The UNTHINKABLE Just Happened. Prep for the Dip!

BWB - Business With Brian•Apr 19, 2026

Summary

Brian presents a thesis centered on the 'shovels and picks' strategy for the AI revolution, identifying nine specific companies that control critical resource constraints. He believes that while the market is fixated on chip designers, the real long-term value lies in the physical infrastructure—power, cooling, and hardware components—that allows AI clusters to function. Brian outlook for 2026 suggests a widening gap between the surging demand for data center capacity and the limited supply of transformers, high-bandwidth memory, and advanced packaging.

Brian organizes his investment framework into three layers: the foundation (power and cooling), the silicon stack (memory and networking), and physical materials (copper and fiber). He highlights that traditional energy and hardware companies are seeing unprecedented growth due to AI's unique physical demands, such as liquid cooling and massive fiber requirements. Brian suggests that the current supply deficits in these areas are not temporary, but structural shifts that will last several years.

Vistra Corp (VST): Vistra is a leader in nuclear power, controlling 3,800 MW of contracted power for AI data centers, which is double its nearest competitor. Brian notes they operate in a deregulated market, allowing them to sign direct, high-margin long-term contracts with giants like Meta and Amazon. He expects 22% EBITDA growth in 2026 and highlights their fixed cost structure as a major advantage during energy price spikes.

Eaton Corporation (ETN): Eaton dominates the power distribution chain, with Brian pointing out that transformer lead times have stretched to nearly three years. The company has a massive 11-year backlog and captures $3.4 million in revenue for every megawatt of data center capacity installed. Brian emphasizes that their operating margins are targeting 30% in the core data center segment by 2026.

US Gold Mining Inc (USGO): As a developer of the Whistler Project in Alaska, this company sits on significant reserves of gold and copper, which are now on the US critical minerals list. Brian mentions that a recent assessment estimates production of 2.7 million ounces of gold and 600 million pounds of copper over a 15-year life. He notes the payback period for capital is extremely short—just 1.2 years at current spot prices.

Vertiv (VRT): Vertiv is the preferred infrastructure provider for Nvidia’s AI factories, specializing in the liquid cooling systems required for high-power AI racks. Brian states that traditional air cooling cannot handle the 132 kW loads of modern racks, leaving Vertiv as a primary beneficiary with a $15 billion backlog. He expects 30% revenue growth in 2026 as they are the only major player providing fully integrated power and cooling systems.

Micron Technology (MU): Micron is one of only three global producers of High-Bandwidth Memory (HBM), which is essential for Nvidia chips and currently sold out through 2026. Brian highlights Micron's advantage in having US-based fabs, shielding them from supply chain disruptions in the Middle East that affect Korean competitors. He points out an incredibly low PEG ratio of 0.25, making it the fastest-growing company on his list.

Amkor Technology (AMKR): Amkor provides advanced packaging (CoWoS) services and acts as a vital second source for companies that cannot get capacity from TSMC. Brian notes that demand for this packaging is growing at 80% annually while supply only grows at 50%, creating a permanent gap. Their Arizona facility is strategically located near TSMC’s US fabs, and Brian expects their advanced packaging business to triple this year.

Broadcom (AVGO): Broadcom holds a dominant position in both custom AI accelerators and Ethernet switching, controlling 80% of the market for high-end switches. Brian mentions they entered 2026 with $73 billion in signed AI orders from the world's five largest spenders. He notes that AI revenue now accounts for 43% of their total business, driven by co-designed chips for Meta, Google, and Amazon.

Marvell Technology (MRVL): Marvell is a leader in optical DSPs, which are processors that convert electrical signals into light for fiber optic transmission. Brian explains that as data centers move to higher signaling speeds, copper becomes inefficient, making Marvell's optical silicon mandatory. He notes that their earnings per share tripled in a single year following their acquisition of Inphi.

Southern Copper Corporation (SCCO): This company operates the largest copper reserves of any listed miner and has the lowest net cash production costs in the industry. Brian highlights that AI data centers require 27 to 33 tons of copper per megawatt, contributing to a projected 10 million ton global shortfall by 2040. He notes the company maintains a 52% operating margin, which is exceptional for the mining sector.

Corning (GLW): Corning is the primary beneficiary of the shift to photonics, as AI data centers require 36 times more fiber than traditional racks. Brian points to a multi-year, $6 billion deal with Meta as evidence of the massive demand for their optical fiber and connectivity solutions. He notes that lead times for ribbon fiber already exceed 60 weeks, ensuring a long-term supply squeeze that benefits Corning's margins.