This week’s Optical Fiber Communication (OFC) conference lands in San Francisco for the first time, shifting the center of gravity in the networking world from San Diego to SoMa. While the physical infrastructure of the internet rarely gets headlines, the companies showing up here—like Applied Optoelectronics (AAOI: NASDAQ), Coherent (COHR: NASDAQ), Lumentum (LITE: NASDAQ), and Fabrinet (FN: NYSE)—move the photons that move the packets that power AI models, trading desks, military operations, and TikTok trends.

Last week, we watched two kinds of exposure surface: one digital, one economic. The Trump administration’s proposed 25% tariffs on Chinese-made EVs look like a gift to Tesla (TSLA: NASDAQ). At the same time, the administration’s own cybersecurity team revealed themselves to be somewhere between amateur hour and a national security crisis. Underneath it all is a message about technical debt worth hearing.

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

Companies Discussed

In This Note:

• Companies Discussed

• AI Sees The Light: OFC 2025

• Red Cat: All Buzz?

• Tweet O’ The Week

• Epistrophy In The News

• 📆 of Epistrophy Events

• Availability This Week

AI Sees The Light: OFC 2025

OFC used to be a telecom show. Fiber was for carriers and the innovations on display—dense wavelength division multiplexing, metro rings, submarine repeaters—lived deep inside the backbone of the internet. You could walk the show floor without hearing the word “data center.” That era is over.

Now the Optical Fiber Communication Conference is about AI. Fiber isn’t just a long-haul solution—it’s a proximity weapon. Every major hyperscaler is racing to build optical interconnects that can keep GPUs talking at scale. And this week in San Francisco, more than 600 exhibitors will converge at OFC 2025 to show how light—split, modulated, and bent into silicon—can solve AI’s most expensive problem: getting data in and out of chips fast enough to matter.

The infrastructure boom—everything from ChatGPT to Tesla’s Dojo—is running into physical limits. Traditional copper-based interconnects can’t scale without burning through power budgets and rack space. As large language models grow and inference moves into production, networking shifts from afterthought to architecture. That’s what makes OFC 2025 more relevant than most earnings calls. It’s where the AI supply chain shows its hand.

Optical networking used to follow a predictable cadence: 200G to 400G, then to 800G. But the jump to 1.6T is a different kind of leap. It’s not just bandwidth—it’s a change in physics, packaging, and economics. Companies like Broadcom (AVGO: NASDAQ) are betting on co-packaged optics (CPO), which marry photonics and compute into a single substrate. The goal: eliminate the heat and inefficiencies of pluggable transceivers altogether.

At OFC, Broadcom will demonstrate its XPU-CPO platform, a 6.4Tbps optical engine mounted directly onto a 2.5D XPU. Lab comparisons show CPO draws 3.5x less optical power than DSP-based pluggables—and 25% less than even the more efficient LPO designs. The TH5-Bailly system, their newest production switch, is already shipping. It’s not theory. It’s SKU.

That power efficiency matters. Optical interconnects are now the limiting factor in data center density. The GPU rack is no longer the star of the show—it’s the wiring diagram that determines how big the cluster can grow, how far traffic can travel, and how many watts are left over for actual compute.

While Nvidia builds the engines, OFC is where we find the roads. And this year’s roads are expanding fast. The show will feature live demos of 800G OpenZR+, multi-span coherent optics, quantum-safe encryption, and multicore fiber—all designed to carry AI traffic faster and cheaper across disaggregated clusters. You’ll also hear about CMIS, a dry-sounding standard that quietly unlocks multi-vendor scale for hyperscalers.

The story isn’t just technical. It’s financial. Amazon Web Services has secured long-term access to optical components just a couple weeks ago,  through two quiet but telling deals. In March, Amazon received a warrant to purchase up to 7.945 million shares of Applied Optoelectronics (AAOI: NASDAQ) at $23.70 per share, tied to up to $4 billion in purchases. Separately, it obtained the right to acquire 381,922 shares of Fabrinet (FN: NYSE) at $208.48 per share, with vesting tied to AWS spending. Neither deal made front-page news—but both signal that Amazon is locking down its optics supply chain before the next AI buildout begins.

Meanwhile, Coherent (COHR: NYSE) and Lumentum (LITE: NASDAQ) continue to blur the line between telecom optics and AI infrastructure. Coherent still plays at the high end with specialized lasers; Lumentum straddles telecom and datacom. At OFC, their relevance may come down to whether the market prioritizes performance or power per dollar.

We expect more than $1 billion in annual PAM4 DSP sales by 2028, driven by 1.6T transceivers alone. That doesn’t include lasers, drivers, alignment systems, or the software stack. Optical networking has become the most important infrastructure layer not named “semiconductors.” It’s also the least understood.

Some will still try to sell vision. Expect startups to push low-cost LPO modules or novel short-reach solutions for AI clusters. But OFC has become a filter. If your design isn’t already in someone’s BOM—or can’t be tuned with CMIS—it’s background noise.

