Hi {{first_name | Investor}},
In 2017, I saw NVIDIA differently than most investors because of two things working in my favor at once: I was a gamer who knew the hardware, and a Googler who understood what it meant to own a developer ecosystem. That combination let me see a moat most people missed.
The same pattern is happening again. And this time, the thing most people are still watching isn't where the constraint actually is.
Everyone is watching chips. I'm watching the wire.
Here's what I mean. Training a modern AI model requires thousands of chips communicating simultaneously — every millisecond of delay costs money. At the scale AI has reached in 2026, the physical limits of copper wiring are no longer a future problem. They're a present one. Copper loses signal over distance. It generates heat. It consumes power at a scale that's starting to determine data center economics.
The longer-term answer is light. Replacing copper with optical connections that transmit data as pulses of light — this is the field called photonics. (Copper still has its own near-term story, and intermediary solutions are bridging the gap. But the direction of travel is clear, and the infrastructure buildout is already underway.) And the reason I've spent months mapping it isn't because it sounds compelling. It's because the inflection has already started.
This isn't a theoretical play.
Hyperscalers are actively deploying next-generation optical networking in real procurement decisions right now. At OFC 2026 — the industry's flagship photonics conference — the language from companies wasn't "we're exploring." It was "we're shipping." The transition from copper to light is underway, and the companies positioned across the supply chain are starting to see it in their revenue.
The Copper Wall
The way I think about photonics is the same way I'd think about the smartphone era: Apple got the headlines, but real value was created at Corning (screen glass), ARM (chips), and Qualcomm (modems). Photonics has the same structure — a full stack, with potential winners at every layer. The component makers. The networking silicon that routes the data. The physical infrastructure. And a layer almost no one talks about: the testing and validation companies that certify every optical component works before it touches a live AI cluster.
Getting this map right matters. Because not every company in a hot sector wins, and the ones that do often aren't the names everyone's watching.
How to Build Your Own Photonics ETF
There is no photonics ETF as of writing of this email. The closest options are broad tech or optical networking funds that bundle in too much noise. If you want real exposure to this theme, you have to build the basket yourself.
Here's the framework I use on how to think about constructing the position.
Think in layers, not names. The photonics opportunity isn't one company — it's a supply chain. A well-constructed basket has coverage across multiple layers: the component makers closest to the optical transition, the silicon and platform companies that make it all function, the physical infrastructure that benefits regardless of which technology wins, and the testing layer that certifies everything works. If your basket only has exposure to one layer, you're not diversified across the theme — you're concentrated in a single bet within it.
Split your conviction into tiers. I think about photonics positions in two buckets. The first is companies already generating revenue from the optical buildout — products shipping, customers named, numbers visible in quarterly results. These are the core of the basket. The second is earlier-stage, higher-optionality names where the thesis is directionally right but execution risk is real. These belong at the edges, sized accordingly.
Size to the risk profile, not the excitement. The names with the most compelling stories in an emerging theme are often the most volatile. A company with a great demo and no shipping revenue deserves a smaller position than one with a confirmed hyperscaler qualification and growing backlog. Enthusiasm is not a substitute for revenue.
Know your rebalancing triggers in advance. Before you build the basket, decide what would cause you to add and what would cause you to trim. For photonics specifically, the signals worth watching are hyperscaler CapEx guidance, qualification announcements from named customers, and the language CEOs use on earnings calls — "in trials" versus "in production" is a meaningful distinction. The full report includes a monitoring framework you can use for exactly this.
Today's Deep Dive on Photonics
I've published a photonics deep-dive — Breaking the Copper Wall — that maps the full photonics meets AI landscape. It covers the technology in plain English, every layer of the supply chain, companies worth adding to your watchlist, and three real risks that could slow the thesis. I also included a monitoring framework so you can track whether this story is accelerating or stalling — without following every earnings call yourself.
As always: I won't tell you what to buy. I'll sharpen the lens you use to look.
Stay disciplined,
Koh