Every time someone asks an AI chatbot a question, thousands of processors inside a data center have to talk to each other quickly. Copper wiring works over short distances, but as data volumes climb, the industry has turned to light. Tiny lasers convert electrical signals into pulses of light, optical fiber carries them across the building, and a receiver turns them back into electrical signals at the other end. These parts, called optical transceivers, have become a quiet bottleneck of the AI boom.
Spending on this equipment is rising quickly. TrendForce, a market research firm, expects sales of optical transceivers built for AI to reach $26 billion this year, up more than 57% from $16.5 billion in 2025. Yole Group, a research and consulting firm based in Lyon, France, that studies the semiconductor and photonics industries, sees the wider transceiver market growing from $23.4 billion in 2025 to $112.3 billion by 2031.
Speed is the main story. Network capacity is measured in steps, and the current workhorse is the 800G module. The next step, 1.6T, is beginning to enter mass production. TrendForce describes this year and next as an important window for suppliers hoping to earn a place in the 1.6T supply chain, because early wins with large customers tend to decide who gets the volume later.Â
Building these parts is difficult. A transceiver needs lasers, modulators, detectors and electronics lined up with extreme precision. TrendForce has pointed to tight supply of certain laser chips and the precision required for optical alignment as limits on how fast production can grow. Power is another concern, because every watt spent moving data is a watt not spent on computing, and it adds heat.Â
The industry’s answer is integration. Photonic integrated circuits place many optical functions on a single chip. Optical engines combine those chips with supporting electronics in a compact unit, and light sources supply the laser light that powers them. Yole calls silicon photonics the leading platform for next-generation connections because it delivers more bandwidth while using less power, and it sees a natural path toward co-packaged optics, where the optical parts sit right beside the processor.Â
That is the market POET Technologies Inc. (NASDAQ: POET) is working to break into. The Canadian company designs photonic integrated circuits, light sources and optical engines built on its own packaging platform, the Optical Interposer, which joins electronic and photonic parts on one chip.Â
Today, POET reported third-quarter revenue of $1.03 million, up 80% from the prior quarter and 244% from a year earlier. That makes seven straight quarters of growth, though $1 million a quarter is a tiny slice of a market measured in tens of billions.Â
The net loss narrowed to $7.6 million, or $0.04 per share, from $11.3 million in the second quarter. A large part of that improvement came from a $6.4 million non-cash gain tied to revaluing warrants, an accounting entry unrelated to selling products. Without it, the loss would have been about $13.9 million.Â
The balance sheet tells a different story. POET ended the quarter with $774.3 million in cash and short-term investments, much of it raised through a $400 million financing completed in May. Operations used $12.7 million in cash during the quarter. With a market value of roughly $888 million, cash makes up close to 87% of what the stock market says the whole company is worth. Investors, in effect, are placing a modest value on the business itself.Â
Management expects fourth-quarter shipments to exceed everything it shipped in the first three quarters of the year combined. It also expects seven of its nine active customers to move into production with its engines in 2027, when its 1.6T engines are scheduled to ramp. The company has a new chief financial officer as well. Giovanni Pacelli, formerly of Silicon Labs, took the role on November 2nd, succeeding Thomas Mika after ten years in the job.Â
The open question is pace. A large cash cushion buys time to build capacity and win customers, but it does not guarantee meaningful revenue. The next few quarters should show whether the order book can turn POET into a real presence in AI networking.