By the end of the week, we’ll have a sharper picture of where AI’s backbone is really being built. Microsoft, Meta, and Google won’t say much directly—but you can read their plans in the booth partners, the interoperability demos, and the whispered sourcing conversations behind the curtained back rooms.

Semiconductors built the brain. Optics is building the spine. If you want to understand where AI infrastructure is going next, skip the model release and watch the link budget. OFC 2025 will be the quietest, brightest place to start.

Red Cat: All Buzz?

Red Cat Holdings (RCAT:NASDAQ) has an exciting story to tell potential investors. It pitches itself as the future of warfare, manufacturing armed drones for the US Army. 

In November 2024, Red Cat announced it had secured the U.S. Army’s Short Range Reconnaissance (SRR) Program of Record contract for its Black Widow drone. This milestone signals a leap into defense primacy and, according to the company, opens a path to $350–400 million in revenue over five years. It implies not only technical credibility, but a sticky, recurring military revenue stream with long-term potential.

And the company is actively expanding its portfolio. Through the September 2024 acquisition of FlightWave Aerospace, Red Cat added the Edge 130 Blue drone, a platform it says is mission-ready and already generating orders. The company claims over $500,000 in combined sales to the Army National Guard and another federal agency, framing this as evidence of growing beyond SRR market. 

Red Cat’s management has reinforced this narrative with bullish 2025 guidance. On top of expected SRR revenue, the company projects an additional $50–55 million in sales from the Teal 3 and Edge 130 models. That would mark a sharp step up from its historical performance and position Red Cat as a top-tier defense drone supplier. 

Or is it?

Budget Reality Check

Red Cat’s SRR claims diverge sharply from Department of Defense documentation. The company has suggested the contract could generate up to $395 million over five years (Red Cat CEO Jeffrey Thompson said on their Q1 2025 call “The upcoming decision on the SRR program... could potentially be worth $79 million in the first year of deliveries.”)

Yet the Army’s official FY 2025 budget request allocates just $21.1 million for SRR procurement, covering 270 systems at roughly $65,000 each. Another $3.5 million is earmarked for training, logistics, and spares.

There is no evidence the Army has committed to Red Cat as an exclusive or long-term supplier. On the contrary, the SRR program was explicitly designed to refresh hardware every 2–3 years to keep pace with evolving requirements and technological advances. The current contract is for Tranche 2. Nothing in Army budget documents suggests a locked-in, sole-source arrangement through 2029.

Adding to the confusion, Red Cat disclosed its contract award before the Army publicly confirmed it. While this may have reflected a strategic effort to capitalize on momentum, defense insiders say the move breached customary norms and could impair the company’s standing in future negotiations. It also fed unrealistic speculation that other military branches or federal agencies would follow suit—despite budget data showing that the Air Force has minimal use for infantry drones, the Marines and Coast Guard maintain separate drone acquisition efforts.

Production Claims vs. Capital Expenditures

Red Cat has repeatedly promised to scale production rapidly. Since 2022, executives have said a high-volume factory was just months away. Yet the company’s SEC filings reveal only $3.1 million in capital expenditures over the past three years—a paltry sum given its manufacturing ambitions. Cap ex even went down last quarter to $23,836.

At its peak, the company’s best-selling Teal 2 drone generated $17 million in revenue, selling 1,100 units over 18 months. Management now claims it will triple that output across three drone models in 2025 alone. That projection is unsupported by its current infrastructure. The factory is not equipped to produce more than one drone type at a time, and in early 2024, management said Teal 2 production had to pause for SRR retooling. That would directly contradict the claim that it can now run multiple production lines concurrently.

CEO Jeffrey Thompson has claimed that Red Cat will have a new production line operational within two months to support Edge 130 manufacturing. But FlightWave, the company behind the Edge 130, had never moved beyond the prototype stage and lacked a factory when acquired. As of early 2025, there is no evidence Red Cat has sourced the necessary tooling, workforce, or suppliers to meet its targets. The new line remains hypothetical.

Flight from the Inside

Red Cat’s own leadership seems unconvinced by its outlook. Within weeks of the SRR contract announcement, several key executives resigned and sold significant blocks of stock. George Matus, CTO and founder of Teal Drones, resigned on November 25, 2024, selling 800,000 shares at an average price of $11.75. CFO Leah Lunger stepped down on December 15 and liquidated 500,000 shares—roughly 60% of her holdings. CEO Thompson and other executives offloaded a combined 1.6 million shares in the weeks that followed.

Matus’s resignation is especially revealing. He had negotiated a 900,000-share equity grant in May 2024, with one-third vesting if Red Cat secured a major military contract. Less than a week after meeting that milestone, he quit. 

The Kesner-Honig Pattern

Red Cat’s associations extend into a murkier financial history. Its legal adviser Harvey Kesner has been linked to more than two dozen Barry Honig-backed firms accused of stock manipulation. Honig himself was charged by the SEC in 2018 for orchestrating fraudulent stock promotions that yielded $27 million in illicit profits (SEC v. Honig, Case 1:18-cv-08175, U.S. District Court, S.D.N.Y., Sept. 2018).

