|
| What are this week’s market risks or opportunities? |
|
|
| Among thousands of conversations shared with WarrenAI this week, one question kept coming up in the AI semi names: should I rotate out of Micron into Broadcom or ASML for better upside? |
|
| People are framing it as a rotation, but the data says it isn’t exactly. |
|
| Nvidia reported Wednesday with Data Center revenue of $75.2B, up 92% YOY. AI infrastructure spending by the major cloud providers (Microsoft, Google, Amazon, Meta, Oracle) is now running around $725B for 2026. The pie is growing faster than any single vendor can capture, opening room for names beyond Nvidia to take real share without Nvidia losing any. |
|
| The harder question is whether end-user revenue ever justifies the spend. Consulting firm Bain puts it at $2 trillion in annual AI revenue to support $500 billion in capex. That gap hasn’t closed. So the real rotation is about whether the spend ever pays back, and it’s worth running against your own watchlist. |
|
|
|
|
|
|
| Which trades should I consider? |
|
|
| The rotation conversation keeps circling the same four names, each playing a different role in the AI stack. |
|
| Nvidia is the company that built the market for AI compute and the most widely held trade in this sector right now. |
|
| Micron (MU) makes the memory chips that go into AI servers, but its revenue rises and falls with memory prices as much as with AI demand, which is why we're seeing people move out of it this week. |
|
| ASML is the only company that makes EUV lithography equipment, the machines required to manufacture every advanced AI chip in this story. |
|
| Broadcom (AVGO) is the one worth understanding before you buy. More on that below. |
|
|
|
| Should I consider this trade? |
|
|
| Broadcom is up 81.85% in the past year and performance chasers drove most of that move without a thesis. |
|
| The true catalyst is custom silicon. |
|
| Google, Amazon, Microsoft, and Anthropic are all designing their own chips now, built for their specific AI workloads, because custom silicon costs less to run at scale than off-the-shelf GPUs. |
|
| Broadcom is the company they’re working with to build it. |
|
| Their deal with Anthropic alone accounts for 3 gigawatts of AI infrastructure demand, a multi-year commitment with years of execution ahead. |
|
| Broadcom reports on June 3. Big earnings weeks separate the investors who own it for a reason from the ones chasing the chart. Two people can hold the same shares without having the same conviction to hold them through a bad week. |
|
|
|
|
|
|
| Data correct to 21.05.2026 |
|
