AI Infrastructure Boom: Are Optical Modules the Next Big Opportunity?

2026-05-22 | AI Data Centers , AI Infrastructure , Optical Communication , Optical Modules , Weekly Market Dive

Optical modules and silicon photonics powering high-speed AI data center infrastructure
Optical modules are emerging as a key bottleneck in the AI infrastructure supply chain.

As AI data centers expand rapidly, optical modules and optical chips are becoming one of the most important bottlenecks in the AI infrastructure supply chain. 

Memory stocks have been soaring. 

Have you gotten in yet?  

AI infrastructure spending continues to accelerate, with no signs of slowing down. Earnings reports from Microsoft, Google, Amazon, and Meta show that major cloud service providers continue to aggressively increase capital expenditure. 

Google even revealed that the backlog of its cloud business nearly doubled from the previous quarter. Management admitted that growth could have been even stronger if not for shortages in AI hardware supply. Other major cloud players have also expressed strong confidence in the resilience of current AI demand. 

Global cloud service provider capex rising from 2023 to 2026 driven by AI infrastructure spending
Global cloud service provider capital expenditure continues to rise as AI infrastructure investment accelerates.

At the same time, memory stocks have already experienced massive rallies. Micron Technology, for example, pulled back from a peak of around $818.67 to roughly $700, highlighting growing valuation concerns across the memory sector.  

The AI infrastructure cycle is still in full swing. 

But investors are now asking a new question: 

What comes after memory stocks? 

One answer may be optical modules and the broader optical communication supply chain. 

The market initially focused almost entirely on computing power. 

Then it discovered memory. 

Now, another bottleneck is becoming increasingly important: 

data transmission speed. 

Imagine a person with Einstein-level intelligence but terrible memory. 

He solves the problem, then instantly forgets it. 

No matter how intelligent he is, his performance slows dramatically. 

That is essentially what happened with AI infrastructure. 

The industry realized that memory chips were critical for supporting AI workloads. 

But computing power and memory alone are not enough. 

The “nervous system” matters too. 

If someone touches a hot stove and only reacts five seconds later, the issue is no longer intelligence. The issue is transmission speed. 

That is now becoming one of the biggest challenges for AI infrastructure. 

As AI models continue expanding, computing workloads grow exponentially. 

That means: 

  • more data transfer  
  • higher bandwidth demand  
  • greater power consumption  

Traditional copper cables are increasingly struggling to keep up. 

As electrical signals travel through copper, large amounts of energy are lost as heat due to resistance. At AI scale, this becomes highly inefficient. 

Light is simply faster and more efficient. 

That is why optical communication is becoming essential for next-generation AI data centers. 

Optical modules are key components that convert optical signals into electrical signals and vice versa. They are critical for: 

  • AI data centers  
  • cloud computing infrastructure  
  • telecom networks  
  • high-performance computing systems  

The global optical module market was valued at USD 11.54 billion in 2024 and is expected to reach USD 47.64 billion by 2035, with a compound annual growth rate of around 13.75%.  

The AI data center market itself is expected to grow even faster. 

Optical transceiver market forecast showing growth to USD 47.64 billion by 2035
The optical transceiver market is entering a high-growth cycle, supported by AI data center demand.

According to Precedence Research

  • AI data center market (2025): USD 17.54 billion  
  • AI data center market (2034): USD 165.73 billion  

This represents nearly a 10x expansion in less than a decade. 

AI infrastructure consumes enormous amounts of electricity. 

By 2030, global data center power consumption is expected to rise from around: 

  • 30 GW today 
    to more than  
  • 90 GW  

That is a threefold increase.  

As AI model parameters expand, the computing workload for a single task increases exponentially. 

More importantly, the industry still believes current AI systems are far from reaching true intelligence. Engineers continue pursuing larger and more powerful systems, with some future AI racks potentially targeting: 

  • 1 megawatt per single rack  

At that scale, traditional copper infrastructure becomes insufficient. 

This is where optical communication becomes critical. 

One of the most important components inside optical modules is the EML chip (electro-absorption modulated laser). 

These chips generate the high-speed laser signals required for data transmission. 

Current 800G optical modules typically require: 

  • 8 EML chips per module  

A single high-performance switch may contain: 

  • 64 ports  

That means one switch alone can consume: 

  • 512 EML chips  

And an AI data center equipped with 100 such switches would require more than: 

  • 50,000 EML chips  

Demand for 800G and higher-speed optical modules is expected to rise sharply: 

  • 2025: 24 million units  
  • 2026: nearly 63 million units  

This translates into annual demand for roughly: 

  • 200 million to 500 million EML chips globally  

The scale of this industry is becoming enormous. 

