Jason Bodner’s “Accelerated AI” pitch rests on a real piece of physics and a forward-looking claim built on top of it. The physics is that light travels faster than electricity through copper, and at the data rates AI clusters now demand, copper is hitting a wall. The forward-looking claim is that this shift “leaves the Magnificent Seven in the dust.” Both parts are worth tracing through the public record.
The pitch is for Bodner’s Brownstone Research newsletter, Inflection Point, priced at $179 for the first year and renewing at $199, with a 30-day refund window. It landed in Stock Gumshoe’s de-tease queue on July 16, 2026. The framing is a “light-speed device” that makes AI 100 times faster, and three paid “Bonus Accelerated AI” picks positioned as the suppliers who ride the next wave. Bodner’s own description, in the copy, is direct: “Photonics fixes this. It replaces copper wires with laser beams. Data travels through glass fibers instead of metal.”
Who Is Jason Bodner
Jason Bodner spent fourteen years trading and selling derivatives institutionally, twelve of them at Cantor Fitzgerald, the last eight as Head of Equity Derivatives North America in New York. He co-founded Mapsignals, a quantitative research firm that tracks unusual institutional buying, and he co-founded Quantum Edge at TradeSmith. He has contributed to Brownstone Research on and off for at least five years alongside Jeff Brown, and Inflection Point is his own newsletter under the Brownstone imprint. His stated methodology is “tracking institutional flows and identifying the stocks that the big money is buying right now.” Bodner is a flows guy, not a component engineer. When he talks about who is buying photonics stocks, that is his lane.
The Mechanism in Plain English
Photonics is the use of light, carried through glass fiber and silicon waveguides, to move data between chips, racks, and data centers instead of pushing electrons through copper. The argument for it is physical. Copper loses signal integrity and burns power as the data rate rises. At 100 gigabits per second per lane, copper carries a signal roughly five meters. At 200 Gb/s, that reach falls to about two and a half meters. At 400 Gb/s, copper stops working reliably even inside a single rack. AI training clusters now run at 800 Gb/s and are moving to 1.6 terabits per second. The copper run is over.
Optical interconnects solve this by converting electrical signals to light at the chip boundary, transmitting over fiber, and converting back. A 1.6 Tbps optical link works the same at 100 meters as at one meter. The energy cost per bit is lower and the cables are lighter — the real thesis, and the architectural direction NVIDIA, Broadcom, TSMC, Marvell, and the entire OFC 2026 conference have been building toward.
The scale of the shift is verifiable in public filings and press releases. NVIDIA has committed at least $6.5 billion to photonics companies since March 2026. Two billion each went into Coherent and Lumentum, both deals paired with multi-billion-dollar purchase commitments and funding for new U.S. fabrication capacity. Another $2 billion went to Marvell, which had acquired the optical interconnect startup Celestial AI in December 2025. Up to $3.2 billion went to Corning through equity warrants and multi-year purchase agreements. Jensen Huang has said publicly that co-packaged optics will ship in 2028 with the Feynman NVLink 8 architecture. None of that is hidden.
The wider read on the photonics transition is that it is an industry-wide shift to optical networking that has been underway for two years and is now accelerating because AI cluster bandwidth outgrew copper. The framing in Bodner’s pitch — a single “100x faster device” — concentrates an industry-wide transition into a specific product claim.
The Picks and What Is Public
The free pick in the presentation is Broadcom, ticker AVGO. Bodner names it openly in the pitch. Broadcom crossed the $1 trillion market cap threshold and designs the custom AI silicon for Alphabet, Amazon, and others, alongside its photonics and networking equipment business. Naming AVGO is fair. It is a megacap everyone knows, and Bodner surfaces it himself in a free presentation.
The three “Bonus Accelerated AI” picks are behind the Inflection Point subscription. Stock Gumshoe has published a full de-tease identifying them. Flak Jacket Finance does not reveal picks from paid products, even ones a third-party site has already solved. What can be said honestly is that all three sit in the optical networking and silicon photonics supply chain, and that the category has been one of the hotter market stories of the past year. A wave of photonics ETFs has launched in the last few months, the largest of which still holds under $500 million in assets. That is the shape of a trade getting crowded, not the shape of a secret.
The structural question on the paid picks turns on whether the specific companies Bodner names capture the photonics transition, or whether the transition is broad enough that capturing it runs through owning the basket. NVIDIA’s $6.5 billion in commitments went to five different companies. The Open CPX multi-source agreement counts Ciena, Coherent, Marvell, Molex, Samtec, and TeraHop as members. The supply chain is wide. The thesis is real. The specific stock selection is where the paid product earns its money.
What the Thesis Depends On
The “100x faster” headline is the first component. The physics supports faster data movement at lower power. It does not support a 100x speedup of the AI itself. The AI’s speed is bounded by the GPU, the memory, the model architecture, and the interconnect. Photonics removes one bottleneck. Removing a bottleneck helps, and at the cluster scale it helps a lot. The leap from “optical interconnect lowers power and extends reach” to “AI runs 100x faster” is a forward-looking bridge from the physics to a system-level claim.
The “leaves the Magnificent Seven in the dust” framing is the second component. NVIDIA is the single largest investor in the photonics supply chain. Broadcom is the free pick. The thesis that the suppliers outperform the customers in a buildout cycle is a historical pattern, and the Magnificent Seven are funding the photonics buildout as it happens rather than standing still while it happens around them.
The third piece is timing. Co-packaged optics, the architecture where the optical engine is integrated into the switch or GPU package, is the part of the photonics story that actually changes the data center. TrendForce projects CPO will reach roughly 35 percent of data center optical deployments by 2030. That means 65 percent will still be pluggable transceivers at the end of the decade. The transition is real and runs on a multi-year timeline rather than a single catalytic moment.
What the Thesis Plays Out Through
The technology is real and the public record on it is extensive. NVIDIA’s GTC 2026 keynote transcripts, the OFC 2026 coverage, and the press releases on the Coherent, Lumentum, Marvell, and Corning deals carry the architecture, the timeline, and the named players. The free pick, Broadcom, is named in the presentation. The paid picks are behind the subscription. The 30-day refund window is the period to inspect the methodology — the flow data on the specific names is the value the subscription carries, and the photonics basket is available in the public market through ETFs and the named suppliers. The thesis plays out through the CPO adoption curve and the flow data on the specific picks, with the 30-day window as the inspection period for the methodology itself.
Flak Jacket Finance covers investment newsletters as an independent third party. We do not reveal paid picks, we do not call gurus scammers, and we do not sell the promos we cover. Read the full disclosure here: /disclosure.