AVGO
The AI story is changing lanes
Broadcom is heading into its fiscal third-quarter 2026 earnings release after the bell on September 2nd, and the stock is basically shrugging its shoulders for now. But BakerAvenue Wealth Management is making the case that this is exactly the kind of setup Broadcom likes: not the flashy, headline-grabbing training phase of AI, but the boring-with-a-big-payoff phase known as inference.
That’s the part where companies actually run AI models across real products and workloads. In other words: once the demo is over, somebody has to pay the electric bill.
Why Broadcom, not just Nvidia?
King Lip of BakerAvenue framed the AI market like a relay race. Nvidia has been the monster in the training leg, but Broadcom is better positioned for the repetitive, high-volume workloads that show up once AI is deployed at scale. The key investor takeaway here is that Broadcom doesn’t need Nvidia to stumble for its own opportunity to grow. That’s a nice feature when you’re shopping for AI exposure and don’t want to bet the whole house on one horse.
Broadcom is also carrying serious expectations into the print:
- adjusted EPS is expected around $3.24
- revenue is expected around $29.43 billion
- the market is watching for clues on how much of that AI demand is real, durable, and not just hype with better branding
The market is getting pickier
Lip also said the AI trade is shifting from “show me the model” to “show me the money,” which is basically Wall Street’s version of a group chat getting tired of excuses. Investors want returns now, not just giant capex budgets and glossy slides.
That matters because Broadcom sits in the sweet spot between custom chips, AI infrastructure, and the kind of repetitive workloads that can turn into sticky, long-term demand. If inference keeps growing the way bulls expect, AVGO could have more room to run even if the broader AI trade gets choosier.
Big picture
This isn’t just a Broadcom story; it’s a reminder that the AI winners list may expand beyond the obvious chip darlings. Training got the hype, but inference could be where the recurring revenue party actually starts.