%), purchasing millions of 800G/1.6T units.
• Order Pre-positioning: Customers have locked in 2026-2027 capacity, with lead times of 800G (8-16 weeks) and 1.6T (over 20 weeks), resulting in firm prices.
IV. Irreversible Fiber-to-the-Home (Fiber to the Home) Replaces Copper (Rigid Replacement)
• Copper Cable Bottlenecks: DAC bandwidth ≤400G, distance ≤5m, high power consumption; 800G+ necessitates optical modules.
• Accelerated LPO/CPO Replacement: Mature silicon photonics and co-packaged optics (CPO) further replace traditional pluggable modules, reducing costs and power consumption.
V. Policy Dividends and Emerging Scenarios (Incremental Supplement)
• China's "East-to-West Computing" strategy: Driving the construction of computing power hubs and releasing massive demand for optical modules.
• Telecom Upgrades: 5G-A/6G and backbone network 400G/800G penetration, resulting in steady growth in the telecom market (approximately 13% market share).
• Emerging Applications: AI video generation, robot computing power, and high-performance computing (HPC) are becoming new growth drivers. VI. Supply Constraints (Supporting Prices and Market Sentiment)
· Shortage of Optical/Electrical Chips: Insufficient capacity for EML, TOSA/ROSA, and DSP, severely limiting high-end supply.
· Yield and Capacity Ramp-up: The yield of 800G silicon photonics is approximately 80%, and even lower for 1.6T. Long expansion cycles lead to a continued short-term supply-demand imbalance.
The high growth of the optical module market in 2026 is a result of the combined effects of AI demand, technological advancements, massive capital investment, the shift from copper to optical fiber, and policy support. Furthermore, supply constraints will further reinforce the "both volume and price increase" pattern, with high-speed products (800G/1.6T) becoming the core of growth.
I. Core Reason: Explosive Growth in AI Computing Power (The Biggest Engine)
· Dual Explosion in Training and Inference: Large model training requires tens of thousands of interconnected GPUs, exchanging terabytes of parameters per second; the inference side has an even higher demand density, 3-5 times that of the training side.
• Dramatic Increase in Single-Unit Usage:
* Traditional Servers: 1–2 100G/400G units
* AI Servers (H100/B100): 8–12 800G units
* NVIDIA GB200: 162 1.6T units
• Continuously Increasing Ratio: GPU: Optical Module ratio from 1:2 → 1:5 → 1:8 (ASIC cluster), optical module growth rate > GPU growth rate.
• Widespread Adoption of 10,000-GPU Clusters: Large-scale deployments by Meta, Google, and NVIDIA have driven 800G/1.6T orders to be booked until the end of the year.
II. Bandwidth Upgrade: 400G → 800G → 1.6T, a Two-Tier Leap in Three Years
• 800G Becomes Mainstream: Demand for 40 million+ units in 2026, Goldman Sachs raised its forecast to 33.5 million units (+58%), with Meta alone demanding 10-12 million units.
• 1.6T Year's Demand Surge: Driven by NVIDIA's GB200/300, demand is expected to reach 15-20 million units, a year-on-year increase of **600%**, with Google's order revised upwards to 15 million units.
• The Irreversible Shift from Copper to Fiber: Copper cables (DACs) have bandwidth ≤400G, distance ≤5m, and high power consumption; 800G+ necessitates optical modules, with LPO/CPO further replacing traditional pluggable modules.
III. Cloud Makers' Capital Expenditure Soars
• The world's four major cloud makers (Meta/Google/Amazon) + NVIDIA will have capital expenditures exceeding $400 billion in 2026, with AI infrastructure accounting for >60%.
• Domestic Follow-up: ByteDance, Alibaba, and Tencent are accelerating their intelligent computing centers, with millions of 800G/1.6T units procured.
• Regional Growth: Computing power projects in the Middle East (Saudi Humain) and Southeast Asia are being implemented, adding 1.6T orders. IV. Market Data (Intuitive Perspective)
· Overall Size: The global optical module market will reach $28.75 billion in 2026 (+25%), with AI accounting for 62.6%.
· High-Speed Market Share: 800G/1.6T will account for over 60% of shipments, with total demand exceeding 100 million units.
· Resilient Prices: 800G average price is $800–$1200, 1.6T is $2000–$3000, with slow price reductions and long delivery times due to supply shortages.
V. Why This Year (2026)?
· AI is transitioning from "experimentation" to "large-scale deployment": 2023–2025 will see investment in training, with a surge in inference demand in 2026, resulting in higher demand density.
· Mature Technology + Capacity Expansion: Improved yield rates for silicon photonics/LPO, ramp-up of 800G capacity, and commercialization of 1.6T will result in supply barely keeping up with demand.
• Order Pre-positioning: Customers are locking in 2026-2027 production capacity in advance, leading to shortages, price increases, and long lead times.
VI. Impact on You (Procurement/Selection)
• Low-end (1G/10G/40G): Stable demand, lower prices, and ample inventory.
• Mid-to-high-end (100G/400G): Steady growth, lead time 2-4 weeks.
• High-speed (800G/1.6T): Severe supply shortage, lead time 8-16 weeks, firm prices, priority given to large customers.
