The AI Trade Is Healthy and Broadening
Nvidia's (NVDA) latest report showed the AI trade is in good order and keeps moving forward. Growth ran at 100% or higher year over year, with a strong forecast for the year ahead. Jensen Huang's message: hyperscaler AI spending is set to make money, and there are many bottlenecks in the system. Those bottlenecks mean many different ways to get exposure to the trade.
The trade began narrow - just semiconductors, broad chip ETFs, or single names like Nvidia (NVDA), AMD (AMD), and Micron (MU). Investors now see distinct layers. An AI chip is a fast engine, but the engine needs fuel. The fuel is high bandwidth memory - the memory chips. Data also has to move from place to place, which is the data highway: photonics, laser, and optics. AI models need large amounts of power and compute to run without breaking down, which requires capacitors, described as the plugs that act like running water inside the AI system.
When investors think the best-known names are priced too high, they look for where the market is heading and for what is needed to keep AI running, then hunt for investment chances there.
Photonics and Memory: PRAM
Photonics uses light to move data. Copper is being phased out because light moves data faster, more efficiently, and with bigger capacity.
The PRAM ETF combines photonics and memory. Many investors know the DRAM ETF, which covers only the memory side. PRAM covers both the memory that fuels the engine and the photonics that moves the data - the two pieces needed to make AI function.
The AIHY ETF covers hyperscalers, a more obvious play.
Capacitors: CAPA
Capacitors may be the most overlooked piece in AI because people rarely think about them. GPUs and AI servers demand huge power densities, so they need MLCCs (multilayer ceramic capacitors). These stabilize power, regulate voltage, and cut down noise so the AI system runs efficiently. Think of a small adapter along the path that moves data and keeps the data center working.
Without capacitors there is no AI, and it does not matter who wins - an Nvidia (NVDA) chip, Micron (MU), AMD (AMD), or anyone else. Every one of them needs the capacitor to run the process. The CAPA ETF targets capacitors and components.
The Full Circle
AI needs to calculate, so it needs compute. It needs to remember, so it needs memory. It needs to communicate, so it needs photonics. It needs stable power, which is where the capacitor comes in. Together these layers round out the picks-and-shovels and bottleneck view of the AI supply chain, each with an ETF built to capture it.
