Our focus on the AI buildout is about to shift from chips to glass. After last week’s capex bursts from Alphabet, Amazon, Meta and Microsoft, this week we get to see whether the optical layer is ready to turn those dollars into photons. Coherent (COHR: NYSE), Fabrinet (FN: NYSE) and Lumentum (LITE: NASDAQ) report, which means we finally get numbers on the packaging, transceiver and enabling sides of the optical chain. That’s where AI data centers go from scaling up and start scaling out.

Have you checked out the website? It’s getting deep! The full archive and research database live here: https://epistrophy.beehiiv.com.

As always, I’m focused on three things:
1) Technology-driven change;
2) the latest in innovation and startup trends, and;
3) stock fraud.

Companies Discussed

In This Note:

1. Companies Discussed

2. Step Into the Light

   1. Table 1: From Electrons to Photons

   2. Table 2: The Expanding Role of Optics in the Data …

   3. Table 3: The Optical Value Chain

3. Tweet O’ The Week

4. Epistrophy In The News

5. 📆 of Epistrophy Events

6. Availability This Week

Step Into the Light

Luminomics 101: Understanding the glass half of the data-center boom.

Three optical companies report next week. Together they’ll show whether the AI build-out has reached its next level of growth—not in silicon, but in light. Fabrinet (FN: NYSE) reports Nov. 3, Lumentum Holdings (LITE: NASDAQ) Nov. 4, and Coherent Corp. (COHR: NYSE) Nov. 5. Each sits on a different rung of the optical chain: Fabrinet packages, Lumentum transmits, Coherent enables. 

Optical is emerging as a crucial technology as AI data centers move from “scaling up” to “scaling out.” Silicon Photonics, or “Optics”, is the physics that lets data outrun copper. Artificial Intelligence needs faster, cheaper ways to move data. Copper can’t keep up, but light can. Every signal leaving a GPU passes through lasers, modulators and glass. Where electrons meet resistance, photons move clean. The industry’s migration from electrical to optical interconnect is becoming a huge business, which we estimate at $12 billion today. You’re right — I slipped and linked the company names, not just the ticker/exchange pair.

But after the latest jaw-dropping capital-expenditure announcements by the likes of Alphabet (GOOG: NASDAQ), Amazon (AMZN: NASDAQ), Microsoft (MSFT: NASDAQ), Meta Platforms (META: NASDAQ) and Oracle (ORCL: NYSE), we estimate the optical market could grow from about $12 billion today to $45 billion by 2030, implying roughly a 30 % CAGR over five years.

Nvidia’s CEO Jensen Huang explained it back in March. “The connection on scale-up is copper,” he said at Nvidia GTC. “We should use copper as far as we can, and that’s, call it, a meter or two. And that’s incredibly good connectivity, very high reliability, very good energy efficiency, very low cost. And so we use copper as much as we can on scale-up. But on scale-out, where the data centers are now the size of the stadium, we’re going to need something for long distance running. And that is where Silicon Photonics comes in.”

Table 1: From Electrons to Photons

The shift from electrons to photons defines the physical boundary of AI scalability.

Cap ex plans announced last week are nothing but bullish for the sector. Amazon said its 2026 capex will “meaningfully increase,” driven by data-center infrastructure. Alphabet told analysts its fourth-quarter outlay will exceed all prior quarters “by a wide margin.” Meta raised its 2025 forecast to as much as $40 billion for “AI hardware and network interconnect.” Microsoft said infrastructure spending “remains elevated.” Oracle, less vocal but equally leveraged to bandwidth cost, keeps adding cloud regions at a rate implying double-digit capex growth. Five companies—roughly $180 billion in combined projects—will shape the optical demand curve for years. A growing share of that money now flows not to chips, but to light.

Optical interconnect is not glamour hardware, but it is essential. The term covers lasers, modulators, silicon-photonic or indium-phosphide chips, transceiver modules and the packaging that binds optics and electronics. Bandwidth transitions mark each generation: 100 G, then 200 G, then 400 G. The next steps—800 G and 1.6 T—are already in production at Coherent and Lumentum. Packaging, alignment, laser reliability, heat dissipation and power now dictate cost and yield. Externally modulated lasers, pump lasers, narrow-linewidth sources and co-packaged optics sit at the frontier, shrinking distance between photons and silicon.

Fabrinet, Lumentum and Coherent occupy distinct positions. Fabrinet converts designs into volume production—precision assembly of lasers, transceivers and opto-mechanical systems. Lumentum designs and manufactures photonic products spanning datacom, telecom and industrial markets. Coherent sits further upstream. Its photonics platform includes lasers, optical materials and components across communications and electronics. All three are poised to benefit from this new “scale out” phase of the AI data center buildout.

Table 2: The Expanding Role of Optics in the Data Center

“Scale Up” captures optics replacing copper inside racks; “Scale Out” shows optics defining the interconnect fabric of AI clusters.

Optical infrastructure is not one industry but a relay—wafer, laser, transceiver, system, network, deployment. Each handoff decides who captures margin and who bears cost.

The constraint isn’t demand but throughput—the capacity to build, test and qualify modules at scale. Fabrinet’s utilization sits near record highs. Lumentum’s shift from telecom to datacom has tilted revenue toward hyperscale clients. Coherent keeps adding indium-phosphide capacity and automation. Each is chasing margin expansion against higher component and labor cost. Packaging yields, component scarcity and power budgets are the known risks. Lead times remain long. Some parts are still tight. But as GPU shipments double, optical interconnect must grow even faster.

Table 3: The Optical Value Chain

From photonic materials to hyperscale deployment, this chain traces the path from photon generation to photon revenue.

The likely outcome is a synchronized up-cycle—rising revenue, swelling order books, early commitments for 800 G and 1.6 T optics. Perhaps the true risk isn’t demand; it’s manufacturing. If they deliver, optics becomes the next capital-spending frontier—the picks and shovels of the data-center boom. If they stumble, the bottleneck is structural: packaging yields, assembly limits, delays in co-packaged optics. The consequences will echo across the AI supply chain. 

Optics has lived in the shadow of semiconductors for two decades. Every jump in compute eventually meets physics. The limits of copper are fixed. The promise of light is not.

Tweet O’ The Week

Epistrophy In The News

On NewsNation I broke down the new “Magnificent Seven,” the AI cycle, and why layoffs rise even as companies boast of efficiency.

In another segment later in the week we examined AI’s political fallout—the possibility that concentrated automation and wealth could fuel a new wave of socialism. Host Connell McShane thanked me describing me as “a really smart guy in Silicon Valley.”

I may have that stitched on a pillow.

📆 of Epistrophy Events

Availability This Week

I’m in San Francisco early in the week, then in Los Angeles Thursday and Friday for client meetings. I’m reachable by text for time-sensitive stuff; email works too, replies later in the day.

Written reports are available to clients, with video summaries on YouTube, and of course our popular summaries of the summaries on Instagram, TikTok, and YouTube Shorts.