The market already recognizes the importance of optical modules. 

But not every part of the industry chain offers the same investment value. 

The highest barriers and strongest pricing power remain concentrated upstream. 

The upstream optical module supply chain includes: 

  • optical components  
  • optical and electrical chips  
  • packaging materials  
  • production equipment  

Among these, optical chips are the most critical. 

Silicon photonics supply chain diagram for AI data centers, optical modules, switches, and fiber networks
Silicon photonics is becoming a critical part of the AI data center supply chain.

Optical chips are mainly divided into: 

  • Indium phosphide (InP)  
  • Gallium arsenide (GaAs)  
  • Silicon photonics  

Each serves different communication requirements. 

  • InP → long-distance, high-speed communication  
  • GaAs → short-distance transmission  
  • Silicon photonics → large-scale integration using CMOS processes  

Lumentum remains one of the global leaders in high-end InP lasers and integrated optical systems, supported by strong IDM capabilities across chip design, wafer manufacturing, packaging, testing, optical modules, and optical switching solutions.  

However, after a massive rally, investors are increasingly asking: 

Where are the next opportunities beyond optical chips? 

Lumentum LITE share price chart showing a rally of more than 10x amid AI infrastructure demand
LITE shares have rallied more than 10x as optical communication demand accelerates.

When capital crowds into one segment, opportunities often emerge elsewhere. 

And right now, equipment may be the next area worth watching. 

Optical module manufacturing is evolving rapidly. 

The industry is shifting from: 

  • traditional plug-in designs  

toward: 

  • co-packaged optics (CPO)  
  • silicon photonic integration  

At the same time, production lines are becoming increasingly automated and intelligent. 

That means equipment standards are rising rapidly. 

Take the mounting process as an example. 

Tiny optical and electrical chips smaller than a fingernail must now be placed with extreme precision. 

Required accuracy has improved from: 

  • 10 microns previously 
    to around  
  • 3 microns today  

That is only a fraction of the diameter of a human hair.  

The chips are also extremely fragile. 

Excessive force or poor temperature control can easily damage them, making precise force application and temperature management critical technical challenges. 

Global leaders in this segment include: 

  • FOUR TECHNOS  
  • MRSI  
  • ASMPT  

After mounting comes the bonding process, which uses ultra-thin gold wires to connect chips for signal transmission. 

This process previously relied heavily on manual work, but increasing manufacturing standards now require far higher levels of precision and cleanliness. 

Even microscopic impurities can affect signal quality and transmission speed. 

Besi remains one of the global leaders in bonding equipment.  

Coupling is one of the most time-consuming processes in optical module manufacturing. 

A position offset of just: 

  • 1 μm  

can result in: 

  • up to 3 dB of optical loss  

As precision requirements continue rising, demand for advanced coupling systems is increasing rapidly. 

ficonTEC remains one of the major global players in this segment. 

Testing is also becoming increasingly difficult. 

As network speeds move toward: 

  • 800G  
  • 1.6T  

testing systems require extremely high bandwidth and advanced algorithms to capture tiny signal fluctuations. 

Keysight and Anritsu currently dominate this market, with the top two players holding roughly: 

  • 53% combined market share 

AI is the biggest opportunity of this era. Companies with true core competitiveness will ultimately survive and thrive through market cycles. 

The AI infrastructure cycle is no longer just about GPUs and memory. 

The next wave may increasingly focus on: 

  • optical communication  
  • optical chips  
  • photonics  
  • precision manufacturing equipment  

As AI systems become larger and more power intensive, data transmission efficiency becomes just as important as raw computing power. 

The market spent the past year chasing AI computing power. 

Then it discovered memory. 

Now, another bottleneck is emerging: 

the speed at which data moves. 

And in the AI era, the companies controlling that “nervous system” may become some of the most important players in the entire industry. 


Market DiveIconBrandElement

article-thumbnail

2026-05-22 | Market Dive

AI Infrastructure Boom: Are Optical Modules the Next Big Opportunity?

As AI data centers expand, optical modules, optical chips, and photonics may become the next major bottleneck and opportunity in AI infrastructure. 

article-thumbnail

2026-05-15 | Market Dive

April 2026 US Nonfarm Payrolls Beat Forecasts as Labor Market Weakens

April 2026 US nonfarm payrolls beat forecasts, but weak wage growth and low participation point to a cooling labor market and cautious Fed.

article-thumbnail

2026-05-08 | Market Dive

Magnificent 7 Earnings: Big Tech Splits Before Nvidia Test

Magnificent 7 earnings show Big Tech is no longer one trade. Traders are watching AI monetization, capex risk, stock dispersion and Nvidia’s May 20 report.