Stocks In AI Bottleneck Soar 1,000%. Now Nvidia Is Demanding 20X More Supply

The AI boom has already burned through one shortage after another. First it was GPUs. Then high-bandwidth memory. Then power infrastructure, transformers, cooling systems, and even land near major electric grids. Yet as hyperscalers continue pouring hundreds of billions into AI data centers, another chokepoint is moving front and center — optics.

That may sound niche, but it sits at the heart of how AI clusters communicate. Without optical networking, the latest AI systems simply cannot move data fast enough between thousands of GPUs. And according to a new report from Rosenblatt Securities, Nvidia (NASDAQ:NVDA) recently asked suppliers to increase indium phosphide (InP) laser capacity by 20x through 2030. Suppliers reportedly countered with a still-massive 12x increase instead.

That gap tells investors something important: demand expectations are enormous, but suppliers remain cautious about overbuilding.

Why Optics Became the AI Industry’s New Pressure Point

AI models are growing too large for traditional networking. Training clusters increasingly require tens of thousands of GPUs connected with ultra-fast optical interconnects so data can move with minimal latency. Copper cables simply cannot keep up at scale.

Rosenblatt estimates the InP optics market could grow from roughly $1.9 billion in 2025 to $22.75 billion by 2030 — a nearly 12-fold increase in annual laser revenue capacity.

Here’s what the firm projects for major suppliers by 2030:

^{Source: Rosenblatt Securities industry update, May 2026.}

Investors have noticed. Over the past year:

• Lumentum climbed 1,048%
• AXT (NASDAQ:AXTI) surged 7,421%
• Lightwave Logic (NASDAQ:LWLG) gained 1,073%
• Applied Optoelectronics advanced 872%, even after a 26% pullback

Those gains look wild at first glance. Surprisingly, they may reflect how unprepared the industry was for AI networking demand rather than pure speculation.

Why Nvidia Wants So Much More Supply

The shift toward co-packaged optics, or CPO, is driving the urgency. CPO moves optical connections closer to the AI chip itself, improving speed while reducing power consumption — a growing problem as racks consume more electricity.

For Nvidia, the math is straightforward. More GPUs require exponentially more bandwidth. A single AI cluster with 100,000 accelerators can create networking bottlenecks that erase the benefit of adding more compute.

That is why Rosenblatt says Nvidia pushed for a 20x increase in optics supply. Regardless of how you look at it, Nvidia appears worried that networking constraints could slow AI deployment before GPU demand slows.

Granted, suppliers have reasons to resist committing fully. Building InP fabrication capacity is expensive, technically difficult, and risky if demand cools. Rosenblatt noted suppliers remain concerned about repeating past telecom cycles where excess optical capacity crushed margins.

In short, suppliers remember the last boom-and-bust cycle.

Which Stocks Rosenblatt Likes — And Which May Lag

Rosenblatt appears most bullish on Coherent and Lumentum. The firm said Coherent could benefit from revenue acceleration and rising production on six-inch wafers, which improve manufacturing efficiency. Rosenblatt also expects Lumentum to expand toward nearly 10% transceiver market share while benefiting from the shift into CPO lasers.

That said, the firm was less enthusiastic about Applied Optoelectronics. Rosenblatt argued the stock’s surge already prices in much of the expected AI demand increase. It also warned Applied Opto’s 1.6T transceiver business may prove weaker than investors expect.

Meanwhile, AXT remains more speculative. The company supplies InP substrates rather than finished optical products, making it more leveraged to raw material demand but also more volatile.

Key Takeaway

When all is said and done, optics may become the next defining AI bottleneck. GPUs grab the headlines, but they cannot operate efficiently without high-speed optical networking tying entire AI clusters together.

Nvidia’s reported request for 20x more supply signals the industry sees years of demand ahead. Yet suppliers agreeing to only 12x tells investors something equally important: management teams still fear overexpansion.

For sharp investors, that creates a balancing act. Companies like Coherent and Lumentum appear positioned to benefit from real infrastructure demand backed by hyperscaler spending. More speculative names such as AXT and Lightwave Logic could deliver larger upside — but with much higher risk if AI spending cools.

In any case, the optics market is no longer a side story in AI. It is becoming the industry’s foundational infrastructure.