Picks-and-shovels investing means owning the suppliers a trend depends on instead of guessing which end product wins. Here is how it worked in the fiber boom and the AI build-out, which suppliers led and which lagged, and the risks that come with it.
Which stocks do Wall Street’s best‑ranked analysts agree on? See them in Exent Gems →
It is investing in the companies that sell the tools a boom needs, instead of the companies competing to win the boom. The name comes from the idea that in a gold rush, sellers of picks and shovels get paid whoever finds the gold. In markets, the “shovels” are equipment makers, component suppliers, networks and infrastructure builders.
The appeal is that you don’t need to pick the winning app, carmaker or AI model. Every competitor has to buy chips, memory, networking gear and power. The catch is that suppliers’ sales depend on their customers’ spending, which swings hard from boom to bust.
Very well on the way up, and very badly on the way down. The late-1990s internet build-out was a classic picks-and-shovels trade: networking equipment, optical components and fiber. Cisco, the leading maker of internet routers, briefly became one of the world’s most valuable companies in March 2000. According to Sherwood News, its shares then collapsed nearly 90%, bottoming in late 2002, and did not close above the March 2000 high of $80.06 until December 10, 2025.
The internet itself was real and grew for decades. The suppliers were priced for spending that slowed sharply once the build-out paused. That is the central lesson: a correct view of the trend does not protect you from paying too much for the suppliers.
By mapping the supply chain: what does every AI data center have to buy? The layers include chip designers, chip-making equipment, memory, networking and optics, and the power and construction work to run the buildings. The table shows how companies in those layers did over the 10 years to Sep 30, 2026, in Exent’s data. Past results, not recommendations.
| Company | Layer | 10-year total return | Worst drop in 10 years | 2022 |
|---|---|---|---|---|
| NVIDIA (NVDA) | AI chips | +13,575% | −66.4% | −50.3% |
| Micron (MU) | Memory | +6,043% | −57.8% | −45.9% |
| Lam Research (LRCX) | Chip equipment | +3,834% | −56.8% | −40.7% |
| Arista Networks (ANET) | Data center networking | +3,727% | −52.2% | −15.6% |
| Quanta Services (PWR) | Power grid construction | +2,240% | −45.6% | +24.6% |
| Lumentum (LITE) | Optical components | +2,225% | −66.9% | −50.7% |
| Applied Materials (AMAT) | Chip equipment | +1,786% | −55.4% | −37.5% |
| Ciena (CIEN) | Optical networking | +1,514% | −49.5% | −33.8% |
| Intel (INTC) | Chips | +293% | −73.5% | −46.6% |
| Equinix (EQIX) | Data center REIT | +239% | −42.8% | −21.1% |
| Digital Realty (DLR) | Data center REIT | +148% | −51.1% | −41.0% |
| S&P 500 (SPY) | Index | +311% | −34.1% | −18.2% |
Besides Nvidia, six of the suppliers in the table (Micron, Lam Research, Arista, Quanta, Lumentum and Applied Materials) were among the 20 best 10-year performers of today’s 500 largest US stocks. Across all semiconductor stocks in the data with 10 years of history, 25 of 35 beat the S&P 500. Being in the supply chain was not enough, though. Intel, a chip maker, and the two big data center landlords trailed the index. More on the winners in how many stocks beat the S&P 500, and on the theme in AI stocks vs the S&P 500.
It depends on the stage of the trend. Early in a build-out, customers spend heavily on capacity before the products earn money, so suppliers’ sales and profits rise first. Later, if spending slows or overshoots, suppliers feel it first too, because a customer can pause orders much faster than it can shut down a product. End-market winners, when they emerge, tend to have steadier and longer-lasting revenue, but they are much harder to identify in advance.
In the fiber era, suppliers rose fastest and fell hardest. In the AI build-out to Sep 2026, many suppliers led. Neither history says what the next five years will look like.
Investors who use this approach often size the theme as a whole, not stock by stock. See how much AI exposure is in your portfolio and thematic ETFs for fund-based versions.
Investing in the suppliers a trend depends on, such as equipment makers, component suppliers and infrastructure builders, instead of betting on which end-product company will win.
They map what every competitor must buy: chips, chip-making equipment, memory, networking, optics and power. Then they check each supplier’s customer concentration, cyclicality and valuation.
Many did over the 10 years to Sep 2026: Micron, Lam Research and Arista Networks were among the 10 best large US stocks. Others in the supply chain, such as Intel and data center REITs, trailed the S&P 500.
Not necessarily. Suppliers are often more cyclical, because customers can cut orders quickly. Several AI suppliers fell 40% to 51% in 2022, and Cisco fell nearly 90% after the dot-com peak.
Exent tracks 4,000+ analysts, keeps only the elite few with proven records, and surfaces the Gems where they agree.
Educational content, not investment advice. Examples are illustrative; past performance does not guarantee future results.