WHAT HAPPENED
This week, Amazon's Andy Jassy announced the company will spend $200 billion on capital expenditure in 2026—almost entirely on AI infrastructure. That's not a marketing number. That's real metal, fiber, and chips being deployed right now. And Amazon's just one player. Across the industry, hyperscalers plan to spend $611 billion on capex in 2026, with projections putting the global AI market at $4.8 trillion by 2033.
Meanwhile, networking stocks are quietly printing money. Broadcom (ANET) got upgraded to Buy this week with a $180 price target. Why? Because back-end data center switching is in "full-blown hypergrowth mode"—up 224% year-over-year to a $3.7 billion market. The Nvidia chips get the headlines. The networking switches? They're the actual bottleneck.
WHY IT MATTERS
During the 1849 Gold Rush, fortunes weren't made digging for gold. They were made selling shovels, pickaxes, and jeans to the miners.
We're in the AI Gold Rush. Everyone's obsessing over whether to own NVDA or bet on OpenAI's valuation. Smart money is watching who's actually building the infrastructure these companies require to operate.
Consider the economics:
Every AI model needs thousands of GPUs
Those GPUs need to talk to each other at insane speeds
That requires specialized networking hardware (Ethernet over InfiniBand)
Companies like Cisco (CSCO), Arista (ANET), and CoreWeave (CORW) are winning here
CoreWeave, a GPU cloud provider, is up 28% YTD. Not because it's a consumer darling. Because it's essential infrastructure for people building AI at scale.
The spread is important too: not all infrastructure bets are equal. Cisco is gaining cloud market share. Arista is losing some ground at cloud hyperscalers but winning in data center switching broadly. This is actionable differentiation.
WHAT TO WATCH
Earnings seasons for infrastructure plays – Watch ANET, CSCO, and semiconductor tooling companies. Guidance on AI capex spending tells you if this is a one-year spike or a decade-long trend. (It's the latter.)
IPO activity in GPU cloud providers – CoreWeave's recent pop shows appetite for pure-play infrastructure. More IPOs = more differentiation. Pick the winners, avoid the hype.
Supply chain tightness – Networking gear has long lead times. If you see shortages or price increases, infrastructure companies have pricing power. That's early money.
Spending announcements from META, GOOG, and TSLA – Amazon's $200B is the headline. But Meta and Google are spending heavily too. Each announcement validates the trend.
THE PRACTICAL MOVE
If you're not a stock-picker, grab an AI infrastructure ETF (like the iShares AI Innovation ETF, but weighted toward infrastructure). If you want to pick individual names, diversify across Cisco (safety), Arista (growth), and a smaller position in a pure-play like CoreWeave.
The myth: AI fortunes go to the companies building the models.
The reality: The steadiest money goes to the people selling the shovels.
DEEP DIVE
Ready to understand AI capex cycles and what they mean for your portfolio? Read our full breakdown on overtimealpha.com.
TOOLS & RESOURCES
Get our AI Infrastructure Stock Watchlist (updated weekly with alerts) on Gumroad.
Position sizing: Never more than 5% per position. Diversification beats being right about timing.