Beyond the Buildout: Why AI Capital is Shifting Toward Monetization

The split nobody's pricing in

For two years the AI trade ran on one engine: capex. Buy the hyperscalers, buy the chipmakers, ride the buildout. That engine still runs, but it isn't the only one producing returns anymore. This week's tape, with cooling inflation easing rate worries and AI spending continuing to lend support, gave enough oxygen for a quiet rotation underneath the AI umbrella. The question is no longer whether data center buildout continues. It's whether your exposure captures the companies being paid to ship, or just the companies being paid to spend.

That distinction is the entire game. A hyperscaler writing checks for GPUs doesn't print money into your brokerage account. The optics suppliers, laser makers, and silicon photonics foundries feeding that buildout do — eventually. As data centers push toward gigawatt scale, moving data between GPUs, switches, and racks through copper becomes a physics problem, and photonics becomes the bottleneck technology. Broad AI ETFs and semiconductor indexes tend to underweight these names because their market caps are small relative to NVIDIA, Broadcom, or TSMC.

The PHOX bet, stripped down

The Aura AI Photonics ETF (PHOX) is the cleanest expression of that rotation hitting public markets. Listed on NYSE Arca under the Tidal Trust III umbrella, with Tidal Investments as adviser and a prospectus dated July 29, 2026, its mandate is narrow: optical interconnects, silicon photonics, and laser-based transceivers — the components moving light instead of electrons through AI stacks.

Two numbers tell the story. The 0.60% expense ratio sits above a plain-vanilla semiconductor index but in line with single-theme AI ETFs. And the fund is too new for a meaningful NAV history, distribution record, or tracking data you can stress-test. Any article claiming otherwise is inventing numbers.

What you can evaluate is the structural argument. Broad semiconductor funds like the VanEck Semiconductor ETF or the iShares Semiconductor ETF give you heavy NVIDIA, TSMC, Broadcom, and AMD weight, but photonics pure-plays typically account for only a low single-digit weight in those funds. If you believe the incremental capex dollar is heading toward optical interconnects, a diluted position inside a mega-cap semi ETF isn't going to move your portfolio the way a concentrated basket will. That's PHOX's real pitch. The 0.60% is the cost of that torque.

Your move

Two options, no third. Buy a broad semiconductor ETF and accept the low single-digit photonics weight as the price of liquidity and lower fees. Or pay 0.60% for concentrated exposure to a thesis that hasn't survived a single quarterly cycle yet. The opportunity cost runs both directions — you're trading breadth for torque, or torque for cost efficiency.

The disciplined move: wait for two quarters of NAV data before sizing. Speculation wrapped in a thematic ETF wrapper is still speculation.