Watchlist beyond NVIDIA

NVIDIA is still the center of the AI trade. It sells the chips that power the boom, and it absorbs most of the attention. But the market has started to look past the obvious winner. The new watchlist is made of companies that sit one layer to the side of the GPU and still collect a toll every time a new AI cluster goes up. That shift matters because AI infrastructure is no longer just a story about compute. It is a story about moving light, moving data, designing chips, and wiring entire racks into something that behaves like one machine. That is why the names showing up beside NVIDIA are so different from one another. Coherent makes the optical parts that let data centers talk at high speed. Broadcom sells the networking silicon and custom chips that keep those clusters fed. Cadence sells the software used to design the chips in the first place. Lattice sells small low-power FPGAs that handle control and housekeeping tasks inside servers. IonQ is the outlier, less an AI infrastructure supplier than a market bet on the next computing platform after this one. Coherent is the cleanest example of the “picks and shovels” logic. The company’s fiscal 2025 revenue rose 23% to a record $5.81 billion, and management tied that growth directly to AI data centers. Its recent technology briefing at OFC, the telecom industry’s big optics conference, underscored where the excitement sits: datacom transceivers and the photonics needed to push more bandwidth through ever denser clusters. (sec.gov) Social posts calling Coherent “streaky” are not wrong. The company has had uneven quarters before. But the current setup is simpler than the chatter makes it sound. If AI racks need more optical links, Coherent has more room to grow. Lattice is a subtler version of the same idea. Its chips are not the stars of the rack. They are the small devices that manage sensors, power, security, and board-level control. That sounds minor until the rack gets crowded with expensive accelerators and high-speed networking. Then the number of helper chips rises fast. Lattice’s fourth-quarter 2025 results showed revenue up 24.2% year over year, with management pointing to strong data-center and physical-AI demand. Third-party summaries of the quarter describe server revenue up about 85% and dozens of Lattice devices going into a single AI rack, which explains why investors are modeling product lines at very different growth rates. (quartr.com) Broadcom sits at the other extreme. It is already huge, and AI is making it bigger. In fiscal Q1 2026, Broadcom reported $19.3 billion in revenue, up 29% year over year, while AI revenue jumped 106%. Management said second-quarter AI semiconductor revenue should reach $10.7 billion. That is why people describe Broadcom as an AI nervous system. It sells the custom accelerators and the networking silicon, especially Ethernet, that connect giant clusters into a usable whole. NVIDIA can dominate the processor socket and Broadcom can still thrive because hyperscalers want both more compute and more tailored interconnect. (broadcom.com) Cadence is even further upstream. It does not ship hardware into a rack at all. It sells the software that chip companies and systems firms use to build the next generation of AI silicon and advanced packaging. Cadence finished 2025 with $5.297 billion in revenue, up more than 14%, and entered 2026 with a record $7.8 billion backlog. The company has been pushing hard on AI-assisted design tools and on 3D-IC workflows, which are becoming more important as the industry stacks memory, logic, and chiplets into tighter packages. The “dominant EDA share” line from social media overstates the case if it implies monopoly. Synopsys is still a giant rival. But the core point is right. As AI chips get harder to design, the value of the design software rises with them. (investor.cadence.com) IonQ is the speculative name in the basket. It is not an AI infrastructure company in the same practical sense as Broadcom or Coherent. It is a quantum company that investors are trying to map onto the same appetite for next-generation compute. IonQ’s latest results explain the appeal. The company reported $130 million in 2025 revenue, up 202% year over year, guided to $235 million for 2026 at the midpoint, and said it ended 2025 with $3.3 billion in cash, cash equivalents, and investments after reshaping itself into what it calls a full-stack quantum platform company. That is real growth, not just a concept stock. But it is still a different kind of bet. Coherent, Lattice, Broadcom, and Cadence all profit from AI buildouts happening now. IonQ is being priced for a future that has not arrived yet. (investors.ionq.com)