Broadcom Earnings Preview: The AI ASIC Bar Is About to Move Again

Broadcom does not report like a quiet component supplier anymore. When the company releases Q2 FY2026 results after the close on Wednesday, June 3, investors will not be grading only revenue, VMware integration, or the dividend. They will be grading the price of custom AI compute.

The old Broadcom story was scale, margins, and disciplined acquisitions. The new one is sharper: can Hock Tan keep turning hyperscaler AI budgets into custom accelerators, networking silicon, and cash flow without letting the valuation get ahead of the order book?

That is why this print matters. In March, Broadcom guided Q2 revenue to about $22.0 billion and said it expected Q2 AI semiconductor revenue of $10.7 billion. Those are not background numbers. They are the bar. On the TECHi AVGO quote page, the stock was recently snapshot at $481.57, up 4.70% on June 2, with a $2.12 trillion market cap and a consensus target sitting almost exactly on top of the latest price. The market is already leaning forward.

Article Brief

Key Takeaways

5 points30s read

1. 01The number to beat–Broadcom guided Q2 FY2026 revenue to about $22.0B and Q2 AI semiconductor revenue to $10.7B. A clean beat matters less than whether management raises the next AI bar.
2. 02The AI ASIC question–Custom AI accelerators are moving from side story to main story. Investors need evidence that demand is broadening beyond a small set of hyperscaler programs.
3. 03Alphabet changes the backdrop–Alphabet said it plans to raise $80B for AI compute infrastructure, including $10B from Berkshire Hathaway, according to AP. That puts a fresh dollar sign on hyperscaler demand.
4. 04Valuation is tight–TECHi quote data shows AVGO near its analyst target, with forward P/E around 33.2x and a 52-week position at the top of the range. Good numbers may not be enough.
5. 05Watch the non-AI cushion–VMware cash flow, infrastructure software margins, and capital returns still matter because they are what keep Broadcom from trading like a pure AI-cycle stock.

The Setup: Broadcom Has to Beat Its Own AI Bar

The awkward thing about Broadcom right now is that the company has already told investors where to look. In its Q1 release, Broadcom reported $19.31 billion of revenue, up 29% from the prior year period. AI revenue was $8.4 billion, up 106% year over year, driven by custom AI accelerators and AI networking. Then management moved the Q2 AI semiconductor guide to $10.7 billion.

That makes the June print less about discovery and more about confirmation. If AI semiconductor revenue comes in around the guide, the stock still needs commentary that says the next six to twelve months are better than the last six. If it comes in above the guide, investors will ask how much of the beat came from pull-forward, one customer, or networking attach. If it misses, even modestly, the market will not treat it like a normal semiconductor wobble.

There is a reason the reaction function is harsher now. AVGO has graduated into the mega-cap AI infrastructure cohort. A $2 trillion stock does not get rewarded for proving that AI is real. It gets rewarded for proving that AI revenue has become durable, margin-rich, and less dependent on one customer calendar.

Alphabet Just Made the Demand Signal Louder

The cleanest reason to care about Broadcom into this report is not a single earnings line. It is the buyer side of the market. AP reported Monday that Alphabet said Berkshire Hathaway agreed to buy $10 billion of Alphabet stock as part of a broader plan to raise $80 billion for computing infrastructure needed to power its AI offerings.

Broadcom does not need to say the word Alphabet for that to matter. The point is broader: hyperscalers are still trying to buy more compute than their existing cash flows and data center schedules can comfortably absorb. When a company with Alphabet scale is willing to sell equity to fund AI infrastructure, the supply chain should hear it as a demand signal for custom silicon, networking, packaging, power, and every bottleneck around inference.

That is where Broadcom sits. NVIDIA owns the general-purpose GPU layer; Broadcom is increasingly the custom silicon and connectivity layer for buyers that want lower cost per inference, tighter control over workloads, and less dependence on merchant GPU supply. The ASIC story is not an anti-NVIDIA trade. It is the next branch of the same AI capex tree.

What the Call Has to Answer

A headline beat would be useful, but it would not settle the stock. The conference call has to clear four harder questions.

