The week earnings will be more than exciting. Palantir (PLTR:NYSE), AMD (AMD:NASDAQ), Alphabet (GOOGL:NASDAQ) and Qualcomm (QCOM:NASDAQ) will surely get plenty of attention, but we’ll also be paying attention to emerging networking powerhouses like Fabrinet (FN:NYSE), Lumen (LUMN:NYSE) and Coherent (COHR:NYSE) as proof that the artificial-intelligence buildout is extending not just up, but out. We’ll be looking at all these all week — and getting ready for another “San Francisco” Superbowl.

Dig out website! It’s a click away and hella imformative pretty: 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. Who Knew Before the News?

   1. Does a Maduro bet prove blatant corruption in pred …

   2. The Hole Magilla

   3. Hollow in the Wild

   4. Corning’s Standing on Hollow Ground

3. Tweet O’ The Week

4. 📆 of Epistrophy Events

5. Availability This Week

Who Knew Before the News?

Does a Maduro bet prove blatant corruption in prediction markets?

“There is no there there,” Gertrude Stein famously wrote in 1937 of her childhood home Oakland, California. 

Some ninety years later, that may well serve as the description for a cutting-edge technology quietly emerging as dominant in the artificial intelligence data centers buildout: hollow core fiber, a fiber optic cable that, at its core, has no cable at all; a cable where there is no there there. 

Corning's (GLW:NYSE)  strong fourth quarter results, reported January 28, showed a sales increase of 20% to $4.4 billion. Its Optical business was 38% of that, $1.7 billion and growing at a 24% clip. An important part of that was the sale of hollow core fiber. 

The day before that earnings announcement, CNBC’s Katie Tarasov broke news that Corning had signed an incredible “up to” $6 billion deal to supply advanced fiber-optic cables and connectivity solutions. Corning says that’s just one of many similar-sized deals it’s pursuing with many hyperscalers. “[There are] similar agreements to Meta that we are in the process of concluding,” CEO Wendall Weeks said on the conference call. “Okay, first, let's size them. They are of a similar size and scale, each of them, to the Meta agreement. So very significant.”

(Here’s a video of our full take on Corning’s quarter.) 

The innovation is, indeed, significant, given that there is no there there.

The Hole Magilla

For four decades, optical fiber has improved like dumb highway “improvements” – adding lanes through more fiber. Doped glass quickly led to high-density fiber cables. The glass content and reflective coatings were regularly tweaked and improved. It meant more wavelengths per fiber. Higher-order modulation. Denser packing of photons into the same glass strand. The material itself stayed constant. Light traveled through silica. Silica slowed it down. The industry accepted that compromise because it was reliably manufactured at massive scale.

Hollow-core fiber challenges the compromise directly.

Instead of transmitting light through a solid silica core, hollow-core fiber guides light through an air-filled central channel, surrounded by a precisely engineered glass microstructure that confines the optical field. It remains a glass product. The glass still determines geometry and confinement. But the photons themselves spend the majority of their path in air, not glass. That distinction changes the physics of propagation in a way the optical industry has not seen since the transition from multimode to single-mode fiber in the 1980s.

Hollow-core fiber is often described as “fiber with a hole in the middle,” which obscures the engineering. The hollow region is surrounded by a microstructured cladding, typically composed of nested glass tubes arranged with micron-level precision. That geometry must remain stable along kilometers of fiber. Small defects increase leakage. Minor asymmetries increase loss. Manufacturing tolerances are tight, and yield matters.

Here’s how to understand it. Optical fiber’s performance is governed by two measures: propagation speed and latency.

One: propagation speed, described as how fast light travels through the fiber, expressed as a fraction of the speed of light in vacuum. 

Two: latency: the accumulated time delay that results from that propagation over distance and scales linearly with each additional kilometer.

(For you music majors out there: “% of c” means the propagation speed expressed as a percentage of the speed of light in vacuum. “Latency per km (µs)” means the amount of time, measured in microseconds, that a signal takes to travel one kilometer of fiber.)

Faster propagation is the most visible advantage, but it is not the only one. Because light interacts less with glass, nonlinear effects are reduced. Signal integrity improves. Launch power limits increase. In early 2024, Corning’s hollow-core designs achieved record-low attenuation at the industry’s standard 1,550-nanometer wavelength, reaching loss levels below those of the best pure-silica single-mode fiber. 

The high speeds and low latency of hollow-core fiber comes at just the right time for artificial intelligence.

Large-scale model training increasingly spans multiple data centers – not “scale up” or “scale out” but "scale across." Inference workloads are being pushed closer to users even as compute gravitates toward regions with available power. 

At metro scale, hollow-core fiber reduces latency enough to change routing decisions. At regional scale, it changes where infrastructure can be placed. A network designed around a fixed latency budget can stretch farther without degrading application performance. Fiber, long treated as passive plumbing, has reentered the critical path.

Hollow in the Wild

What we learned from Corning’s fourth quarter announcement is that AI data centers – and hyperscalers in particular – are aggressively increasing their fiber spend, expanding the initial market for HCF.

