5 Things Failing in AI. One Technology Fixes Them All!
Summary
Brian's main thesis is that the AI industry is hitting a "physics wall" with copper connectivity as data rates move toward 1.6 terabits per second. At these speeds, copper cables generate excessive heat and signal degradation, whereas silicon photonics—using light instead of electricity—is 3 to 3.5 times more power efficient. Brian outlines a massive market expansion, with current valuations at $3 billion projected to reach up to $30 billion by 2034. He structures the investment opportunity into three layers: the base layer of established giants, the middle layer of specialized foundries, and the top layer of pure-play photonics companies.
Brian provides a breakdown of several key players in the sector:
Mentioned Stocks
Reasoning: Brian identifies Broadcom as the industry leader in co-packaged optics, currently on its third generation of technology and scaling toward 400 gigabits per lane. He believes photonics will become a significant growth layer on top of their already dominant networking business.
Reasoning: Brian views Nvidia as the primary driver of the silicon photonics transition, integrating it directly into their next-gen Vera Rubin platform and Spectrum 6 switches to achieve 1.6 terabit speeds. He emphasizes their $4 billion investment in the space as a signal that the entire supply chain will follow their lead.
Reasoning: Brian calls Intel the most advanced research company in photonics with over 32 million on-chip lasers shipped. He highlights their OCI chiplet's 3x power efficiency and their unique position as the only optics-based foundry option for third-party customers.
Reasoning: Brian views TSMC's commitment to a dedicated photonics production line in 2026 as a major validation signal for the entire industry, marking the shift from research to mass production.
Reasoning: Brian notes Marvell owns 50% of the DSP market for transceivers and recently acquired Celestial AI for $3.2 billion. He points to a low PEG ratio of 0.64 and strong cash flows despite GAAP accounting losses from acquisitions.
Reasoning: Brian identifies Lumentum as the cheapest valuation in the space with a PEG ratio of 0.61. They provide the lasers essential for all photonic chips and recently secured a massive $2 billion purchase commitment from Nvidia.
Reasoning: Brian highlights Coherent's vertical integration and a $2 billion strategic investment from Nvidia. He considers the stock attractive with a PEG ratio of 1.1 and record quarterly revenue.
Reasoning: Brian states that Cisco is using silicon photonics to stay relevant in the data center market, utilizing tech from their Luxtera acquisition to build 800 gigabit transceivers and prototype 25 terabit switches.
Reasoning: Brian highlights this sponsored company for its GPS-free drone navigation technology and its expansion into active defense markets like Ukraine. He notes a stock price just over $1 and 34% insider ownership.
Reasoning: Brian notes Tower Semiconductor is a leading specialty foundry for photonic circuits with a direct partnership with Nvidia. Over 70% of their expanded capacity is already reserved through 2028 via customer prepayments.
Reasoning: Brian points to GlobalFoundries' goal of a $1 billion photonics run rate by 2028. While total revenue is flat, their ability to double photonics revenue annually makes them a key scale player in the transition.
Reasoning: Brian recommends Fabrinet as the safest entry point because they handle the high-precision manufacturing for major players like Coherent and Lumentum. They benefit from industry growth without taking on direct technology R&D risk.