Deep Dive: Understanding the AI Trade

ChatGPT launched in November 2022 and hit 100 million users within months, kicking off a whirlwind AI trade — from ’s massive rally to more recent turbulence in software stocks.

Despite the flurry of headlines this week, one item stood out: Nvidia investing $2 billion in and $2 billion in . That’s a reminder that the AI buildout isn’t just chips — it’s also “optics and photonics,” where bandwidth demands are exploding inside data centers. COHR and LITE are up 37.5% and 77% year-to-date, respectively.

Understanding the AI Trade

Chips were the first, most obvious winners — Nvidia, , , , and the toolmakers like , , , and . But the opportunity set is broader.

The AI infrastructure trade is really a “picks-and-shovels” buildout: compute, networking, and the physical systems required to power and cool dense AI racks. Beyond silicon, that includes networking (Broadcom, , and ), optics (, Coherent, Lumentum), and increasingly power and thermal enablers like and as electricity and cooling become gating factors. Datacenters are the “housing” layer for all of that gear — the physical campuses and leased capacity that determine how quickly new AI compute can actually be deployed. That’s where operators like and fit in.

Finally, there’s the “software infrastructure” layer: Firms like and help enterprises monitor and manage data and AI workloads in production, while companies like and operate on the cybersecurity side of the expansion as AI increases complexity, automation, and the attack surface.

Diving Deeper — Sentiment & Risk

We’ve seen how sentiment can ebb and flow, particularly within this group. Take Nvidia for instance. Despite beating on earnings and revenue estimates and providing strong guidance, shares have moved lower since earnings. Its valuation now hovers near multi-year lows. In other words, sentiment has dampened despite a strong fundamental backdrop.
A key dynamic in the AI infrastructure trade is bottleneck rotation. It began with scarce accelerators and advanced-node capacity, then shifted to networking and optics as moving data between chips became the limiter. Now power delivery and cooling are increasingly becoming key factors, with grid constraints and long-lead electrical and thermal gear dictating how fast new capacity can come online.

In other words, these dynamics can shift quickly, with sentiment flipping from hot to cold — and back again.

Risks

Sentiment swings are one thing to be cognizant of, but there are other risks too. Some include: timing risk on “next leg” themes, power/permitting delays that push revenue recognition out even when demand is intact, and concentration and pricing risk in the enablers, where a small set of customers can re-source, insource, or renegotiate as volumes scale. That doesn’t even take into account bigger-picture macro risks.

The Bottom Line

AI is still a multi-year buildout with real, investable opportunities across the stack as spending expands from compute into networking, optics, and power infrastructure. The trade won’t be linear — bottlenecks and sentiment will rotate — but the direction of travel is clear, and the winners tend to emerge where constraints are tightest and demand is most durable. Still, investors should do their own research to determine which risk-reward setup best fits their style.

***
Disclaimer: Please note that due to market volatility, some of the prices may have already been reached and scenarios played out. Content, research, tools, and stock symbols displayed are for educational purposes only and do not imply a recommendation or solicitation to engage in any specific investment strategy. All investments involve risk, losses may exceed the amount of principal invested, and past performance does not guarantee future results.