Broadcom Earnings Preview: Date, Time, and 4 AI Numbers to Watch

Broadcom’s last earnings report contained almost everything investors normally claim to want.

AI semiconductor revenue increased 143% year over year to $10.8 billion. Total revenue grew 48% to $22.2 billion. Adjusted earnings reached $2.44 per share, and free cash flow topped $10 billion for the quarter.

Shares still fell roughly 13% following the print.

The market was no longer judging Broadcom against ordinary expectations. An enormous AI ramp was already priced in, and the company’s forecast for $16 billion in third-quarter AI semiconductor revenue landed slightly below the most aggressive whisper numbers. Management also warned that a shifting revenue mix would naturally weigh on gross margins.

Expectations remain elevated heading into Wednesday's fiscal third-quarter report. Wall Street is looking for roughly $29.24 billion in revenue and $3.22 per share in adjusted earnings. Because Broadcom already guided to $29.4 billion in revenue, the consensus is essentially just asking the company to deliver what it promised.

That makes a simple headline beat or miss fairly useless. To get a real picture of where Broadcom’s business is heading, four specific AI metrics matter much more.

When Does Broadcom Report Earnings?

Broadcom is scheduled to release fiscal third-quarter 2026 results after the market closes on Wednesday, September 2, 2026. The earnings call begins at 5:00 p.m. Eastern Time (2:00 p.m. Pacific Time).

What Broadcom Actually Sells Into AI

Broadcom’s AI business sits in two distinct buckets.

The first involves custom AI accelerators. What Broadcom calls XPUs. Instead of selling off the shelf processors, the company partners with a handful of hyperscalers to design silicon tailored to specific workloads, most notably Google’s Tensor Processing Units.

The other side of the business is networking. Accelerators are largely useless unless they can move data rapidly enough to function as a unified computing system. Broadcom supplies the switches, optical components, and connectors that make those massive clusters possible.

Custom accelerators generate massive revenue figures, but they typically carry lower gross margins than Broadcom's networking products. A quarter heavily skewed toward accelerator shipments will naturally pair explosive revenue growth with a lower consolidated gross margin. Depending on the operating leverage supporting those sales, that isn't necessarily a deterioration in the business.

1. Q3 AI Semiconductor Revenue: $16 Billion

Broadcom called for roughly $16 billion in third-quarter AI semiconductor revenue. That would represent more than 200% growth from the $5.2 billion generated in the same quarter last year, and a 48% sequential jump from Q2.

The company has a solid track record of hitting these near-term AI forecasts. It guided to $8.2 billion for Q1 and delivered $8.4 billion, then forecasted $10.7 billion for Q2 and reported $10.8 billion.

A modest beat here, however, won't settle the broader debate surrounding the stock. Google recently expanded its custom-chip relationship with Marvell, heavily incentivizing Marvell to secure a larger portion of Google’s future TPU-ecosystem spending. It makes strategic sense for Google to diversify its supplier base as its compute demands scale. The real question is whether the overall AI infrastructure market is expanding fast enough for Broadcom’s absolute revenue to keep climbing even if it shares the hyperscaler pie. Clearing the $16 billion hurdle keeps that bullish thesis intact.

2. The Implied Q4 AI Target: About $21 Billion

Management expects to hit $56 billion in AI semiconductor revenue for the full fiscal year. Given the $8.4 billion generated in Q1 and the $10.8 billion in Q2, the math gets steep.

Assuming Broadcom hits its $16 billion Q3 guidance, it still needs approximately $21 billion in Q4 to fulfill that $56 billion promise.

That requires AI semiconductor revenue to jump another 30% sequentially in the final quarter. Broadcom carried more than $30 billion in AI semiconductor bookings during Q2, giving them massive visibility as customers line up chips, power, and data center space months in advance. But revenue is only recognized upon shipment. Guidance anywhere near $21 billion proves the second-half shipment ramp is executing on schedule.

3. FY2027 AI Semiconductor Revenue: More Than $100 Billion

Even if Broadcom clears $56 billion this year, getting above management's stated $100 billion target for 2027 will require at least 79% growth from a significantly enlarged base.

Broadcom plans to support this by helping deploy roughly 10 gigawatts of AI computing capacity during 2027, servicing core programs across Google, Anthropic, OpenAI, Meta, and others.

Investors need to separate theoretical demand from committed capital. Large AI labs often want more computing capacity than they can immediately finance or power. Broadcom’s recent 10-Q explicitly warned that alternative financing and deferred-payment models used to bridge this gap could expose the company to credit risk or drag on free cash flow. Simply reiterating the $100 billion target isn't enough; the market needs updated bookings and purchase orders to confirm that real capital is supporting these deployments.

4. AI Networking Mix: Approximately 30%

Networking made up nearly 40% of Broadcom’s AI semiconductor revenue in Q2. Management noted this was unusually high, pointing to a normalized run rate closer to 30%.

If networking falls toward 30% as lower-margin custom accelerator shipments surge, total AI revenue will look spectacular while consolidated gross margins sink. Broadcom has already modeled this, forecasting non-GAAP gross margin to fall from 77.1% in Q2 to roughly 74% in Q3.

Critically, management expects non-GAAP operating margins to hold steady around 67%.

This operating leverage is the key to the Broadcom thesis. Designing a custom chip requires heavy R&D upfront, but shipping twice as many completed units doesn't require doubling the headcount. If gross margins slip due to product mix while operating margins remain rock solid, the unit economics are working exactly as designed.

The Non-AI Business

Broadcom's legacy segments still have to fund the inventory and supply commitments driving the AI ramp. Management expects about $4.5 billion in non-AI semiconductor revenue (up 12% year-over-year) and roughly $8.9 billion in infrastructure software (up 31%). To support the second-half accelerator surge, Broadcom built its inventory up to $4.33 billion in Q2. We need to see that begin converting into cash.

A standard earnings beat may not satisfy the market this week. To reclaim its momentum, Broadcom may need to prove it can convert historic demand into a rapidly escalating shipment schedule without sacrificing operating margins or taking on unnecessary financing risk to get there.