Photonics Computing: What It Means and Why It Matters

Photonics computing is the use of light to process and transmit data inside computing systems. In the context of AI, it means replacing the copper connections between GPUs with optical connections that use laser beams to move information. Jason Bodner’s Accelerated AI presentation makes the case that this transition is the key breakthrough that will enable the next phase of AI growth.

The concept is distinct from optical computing, which attempts to use light for computation itself. Photonics computing, as Bodner describes it, is primarily about data transmission: moving information between processors at the speed of light rather than the speed of electrons in copper. This is an important distinction because data transmission is the bottleneck in modern AI systems, not computation. The GPUs can process data fast enough. The problem is getting data to and from the GPUs quickly enough to keep them busy.

The Problem with Copper

Every time you ask ChatGPT a question, it burns roughly 10 times more electricity than a Google search. A single AI data center can consume as much power as 100,000 homes. The problem is not the chips themselves but the connections between them. AI systems require thousands of GPUs to work together simultaneously, and the data moves between those chips over copper wires.

Copper has served the computing industry well for decades, but it has physical limits. Electrons in copper travel at less than 1 percent of the speed of light. They generate heat, which requires additional power for cooling. They consume significant amounts of electricity. And they cannot keep up with the bandwidth demands of modern AI workloads.

Bodner frames this as Phase 2 of his Acceleration Curve. Phase 1 was the initial AI boom. Phase 2 is the wall: the copper bottleneck. Phase 3 is the breakthrough: photonics. For more on this framework, see our copper to fiber article.

How Photonics Computing Works

Photonics computing uses several key components:

Optical transceivers convert electrical signals from GPUs into light pulses. Broadcom debuted the industry’s first 400G optical chip, which can handle 400 gigabits per second of data throughput. That is enough bandwidth to transfer a high-definition movie in a fraction of a second. For more on this component, see our photonic chips article.

Fiber optic cables carry the light signals between components. Unlike copper wires, fiber optic cables do not generate heat and barely consume power. They can carry far more data over longer distances with minimal signal loss.

Photonic integrated circuits combine multiple optical components on a single silicon chip. These are fabricated using standard semiconductor manufacturing processes, which means they can be produced at scale using existing fabrication facilities.

The result is a system where data moves between GPUs at 124,000 miles per second, the speed of light, instead of less than 1 percent of that speed for electrons in copper. Bodner says this makes AI “100 times faster and 100 times more energy efficient.” The physics supports the directional claim, though the actual improvement depends on the specific implementation and workload.

The Investment Opportunity

Bodner reveals Broadcom (AVGO) for free as the most direct large-cap play on photonics computing. Broadcom’s AI revenue was $10.8 billion last quarter, up 143 percent year over year. The company designs custom AI silicon, builds networking switches, and makes the 400G optical chip that enables the transition. For more on Broadcom, see our Broadcom stock article.

The three bonus picks behind the paywall offer higher-upside exposure. One is a pure photonics play that moves data as light between data centers across cities and states, with a new product that cuts networking power use by 70 percent. Another builds parts that turn electrical signals to light inside AI clusters and has a multi-billion partnership with Nvidia. Jensen Huang calls it “the next trillion-dollar company.” For more on the full investment thesis, see our Accelerated AI review.

The Structural Evidence

The investment in photonics computing is substantial and comes from the largest players in the industry:

  • Nvidia has invested over $7 billion in photonics companies and partnered with Corning to build three optical factories in the United States.
  • AMD is building a $280 million photonics research hub.
  • Ayar Labs raised $500 million from ARK Invest and Sequoia Capital.
  • Bill Gates has personally invested over $200 million in two photonics companies.
  • The Optical Interconnect Alliance includes Nvidia, AMD, Broadcom, Microsoft, Meta, and OpenAI.

These investments confirm that photonics computing is not one presenter’s opinion but an industry-wide strategic priority. For more on the alliance, see our optical interconnect article.

Considerations

Photonics computing is a real technology with real applications in AI data centers. The transition is already underway. The risks are execution and timing. The technology needs to scale from current implementations to the bandwidth and reliability levels required for mission-critical AI workloads. Broadcom is already shipping 400G optical chips, but the broader transition will take years.

Bodner’s track record, including Nvidia at $4.50, Super Micro before ChatGPT, and Vertiv before Goldman Sachs, suggests he has a genuine ability to identify infrastructure shifts. But he acknowledges: “I don’t always get it right.” For more on his background, see our Jason Bodner profile.

If you want to explore the full thesis, you can access the Accelerated AI presentation through Brownstone Research.

This is not financial advice. Always do your own research before investing.