Kesner’s past work includes setting up legal frameworks that concealed ownership stakes and facilitated aggressive stock sales. At his former firm, Sichenzia Ross Ference, he was named in litigation by MabVax Therapeutics (MBVX: NASDAQ) for allegedly designing insider-friendly transactions that contributed to the company’s delisting (MabVax Therapeutics Holdings, Inc. v. Sichenzia Ross Ference, Superior Court of California, 2019, p. 18).

That pattern is visible here. Since 2022, Red Cat has issued frequent press releases emphasizing big potential contracts, but those rarely translate into sustained revenue. In parallel, its stock has shown volatile spikes—followed by insider sales.

Trump and Unusual Machines

In July 2024, Red Cat quietly sold its 49% stake in Unusual Machines—the company it spun off its FPV drone business to—for $4.4 million. After acquiring the FPV assets, Unusual Machines (UMAC: NASDAQ) conducted a private placement in October 2024, issuing over 1.28 million units at $1.52 apiece and raising roughly $1.95 million. Company insiders—including CEO Allan Evans—participated in the deal. While not illegal, this level of insider access to discounted stock mirrors the dynamics of companies previously associated with Barry Honig. 

There is also a potential indirect link: SEC documents from Unusual Machines’ 2023 Form S-1 disclose that Jonathan Honig, Barry Honig’s brother, had voting and dispositive control over shares held by Titan Multi Strategy Fund I Ltd, a Boca Roton, FL-based major stakeholder in the company. While there is no proof Barry Honig himself is involved in Unusual Machines, the overlap in governance patterns, share structures, the use of convertible securities before a sudden run in the stock and insider enrichment mechanisms echoes past setups.

​In November 2024, Unusual Machines appointed the president’s son Donald Trump Jr. to its advisory board. Trump Jr. participated in a private placement with Unusual Machines in October 2024, acquiring 331,580 shares at $1.52 per unit. Following the announcement of his appointment to the company's advisory board on November 27, 2024, Unusual Machines' stock experienced a significant surge, more than doubling in value during that day's trading session.

Following the Trump Bump, Honig has yet to convert his Series C Convertible Preferred Stock into common shares. But priced on pre-surge valuation, Honig could benefit from the post-announcement spike. The conversion mechanism, structured around a volume-weighted average price (VWAP) prior to the stock’s run-up, would allow Honig to obtain shares at a substantial discount relative to the new market price—creating a dynamic similar to past setups where affiliated funds profited from orchestrated publicity boosts.

No NATO, No Market

Red Cat frequently references interest from NATO allies, but these opportunities have failed to materialize. European militaries tend to rely on domestic drone suppliers or modified Chinese hardware, which is cheaper and often more capable than Red Cat’s offerings. While Red Cat claims compliance with U.S. defense export rules, its record of closing international contracts is limited.

The broader drone market is also saturated. Competitors such as Skydio, Anduril, and AeroVironment have better funding, deeper government relationships, and actual delivery capacity. Red Cat’s strategy appears to hinge on the hope that one big contract win can snowball into industry-wide adoption. But as the SRR numbers show, even its biggest win doesn’t pencil out.

Gravity Wins Again

Red Cat Holdings markets itself as a breakout defense technology company. But its core assumptions—that SRR is a transformative contract, that production can scale, and that non-Army demand will surge—fail under scrutiny. Budget documents don’t support its revenue claims. The company lacks the infrastructure to meet its own manufacturing goals. Insiders are selling. And the financial history surrounding its leadership and affiliates resembles a familiar blueprint for short-term stock hype.

There may be a future in military drones. But Red Cat’s story, as told by its own filings, doesn’t hold altitude.

Tweet O’ The Week

Epistrophy In The News

On Bloomberg BNN, I broke down GameStop’s surprise move to add Bitcoin to its treasury. The nostalgia play is over—this is a pivot to programmable money, and it’s either a desperation tactic or a small stroke of genius. Watch the clip.

On Schwab Network, I explained why DOGE—the Department of Government Efficiency—poses an existential threat to IBM (IBM: NYSE), Booz Allen Hamilton (BAH: NYSE), and others weighed down by services contracts. Meanwhile, companies like Elastic (ESTC: NYSE) and Palantir (PLTR: NYSE) are skating into the vacuum, offering scalable platforms with lower onboarding costs and tighter AI hooks. Here’s the segment.

And on NewsNation, I joined Connell McShane to talk about how the Trump tariffs create not just pain for Chinese EV makers, but real market power for Elon Musk. There’s a geopolitical story hiding in the drivetrain—and Tesla might be writing the next chapter. Full video here.

📆 of Epistrophy Events

Availability This Week

I’ll be at the OFC conference in San Francisco early this week, listening to people who make photons go faster than you read emails. If you’re there, come find me—I’ll be easy to spot: the guy asking weirdly specific questions about optical transceivers.

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

Later in the week, I’ll be in New York City for customer meetings. If you’d like to connect in person or follow up on anything from the newsletter, reach out. And if I don’t respond right away, text me. I’m probably in transit, not ignoring you.