Financial trading networks were among the first HCF adopters, where microseconds translate directly into competitive advantage. Metro links connecting trading venues followed. Research backbones and test networks came next. These are environments where cost is secondary to performance and where operators tolerate complexity in exchange for speed.

In 2022, Microsoft (MSFT: NASDAQ) saw these developments and acquired a small Corning competitor, UK-based hollow-core specialist Lumenisity. It integrated the technology into its Azure network engineering organization. Lumenisity emerged from the University of Southampton’s Optoelectronics Research Centre and brought with it manufacturing capability, including a roughly 40,000-square-foot production facility in Romsey, England. Since the acquisition, Microsoft has emphasized hollow-core fiber as part of its “purpose-built infrastructure” for AI workloads.

Microsoft has described HCF as part of Azure’s “purpose-built infrastructure” for latency-sensitive and AI workloads, citing propagation speeds “up to 47% faster than standard silica glass” and record-low optical loss at the industry’s 1,550-nanometer wavelength. In technical disclosures, Microsoft engineers have said pilot deployments already exceed 1,000 kilometers of installed hollow-core fiber carrying live traffic, with plans to scale materially as supply allows. The emphasis has been less on raw speed than on topology: extending the practical radius of a cloud region from roughly 80 kilometers toward 120 kilometers without breaching latency budgets. 

Amazon (AMZN: NASDAQ) has been more candid than is often assumed, particularly within Amazon Web Services. AWS has said its global network includes “nine million kilometers of fiber-optic cabling” and that “hollow-core fiber optics” can deliver a “30% improvement in transmission latency” relative to conventional fiber. In industry presentations, AWS engineers have described deploying hollow-core fiber to connect roughly ten data centers, framing the technology as a way to go “farther for the same latency target” rather than as a broad replacement for existing fiber. AWS has also acknowledged limits, citing poor manufacturing yields and high cost as the primary barriers to wider rollout. That framing makes clear that adoption is gated by supply and economics, not by hyperscaler appetite.

Google (GOOGL: NASDAQ) has publicly emphasized advances in multi-core fiber for subsea systems, highlighting capacity density rather than propagation speed, and has disclosed no hollow-core fiber deployments, pilot routes, or performance benchmarks. The absence of commentary is notable given Google’s history of custom network engineering and willingness to publicize infrastructure breakthroughs when they align with its operating model. If hollow-core fiber appears in Google’s backbone, it is likely to do so only after manufacturing economics converge toward conventional fiber, a threshold that has not yet been reached.

Meanwhile Oracle (ORCL: NYSE) has said little. Its public cloud narrative centers on pricing, bandwidth guarantees and service isolation rather than physical-layer innovation, and there have been no disclosed hollow-core trials, supplier relationships, or deployment metrics. That silence suggests Oracle is content to consume improvements as they become standardized rather than sponsor early manufacturing risk. 

Traditional carriers have responded to next-generation optical demands at the scale of their business models rather than by chasing early-stage physics. Lumen Technologies (LUMN: NYSE) has anchored its strategy publicly on expanding what it calls its Private Connectivity Fabric, a multibillion-dollar intercity network designed to support AI workloads with fiber that Lumen says delivers roughly 25% less optical loss per kilometer and 60% more capacity than older designs, and that achieves sub-five-millisecond edge latency across North America. Lumen’s leadership has framed this build as part of “the backbone for the AI economy. ” On Lumen’s first-quarter 2025 earnings call, management said the company has 57 In-Line Amplifier (ILA) sites under construction as part of its multi-billion-dollar network expansion to support AI-era connectivity. If hollow-core fiber achieves lower loss over distance, it can extend the spacing between ILAs or reduce their number, which directly lowers capital cost, power consumption and operational complexity. Lumen noted that conversion of its pipeline involves “net new routes” that must be “designed exactly right” with customers before they proceed, underscoring the engineering and commercial complexity.

Verizon Communications (VZ: NYSE) has made fewer public statements about advanced fiber physics specifically, instead stressing broad fiber densification and 5G backhaul as core network priorities. For carriers, hollow-core and other next-gen fiber technologies sit inside long planning and build cycles where manufacturability, supply control and predictable yields matter far more than vendor hype.

Corning’s Standing on Hollow Ground

Conventional single-mode fiber has benefited from more than forty years of process refinement. Global production runs into the hundreds of millions of miles annually. Yields are predictable. Handling and installation are standardized. Hollow-core fiber does not yet share those advantages. It is harder to fabricate, yields are lower, splicing is more delicate, bend-radius constraints are tighter and costs remain high.

That gap explains why Corning is the clearest signal in the hollow-core debate. Corning dominates the fiber market because it manufactures at scale, not because it dabbles in novelty. Hollow-core fiber is not arriving as a replacement but as an overlay, expanding only as manufacturing realities allow. 

In fiber, physics opens the door, but economics decides who walks through it. Corning may well lead the way.

Tweet O’ The Week

📆 of Epistrophy Events

Availability This Week

I’ll be in our San Francisco office at the Ferry Building all week, though much of it will be taken over by the NFL in preparation for the SuperBowl. Reach out for background conversations, or some longer off-the-record discussions if you want to bounce some ideas off me — I’ve had a bunch of these lately and I think it’s been a small help to just the right people!

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.