The rapid expansion of artificial intelligence infrastructure is creating new pressure on materials beyond the better-known shortages involving advanced chips and memory. Indium phosphide (InP), a compound semiconductor material used in high-speed optical communications, is seeing sharp price increases as demand from AI data centres grows and supply remains concentrated in China.
According to industry sources cited in a recent Nomura report, the price of two-inch InP substrates rose to as much as 880 yuan ($130.54) per wafer in June, up from around 500 yuan at the end of 2025. Prices for three-inch wafers increased to about 3,200 yuan from approximately 1,800 yuan over the same period. The increases highlight how the AI infrastructure buildout is putting pressure on increasingly specialized parts of the technology supply chain.
Why Indium Phosphide Matters for AI Data Centres
Indium phosphide is a semiconductor material with properties that make it particularly useful for high-speed optical communications.
AI data centres increasingly depend on optical links to move enormous volumes of information between servers, accelerators and networking equipment.
As AI models become larger and data-centre clusters become more powerful, the amount of information that must move between computing systems also increases.
Conventional electrical connections can face limitations involving power consumption, signal loss and transmission distance.
Optical communication can address some of these challenges by transmitting information using light.
InP is used in components such as lasers and photonic devices that form part of these high-speed optical systems.
AI Is Creating a New Bottleneck
The AI boom initially created shortages and price pressure in areas such as GPUs, high-bandwidth memory and advanced packaging.
The supply-chain impact is now spreading into other components needed to build large AI computing systems.
Data-centre operators need more than processors.
A modern AI facility requires:
- GPUs and AI accelerators
- High-bandwidth memory
- Advanced networking
- Optical transceivers
- Fibre-optic connections
- Power systems
- Cooling equipment
- Storage
- Semiconductor components
- Photonic devices
As investment in AI infrastructure accelerates, demand for each of these layers can increase.
This is creating new potential bottlenecks in materials and components that previously received less attention.
InP Prices Have Climbed Sharply
The increase in InP substrate prices illustrates the speed of the supply squeeze.
Two-inch substrates reportedly rose from around 500 yuan per wafer at the end of 2025 to as much as 880 yuan in June.
That represents an increase of approximately 76%.
Three-inch wafers increased from roughly 1,800 yuan to around 3,200 yuan during the same period, an increase of about 78%.
| InP Substrate | End-2025 Price | June Price | Approx. Increase |
|---|---|---|---|
| Two-inch wafer | 500 yuan | 880 yuan | ~76% |
| Three-inch wafer | 1,800 yuan | 3,200 yuan | ~78% |
The price movement shows that the supply constraint is already affecting the upstream materials used by optical component manufacturers.
China Has a Major Role in InP Supply
China is a major producer of indium phosphide materials and related components.
That concentration has become increasingly important as global demand for optical infrastructure rises.
When supply chains depend heavily on a small number of producers or a particular geographic region, sudden increases in demand can quickly create shortages.
The current situation also highlights the broader strategic importance of China’s position in semiconductor materials.
The country is a major supplier of numerous materials used across the global electronics industry.
Optical Networks Are Becoming More Important
The growth of AI computing is changing data-centre networking requirements.
Traditional data-centre workloads generally involve significant amounts of communication, but AI clusters can generate much larger and more frequent data transfers.
Thousands of accelerators may need to communicate with each other during the training and inference of large models.
That makes high-speed networking a critical part of AI infrastructure.
Optical communication is increasingly being used to handle these requirements.
AI Clusters Need Faster Connections
Training a large AI model involves distributing workloads across many processors.
The processors constantly exchange information.
If networking becomes a bottleneck, expensive GPUs can remain underutilized.
Data-centre operators therefore have a strong incentive to increase network bandwidth and reduce communication latency.
This is creating demand for faster optical transceivers and other photonic components.
As these systems become more sophisticated, the materials used to manufacture optical components can become strategically important.
InP Competes With Other Semiconductor Materials
Indium phosphide is not the only material used in advanced optical communication.
Silicon photonics is another major technology being developed for high-speed data transmission.
Silicon-based photonics can benefit from the existing semiconductor manufacturing ecosystem and offers potential advantages in scalability.
However, InP has important optical properties that make it valuable for lasers and other photonic applications.
