When you send a text, stream a video, or ask an AI chatbot a question, it can feel like the information appears instantly. In reality, every digital interaction depends on data moving across networks, servers, and chips at extraordinary speed.
For years, much of this movement happened through electrical signals traveling over copper wires. Copper is familiar, reliable, and cost-effective over short distances. But AI is changing the scale of the challenge. Modern AI systems require massive amounts of data to move between chips, servers, and data centers. As those systems grow larger and more complex, the speed and efficiency of moving data is becoming just as important as the chips doing the processing.
That is where photonics and optical technology are becoming increasingly vital.
Photonics refers to using light to transmit information. In data centers, that often means using optical links, lasers, fiber, and related components to move data faster and farther than traditional electrical connections.
A simple way to picture it is with two pipes:
In the first pipe, imagine sending information by firing tiny steel balls. As you try to send them faster, they collide more with the pipe, lose energy, and create heat.
Put another pipe nearby and the vibrations from the first can leak into this second pipe, causing interference.
That is not exactly how copper works, but it captures the problem: Electrical signals face resistance, heat, signal loss and interference as speeds rise.1
Now instead, imagine sending a laser through a glass tube. The light is guided down the tube, can travel long distances with relatively low loss, and is far less affected by neighboring signals.2
This matters because AI infrastructure is not just about a single chip or server. It is about connecting thousands of chips so they can work together as one system.
AI models are trained and run across large clusters of processors. These processors constantly exchange information with one another. As models grow larger, they require more coordination. The more coordination required, the more data needs to move quickly across the system.
Today, GPUs get most of the attention because they perform the calculations that power AI. But those calculations are only useful if data can reach the right processor at the right time. If data movement slows down, expensive chips can sit idle. That creates a bottleneck in the system.
In traditional computing, performance was often limited by processing power. In AI, it is increasingly limited by how quickly data can move between chips, servers, and data centers. The bottleneck is shifting from computation alone to the movement of data around the system.
Copper is not going away. For short distances, especially inside racks, it can still be the lowest-cost and most practical solution. But as data rates increase, copper faces physical limits that become harder to manage.
At higher speeds, electrical signals weaken more quickly in copper cables. Effects like resistance and electromagnetic interference can reduce signal quality. Maintaining performance often requires more power to amplify the signal, which creates extra heat and complexity. Copper cables are also relatively thick, making them harder to fit into densely packed systems.
Optical technology solves many of these problems by using light instead of electricity to transmit data. Fiber-optic cables can carry much more data, lose less signal over distance, and are largely immune to electromagnetic interference. They also use less power for long-distance connections and take up less space, allowing for higher cable density and better airflow.
The industry is no longer debating whether optics will play a larger role—it is deciding how quickly the transition will happen. Copper will remain useful for short-reach connections, but its limits grow as bandwidth demands rise. As AI clusters continue to scale, light is increasingly becoming the preferred medium for moving data.
The shift to optical technology did not begin with AI. For decades, fiber has moved steadily closer to where data is created and consumed. Early fiber systems were used for long-distance communications, with later advances allowing multiple wavelengths of light to travel over the same fiber and substantially increase capacity.2 Over time, optics expanded from long-haul networks into internet backbones, mobile networks, homes, and now deeper into the data center.
The Next Frontier for Optics Is Inside the Data Center
What began in long-distance networks is now moving into AI infrastructure
That inward movement is important. Light tends to win the longest distances first, then gradually moves closer to the compute layer as bandwidth requirements rise and the economics improve. The AI buildout may represent the next major step in that progression, bringing optics from the network edge of the data center toward the rack, the board and eventually the chip package itself.
We view the ability to move data quickly between chips, servers, and data centers as a critical bottleneck in the AI infrastructure buildout.
For investors looking to access this long-term secular opportunity, the Tema Photonics & Optical ETF (LAZR) offers targeted, pure-play exposure to innovators in photonics and optical technology through a high-conviction, actively managed portfolio.