1. Is the ASIC customer base broadening?

Broadcom has been careful about customer names, and that is normal for custom silicon. Investors still need evidence that revenue visibility is expanding across more programs and more generations. One large ASIC ramp can create impressive growth. Several large ASIC ramps can create a franchise.

The market will listen for customer-count language, multi-year program references, and anything that suggests the next AI revenue leg is not just one hyperscaler moving one internal chip family through the channel.

2. Is AI networking keeping pace with accelerators?

Custom accelerators get the cleaner headline, but networking is the tell. Training and inference clusters do not scale because a buyer adds more chips in isolation. They scale when the fabric lets those chips behave like one machine. That is why Broadcom commentary on switching, optics, Ethernet, and AI networking attach deserves as much attention as accelerator revenue.

If management says accelerator demand is strong but networking grows slower, the market may read that as narrower exposure. If networking accelerates with the ASIC programs, the story becomes more powerful because Broadcom is taking a larger share of the cluster bill of materials.

3. Is VMware still an earnings cushion?

The stock is being priced around AI, but VMware is still the margin ballast. Infrastructure software was 35% of Q1 revenue, and the entire Broadcom model depends on software cash flow smoothing the semiconductor cycle. Investors will want retention, renewal, margin, and integration commentary that shows VMware is not being used to hide volatility elsewhere.

This matters more than it sounds. If ASIC revenue gets lumpy, Broadcom can still deserve a premium multiple if software cash flow holds. If VMware starts to look tired at the same time the AI guide gets less aggressive, the multiple argument changes quickly.

4. How much of the guide is already in the stock?

TECHi data shows the difficult part of the setup. AVGO has a Buy consensus from 56 analysts, but the tracked consensus target of $481.97 was barely above the latest $481.57 snapshot. That does not mean analysts are bearish. It means the stock has outrun the old target deck.

That is the price of success. The forward P/E near 33x is not absurd for a company with Broadcom margins, software cash flow, and AI growth. It is also not cheap enough to forgive vague guidance. The call needs to give analysts a reason to move numbers, not just a reason to keep their Buy ratings.

The Bull Case Into the Print

The bull case is straightforward and still credible. Broadcom is one of the few companies that can turn AI infrastructure spending into high-margin semiconductor revenue and software-like cash conversion. Q1 free cash flow was $8.01 billion, and adjusted EBITDA was 68% of revenue. That is not a speculative AI supplier profile. That is a cash machine with an AI accelerator attached to it.

If Q2 AI semiconductor revenue clears $10.7 billion, if management raises the second-half AI run-rate, and if VMware remains stable, the market will have to think about Broadcom less like a cyclical chip stock and more like an AI infrastructure platform. In that case, the current valuation could still be defensible, especially if custom silicon becomes a recurring multi-generation relationship with the largest cloud buyers.

The Bear Case Is Not That AI Goes Away

The bear case is subtler. It is not that AI demand disappears. It is that Broadcom’s AI revenue becomes too concentrated, too lumpy, or too well understood by the market to keep surprising. Once investors can model the ASIC ramps cleanly, the multiple starts depending on growth duration rather than growth revelation.

There is also the customer-economics question. Hyperscalers are spending aggressively because they believe AI demand will pay them back through cloud growth, ads, subscriptions, enterprise tools, and internal productivity. If those returns are slow to show up, buyers may still spend, but they may push harder on supplier pricing, architecture control, and second-source leverage. Broadcom is powerful in that negotiation. It is not immune to it.

What Would Be Enough?

For the stock to move cleanly higher after earnings, Broadcom probably needs three things at once: an AI semiconductor number above the $10.7 billion guide, language that extends the ramp into the back half of fiscal 2026, and software commentary that keeps the cash-flow story intact. One or two of those may prevent damage. All three can reset the target deck.

The strongest version of the report would sound something like this: custom accelerator demand is still exceeding supply, AI networking is attaching to more cluster builds, customer programs are moving into later generations, and VMware remains a margin engine rather than a distraction. That would make the Alphabet capital raise feel less like a headline and more like a map of where Broadcom revenue can go next.

The weaker version is a clean Q2 with a careful guide. That might be fundamentally fine and still trade poorly. At $481, investors are not paying to learn that Broadcom is participating in the AI buildout. They are paying to learn that the ASIC bar is going up again.