The two technologies can therefore coexist rather than one immediately replacing the other.
Silicon Photonics Is Also Expanding
Major technology companies and semiconductor manufacturers are investing heavily in silicon photonics.
The technology is increasingly being considered for connecting AI accelerators and other high-performance computing systems.
As AI data centres grow, demand for optical communication is expected to increase regardless of which specific photonic technology becomes dominant.
This means the broader optical-component supply chain could remain under pressure even as companies develop alternatives to InP.
The Supply Crunch Could Raise Data-Centre Costs
Higher material prices can eventually affect the cost of optical components.
Optical transceivers and related networking hardware are only one part of the overall cost of an AI data centre, so an increase in InP prices may not have a dramatic impact on the total project cost.
However, if multiple materials experience simultaneous shortages, the combined effect could become significant.
Data-centre construction already requires large investments in chips, power infrastructure, cooling and networking.
Higher component costs could add another layer of pressure.
AI Infrastructure Spending Remains Strong
Technology companies are continuing to invest heavily in AI computing capacity.
Hyperscalers and other major infrastructure providers are building new data centres and expanding existing facilities to support AI services.
The investment cycle is increasing demand for GPUs, networking equipment and power infrastructure.
As the industry scales, supply shortages are increasingly moving beyond the most visible components.
The InP market is one example of this broader phenomenon.
Specialized Materials Can Become Strategic Assets
The situation demonstrates why semiconductor supply chains extend far beyond chip fabrication.
Advanced computing systems depend on a large number of specialized materials.
These include:
- Compound semiconductors
- Rare metals
- Wafer materials
- Packaging substrates
- Optical materials
- High-performance chemicals
- Specialty gases
- Advanced electronic components
A shortage in any one of these areas can potentially slow production elsewhere in the chain.
This makes upstream material availability increasingly important to AI infrastructure planning.
China Supply Risks Add a Geopolitical Dimension
China’s role in semiconductor materials has also become more significant amid growing technology tensions between China and the United States.
Governments are increasingly focused on securing domestic or allied supply chains for strategically important materials.
The concentration of InP production in China could therefore encourage companies and governments to look for alternative sources.
Building new production capacity outside China, however, would require significant investment and time.
Prices Could Remain Elevated if Demand Continues
The future direction of InP prices will depend on the balance between demand and new supply.
If AI data-centre construction continues at its current pace, demand for optical components could remain strong.
Manufacturers may respond by expanding production capacity.
However, new semiconductor-material facilities cannot necessarily be built quickly.
Production involves specialized equipment, technical expertise and quality-control processes.
That could allow supply constraints to persist for some time.
Optical Component Makers Face Margin Pressure
Higher substrate prices could put pressure on companies that manufacture optical devices.
If material costs rise faster than selling prices, manufacturers could see their margins decline.
Companies with long-term supply agreements or greater purchasing power may be better positioned to manage the increase.
Smaller manufacturers could face greater difficulties if they cannot secure sufficient material at predictable prices.
Higher Prices Could Accelerate Substitution
Sustained price increases often encourage companies to seek alternatives.
Optical component manufacturers could increase investment in silicon photonics or other technologies if InP becomes significantly more expensive or difficult to obtain.
However, replacing a semiconductor material is not as simple as switching suppliers.
Different materials have different performance characteristics and manufacturing requirements.
Any transition would therefore take time.
AI Is Expanding the Definition of a Chip Shortage
The current situation suggests that AI-related supply constraints are becoming more complex.
The industry is no longer focused only on whether enough advanced GPUs can be produced.
Companies now need to secure memory, networking, optical components, power equipment and specialized materials.
A shortage in any one area can potentially affect the deployment schedule of an entire data centre.
This makes supply-chain planning increasingly important for AI infrastructure companies.
Data Centres Are Becoming More Network-Intensive
As AI clusters become larger, communication between processors becomes a larger part of overall system performance.
This is increasing demand for high-bandwidth optical links.
The trend is likely to continue as AI systems use larger models and more distributed computing architectures.
The result could be sustained growth in the market for optical transceivers, lasers and photonic components.
That would keep materials such as InP strategically relevant.
The Opportunity for Material Suppliers
The supply crunch also creates an opportunity for companies capable of expanding InP production.
