The rapid rise of artificial intelligence has brought unprecedented attention to GPUs, HBM memory, advanced packaging, and computing power. However, beneath these technologies lies a fundamental challenge that is becoming increasingly important:

How can massive volumes of data be transferred efficiently, at high speed, and with minimal power consumption?

Modern AI infrastructure is not built solely on powerful processors. Large-scale AI data centers depend on extensive communication networks that move enormous amounts of information between servers, accelerators, storage systems, and network switches. As AI workloads continue to grow, the demand for higher-bandwidth optical links and lower energy consumption per transmitted bit is accelerating.

In the AI era, the ability to process data is important—but the ability to move data efficiently may become equally critical.

The Growing Pressure on AI Optical Interconnects

Future AI clusters require:

• Higher data transmission rates
• More optical links per system
• Lower power consumption
• Reduced cost per transmitted bit
• Greater scalability

To address these requirements, the photonics industry is increasingly turning toward photonic integration, where multiple optical functions are integrated onto a single chip platform.

An ideal Photonic Integrated Circuit (PIC) must simultaneously achieve:

1. Mass production capability
2. Ultra-low optical loss
3. Efficient electro-optic control

Achieving only one or two of these requirements is insufficient. A practical optical interconnect platform must combine all three while maintaining manufacturability and reliability.

Within these systems, optical modulators play a crucial role. They serve as the interface between electronic signals and optical carriers, directly impacting transmission speed, energy efficiency, and overall system performance.

In other words, the future success of photonic chips depends not only on guiding light efficiently but also on modulating it effectively.

Why Thin-Film Lithium Niobate Matters

Existing photonic platforms each have strengths and limitations.

Silicon Photonics

Silicon photonics offers mature semiconductor manufacturing infrastructure and excellent scalability. However, modulation mechanisms based on carrier injection or depletion can introduce optical losses and performance trade-offs.

Silicon Nitride

Silicon nitride provides exceptionally low optical loss and is highly suitable for passive photonic circuits. However, it lacks a strong intrinsic electro-optic effect, limiting its ability to perform efficient high-speed modulation.

The Advantage of Lithium Niobate

Lithium niobate possesses a naturally strong Pockels effect, enabling direct and highly efficient electro-optic modulation.

Key material advantages include:

These characteristics make lithium niobate particularly attractive for high-speed optical communication systems requiring low insertion loss and wide modulation bandwidth.

From Excellent Material to Scalable Platform

Historically, lithium niobate's primary limitation was integration.

Conventional lithium niobate modulators often featured:

• Device lengths approaching 10 cm
• High manufacturing costs
• Significant power consumption
• Dependence on external electrical amplifiers

Such characteristics made large-scale deployment in AI data centers challenging.

The emergence of Thin-Film Lithium Niobate on Insulator (LNOI) has fundamentally changed this situation.

Advances in nanofabrication and wafer processing have enabled:

• Wafer-scale manufacturing
• UV stepper lithography processes
• Highly reproducible fabrication
• Dense photonic integration

Today, state-of-the-art LNOI platforms can achieve:

• Waveguide losses as low as 0.05 dB/cm
• Quality factors (Q) around 6,000,000
• Integrated modulators, filters, resonators, and frequency comb generators

This transformation has elevated lithium niobate from a high-performance material into a complete photonic integration platform.

Enabling Next-Generation Optical Modulators

One of the most promising achievements of LNOI technology is its electro-optic modulator performance.

Compared with traditional lithium niobate Mach-Zehnder modulators (MZMs), LNOI devices offer substantially improved efficiency.

Typical performance includes:

A 2 cm LNOI modulator can operate directly at approximately 1 V CMOS drive levels, potentially eliminating the need for dedicated electrical amplifiers.

For AI optical interconnects, this translates into:

• Lower system power consumption
• Simpler packaging
• Reduced infrastructure costs
• Higher overall efficiency

Frequency Combs and WDM Integration

Beyond modulation, future optical networks require advanced wavelength management technologies.

Wavelength Division Multiplexing (WDM) enables multiple data channels to be transmitted simultaneously over a single optical fiber, dramatically increasing bandwidth.

To support next-generation WDM systems, ideal optical frequency combs should provide:

• Flat spectral output
• High optical power
• Precise frequency spacing
• Chip-scale integration

LNOI has demonstrated remarkable capabilities in this area.

Recent demonstrations have achieved:

• 430 comb lines across an 85 nm bandwidth
• 25 GHz channel spacing
• RF power consumption of approximately 740 mW

Other highly efficient electro-optic comb architectures have generated:

• 47 comb lines
• 25 GHz spacing
• RF power consumption as low as 0.6 W

These developments indicate that LNOI is capable of supporting highly scalable optical communication architectures.

Moving Beyond the Laboratory

Perhaps the most important milestone is that LNOI is no longer limited to laboratory demonstrations.

Real-world transmission experiments have validated its potential for practical deployment.

Using a flat-top 50 GHz electro-optic frequency comb and WDM technology, researchers demonstrated:

• 53 km fiber transmission distance
• 6.48 Tbps aggregate data rate

Such results suggest that LNOI is rapidly progressing from individual device innovation toward system-level optical interconnect solutions.

Conclusion

Thin-Film Lithium Niobate represents far more than a smaller modulator or a lower-loss waveguide.

It brings together several critical capabilities within a single platform:

• Ultra-low optical loss
• Intrinsic electro-optic modulation
• High-bandwidth signal processing
• Wafer-scale manufacturing
• Integrated frequency comb generation
• Advanced WDM functionality

These capabilities directly address the most pressing challenges facing AI data center infrastructure:

• Increasing bandwidth requirements
• Lower energy consumption
• Reduced cost per transmitted bit
• Higher integration density

As AI systems continue to scale, future performance may depend not only on computational power but also on how efficiently data can move between electrical and optical domains.

For this reason, Thin-Film Lithium Niobate is increasingly viewed as one of the most promising foundational platforms for next-generation AI optical interconnects.