In short: The AI computing power revolution + bandwidth necessity + cloud companies investing heavily + fiber-to-the-home (FTTH) replacement of copper—these four factors combined have transformed optical modules from "small components" into "strategic necessities," resulting in explosive demand growth this year.
What is the market size for optical modules?
I. Global Market Size (2026)
· LightCounting / Most Industry Reports: $28.75-30 billion
o YoY: +25%～30%
· Goldman Sachs (Optimistic): $43-52 billion
· 2025 Base: Approximately $23 billion
II. Shipments (2026)
· Total Speeds: Approximately 420-450 million units
· High-Speed Mainstream:
o 800G: 34-40 million units
o 1.6T: 14-25 million units (only 5 million units in 2025)
III. Market Structure (Where is the Money Going?)
· AI Data Centers: $18-26 billion, accounting for 60%+
· Cloud Computing (Non-AI): Approximately 24%
· Telecommunications (5G/Transmission): Approximately 13%, basically stable
IV. China Market
· Size: Approximately $56-65 billion =RMB 100 million (≈80-95 billion USD)
· Global market share: 20%-25%
· Production capacity: Chinese manufacturers account for approximately 70% of the global market (almost monopolistic in the low-to-mid-range market, rapid growth in high-end 800G/1.6T)
What factors mainly influence the growth of the optical module market?
The core drivers of the optical module market growth can be summarized as: the essential demand for AI computing power, accelerated technological iteration, expanded capital expenditure of cloud manufacturers, the irreversible shift of fiber optics to copper, and policy support and emerging application scenarios. These five factors resonate, propelling the industry into a high-growth cycle of "both volume and price increases."
I. Explosive Growth in AI Computing Power (Core Engine, Contributing 60%+ to Incremental Growth)
· Dual-engine drive of training and inference: Large model parameters reach trillions, training clusters expand from thousands of cards to hundreds of thousands of cards, exchanging petabytes of data per second; optical interconnect is the only feasible solution.
•Surge in Single-Unit Usage:Traditional servers use 1-2 optical modules; AI servers (H100/B100) use 8-12 800G modules; NVIDIA GB200 servers reach 162 1.6T modules, with the GPU:optical module ratio increasing to 1:8-1:12.
•Cluster Scalability: Meta, Google, and NVIDIA are widely adopting 10,000-card-level clusters. AI server shipments are projected to reach 3.5 million units in 2026 (+40%).
II. Accelerated Technology Iteration (Core Factors: Simultaneous Increase in Quantity and Price)
Shortened Speed Upgrade Cycle: The iteration cycle from 400G to 800G to 1.6T has shortened from 3-4 years to 2 years. 3.2T is expected to be commercially available in 2027-2028.
High-Speed Penetration:
800G: Demand is projected at 40-50 million units in 2026, becoming mainstream.
o1.6T: Shipments of 15-25 million units in 2026 (+600%), with NVIDIA and Google as core buyers.
· Cost and power consumption optimization: 800G reduces unit bit cost by 30% and power consumption by 25% compared to 400G; 1.6T/CPO further reduces power consumption by 30%-50%.
III. Global cloud vendors' capital expenditure surges (direct driver of demand)
· North American top five cloud vendors + NVIDIA: Capital expenditure exceeding $400 billion in 2026, with AI infrastructure accounting for >60%; Google's expenditure of $175-185 billion (doubling), Meta purchasing 10-12 million 800G units.
· Domestic follow-up: ByteDance, Alibaba, and Tencent's capital expenditure of 163.7 billion yuan in 2026 (+92%), purchasing millions of 800G/1.6T units.
• Order Pre-positioning: Customers have locked in 2026-2027 capacity, with lead times of 800G (8-16 weeks) and 1.6T (over 20 weeks), resulting in firm prices.
IV. Irreversible Fiber-to-the-Home (Fiber to the Home) Replaces Copper (Rigid Replacement)
• Copper Cable Bottlenecks: DAC bandwidth ≤400G, distance ≤5m, high power consumption; 800G+ necessitates optical modules.
• Accelerated LPO/CPO Replacement: Mature silicon photonics and co-packaged optics (CPO) further replace traditional pluggable modules, reducing costs and power consumption.
V. Policy Dividends and Emerging Scenarios (Incremental Supplement)
• China's "East-to-West Computing" strategy: Driving the construction of computing power hubs and releasing massive demand for optical modules.
• Telecom Upgrades: 5G-A/6G and backbone network 400G/800G penetration, resulting in steady growth in the telecom market (approximately 13% market share).
• Emerging Applications: AI video generation, robot computing power, and high-performance computing (HPC) are becoming new growth drivers. VI. Supply Constraints (Supporting Prices and Market Sentiment)
· Shortage of Optical/Electrical Chips: Insufficient capacity for EML, TOSA/ROSA, and DSP, severely limiting high-end supply.
· Yield and Capacity Ramp-up: The yield of 800G silicon photonics is approximately 80%, and even lower for 1.6T. Long expansion cycles lead to a continued short-term supply-demand imbalance.
The high growth of the optical module market in 2026 is a result of the combined effects of AI demand, technological advancements, massive capital investment, the shift from copper to optical fiber, and policy support. Furthermore, supply constraints will further reinforce the "both volume and price increase" pattern, with high-speed products (800G/1.6T) becoming the core of growth.
Leave a Reply
Your email address will not be published.Required fields are marked. *