Higher prices can improve the economics of investing in new capacity.
Manufacturers outside China could also see an opportunity to establish alternative supply chains.
Governments may support such projects because semiconductor materials are increasingly viewed as strategically important.
Over time, this could lead to a more geographically diversified InP supply network.
The Broader Semiconductor Supply Chain Is Changing
The InP situation is part of a much larger transformation in the semiconductor industry.
AI has changed demand patterns across several segments.
Manufacturers are prioritizing high-performance applications, while companies are securing capacity further upstream.
This can create competition for materials and manufacturing resources that were previously sufficient for conventional electronics.
The result is a supply chain increasingly shaped by AI infrastructure demand.
What Companies Will Watch
Technology and data-centre companies will likely monitor several factors closely:
- InP substrate prices
- Optical transceiver demand
- New InP production capacity
- Silicon photonics adoption
- AI data-centre construction
- Chinese export and supply policies
- Alternative material development
- Optical component pricing
- Semiconductor supply-chain diversification
These factors will help determine whether the current price surge becomes temporary or develops into a longer-term supply constraint.
Key Facts at a Glance
| Metric | Details |
|---|---|
| Material | Indium phosphide (InP) |
| Primary relevance | High-speed optical communications |
| Two-inch wafer price, end-2025 | ~500 yuan |
| Two-inch wafer price, June | Up to 880 yuan |
| Three-inch wafer price, end-2025 | ~1,800 yuan |
| Three-inch wafer price, June | ~3,200 yuan |
| Main supply factor | China production concentration |
| Major demand driver | AI data-centre expansion |
| Key application | Optical communication components |
| Alternative technology | Silicon photonics |
Infographic: How AI Demand Is Pushing Up InP Prices
AI BOOM
↓
MORE AI DATA CENTRES
↓
MORE GPUs + ACCELERATORS
↓
MORE DATA MOVEMENT
↓
HIGHER BANDWIDTH REQUIREMENTS
↓
MORE OPTICAL NETWORKING
↓
GREATER DEMAND FOR
LASERS + PHOTONIC COMPONENTS
↓
INDIUM PHOSPHIDE
↓
SUPPLY CONCENTRATED IN CHINA
↓
SUPPLY PRESSURE
↓
TWO-INCH InP
500 YUAN
↓
UP TO 880 YUAN
↓
THREE-INCH InP
1,800 YUAN
↓
UP TO 3,200 YUAN
↓
POTENTIAL IMPACT
HIGHER COMPONENT COSTS
+
SUPPLY-CHAIN DIVERSIFICATION
+
GREATER INTEREST IN SILICON PHOTONICS
The Bigger Picture
The sharp rise in indium phosphide prices shows how the AI infrastructure boom is spreading pressure beyond GPUs, memory and advanced packaging. InP is an important material for optical communication components that help move data through increasingly powerful AI data centres. With two-inch substrate prices reportedly rising to as much as 880 yuan from around 500 yuan at the end of 2025, and three-inch wafers reaching about 3,200 yuan from roughly 1,800 yuan, the material is emerging as another potential AI supply-chain bottleneck.
The development also highlights the strategic importance of semiconductor materials and China’s role in global supply chains. As AI clusters require faster and more energy-efficient communication, demand for optical technologies is likely to continue growing. That could encourage investment in additional InP capacity, alternative suppliers and technologies such as silicon photonics. The broader lesson for AI infrastructure companies is that securing processors alone is no longer enough; access to the specialized materials and components surrounding those processors can also determine how quickly new computing capacity can be deployed.
Looking Ahead
The immediate focus will be on whether InP producers can increase supply quickly enough to meet growing demand from optical-component manufacturers. If AI data-centre construction continues expanding rapidly, material suppliers could face sustained pressure, while manufacturers of optical components may need to negotiate longer-term supply arrangements to secure capacity. Higher prices could also encourage new investment in production facilities outside China.
Over the longer term, the growth of AI networking could reshape the market for compound semiconductors and photonic technologies. InP is likely to remain important for high-performance optical applications, while silicon photonics and other technologies could gain ground as data-centre operators seek greater scale and lower costs. The emerging InP shortage is therefore another example of how the AI boom is transforming supply chains well beyond the processors at the heart of today’s data centres.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
