The picks-and-shovels of the AI buildout
Every technology boom has a supply chain underneath it. During the gold rushes of the 1800s, the reliable fortunes went to the people selling picks and shovels rather than the people digging for gold. The same logic runs through artificial intelligence today. The companies selling the chips, the memory, and the tools that make AI possible sit one layer beneath the consumer-facing names, and that is where a large share of the value has actually been created.
Semiconductor stocks have been the clearest expression of that idea. When a data center orders thousands of accelerators, someone has to make the silicon, and someone has to make the memory that feeds it. Those suppliers have been among the best-performing stocks of the past few years, and for a simple reason: demand for compute keeps growing, and every additional chip requires more of everything else.
How the value chain breaks down
The semiconductor industry is not one industry. It splits into a handful of layers, and each one has its own economics and its own leaders.
At the top sit the chip designers, the companies that draw the blueprints for graphics processors, central processors, and custom accelerators. Below them sit the fabricators, the companies that actually etch those designs onto silicon. Then there are the equipment makers that build the machines the fabricators use, and finally the memory makers that produce the DRAM and other chips every system needs.
The AI buildout has concentrated enormous value in two places in particular. The first is the accelerator layer, where one company holds a commanding share of the market. The second is the memory layer, where just three companies control the supply. We walk through the pitch that frames all of this in our review of the Elon Musk 10X Project.
Memory is the chokepoint
The less obvious part of the semiconductor story is memory. Most investors focus on the processing chip, the part with the highest profile. But every accelerator needs fast memory sitting right beside it, and as chips have gotten faster, the memory requirement has grown faster still.
That is why memory has become one of the tightest chokepoints in the entire hardware stack. Each new generation of accelerator carries more high-bandwidth memory, and the companies that make that memory have ridden the same demand wave as the chip designers. Our primer on HBM memory explains how that specific product works and why it matters so much.
Where the 10X Project tease fits
The “Elon Musk’s 10X Project” pitch from Brownstone Research runs directly through this memory chokepoint. The presentation’s “silent partner” is Micron Technology (MU), the only major US-based maker of DRAM and high-bandwidth memory. Micron is vertically integrated, handling everything from chip design through fabrication in its own facilities.
Micron is a legitimate, well-run AI-memory leader. It trades at roughly 15 times forward adjusted earnings and sits inside a three-way oligopoly with SK Hynix and Samsung. That is a real investment thesis, the same one we unpack across the wider Brownstone Research catalog. The catch is in the framing: Micron is a roughly trillion-dollar company, not the “little-known supplier” the pitch describes.
The cyclical catch
The other thing to know about semiconductor stocks, and memory stocks in particular, is that the industry is cyclical. DRAM has spent decades swinging between oversupply and shortage. When prices are high, all three memory makers expand capacity, which eventually floods the market and crushes prices, which then forces cutbacks and starts the cycle again.
That boom-and-bust rhythm does not invalidate the AI thesis. It does mean that buying a semiconductor stock at the top of the cycle is a different decision from buying it at the bottom. The pitch’s “+911%” figure is a backward-looking scoreboard, not a promise about what happens next. Investors who understand the cycle, and where the specific company sits inside it, are in a much better position than investors who buy on a teaser.
How to evaluate a semiconductor stock
The useful question with any semiconductor stock is not whether the technology is real. It usually is. The question is whether the company can turn that technology into durable earnings, and whether today’s price already reflects the answer.
Three checks matter most. First, where the industry sits in its cycle, since buying at the peak of a shortage is a different bet from buying at the bottom. Second, pricing power, which in memory comes from the fact that only three companies control the supply. Third, the design wins that lock in revenue for the next product generation. A company that earns a spot in the next accelerator generation has revenue visibility that its rivals do not.
None of these checks is secret. They are the same filters professional analysts apply, and they beat trying to guess which “hidden supplier” a teaser is describing.
The bottom line
Semiconductor stocks are the picks-and-shovels of AI, and the memory layer is one of the most concentrated parts of that supply chain. The 10X Project pitch points at a real company and a real chokepoint, but it wraps both in a “secret supplier” story that no longer describes reality. The opportunity is real; the mystery is not.
Ready to see the research? Click here to access the Near Future Report.
NewsletterVetter is an independent publication. We receive compensation from some of the services we review through affiliate links. Nothing on this site is investment advice. Always do your own research.