Broadcom Q2 Earnings Prep: Racks on racks on racks
Broadcom (AVGO) has always been the company that sells the unglamorous half of every semiconductor system. Then came AI, and that unglamorous half -- custom accelerators, networking switches, the optics between racks -- became critical to everyone with a frontier model. Q2 results Wednesday will be about whether the customer pile-up that drove AI revenue to double last quarter is still expanding, or whether the financing behind the racks Broadcom is now shipping has started to slow down.
Broadcom generates revenue through two segments: Semiconductor Solutions ($36.86B in FY25, 58% of consolidated revenue), which sells custom AI accelerators (”XPUs”) and Ethernet networking silicon to a handful of hyperscale customers alongside legacy wireless, broadband, and storage businesses, and Infrastructure Software Solutions ($27.03B, 42%), which is largely the VMware portfolio acquired in late 2023, repackaged into VMware Cloud Foundation subscriptions. Two years ago, before VMware closed, the mix was closer to 80%/20% Semi/Software. Then VMware flipped it, and now AI is flipping it back: Q1 FY26 semi revenue grew $8.21B → $12.52B (+52% YoY) while software stayed near flat sequentially.
The strategic response from Hock Tan has been to guide custom-silicon customers from chip-only sales toward XPU-based rack systems that bundle Broadcom networking; keep the cost base disciplined while doing it, with headcount going 37,000 → 33,000 between FY24 and FY25 as revenue grew $51.57B → $63.89B (+24%); and finance the customers when they can’t finance themselves.
This last point should jump out. The latest 10-Q flags it in its own language as “alternative financings or novel or deferred payment models” for AI racks. If there are cracks in the overall AI infrastructure buildout happening across the industry, as the morbid rubberneckers among us are on high alert for, this is a place they would show up.
So heading into next week, here’s what we’re watching:
AI Semiconductor Revenue. Q1 was a record $8.4B (+106% YoY) and management’s own Q2 guide is $10.7B, which would be another +27% sequential and roughly half of total company revenue from a single product line. Look for whether the guide gets raised on the call. Hock Tan has framed a $100B AI revenue trajectory by FY27, and the gap between current run-rate and that bar will narrow or widen with this report.
The $45B Multi-Year Backlog. Firmly committed multi-year RPO across Semi and Software went $22.4B (Q1 FY25) → $45B (Q1 FY26), a doubling that lines up with OpenAI joining the named XPU customer list as the sixth name alongside Google, Meta, Anthropic, and two unnamed hyperscalers. Over the same span the
12-month recognition share fell 40% → 33%. Backlog roughly doubling while duration stretches is bullish if you trust the customers and a sign of revenue being pushed out if you don’t. If the multi-year number ticks up again, pay attention to revenue recognition to understand if it is a sign of health or a sign of dollars being pushed out.
Customer Concentration at 50%. The top five end customers were 50% of Q1 FY26 revenue, up from 40% a year earlier, and the single largest semi customer was 32% of consolidated revenue in FY25 vs 28% the year before. This reflects the shape of the AI capex build among hyperscalers. Listen for any commentary, e.g. that frames the 6th customer’s contribution or talks about diversifying the buyer set, to glean signals on the overall industry build-out.
Gross Margin in the Rack Era. Consolidated gross margin recovered 63% (FY24) → 68% (FY25) as the VMware integration costs annualized away. The Q1 10-Q itself now flags that selling “AI racks or systems based on our XPUs will likely increase our operating margin but compress or lower future gross margin.” That is the trade Tan has chosen. Look for Q2 consolidated gross margin below 67% as the early signal that the rack-system mix is starting to bite.
The OpenAI Financing Footnote. Bloomberg reported on May 7 that OpenAI is having trouble closing roughly $18B of financing tied to its Broadcom custom-silicon partnership. The 10-Q risk factors now explicitly call out “credit or customer default risks” from AI customers who cannot fund their own infrastructure. None of this changes Q2 reported revenue, since the chips will ship regardless of who pays at the rack level. But any commentary on customer payment terms, factoring, or vendor financing exposure is critical and has the potential to create headlines.
Consensus is roughly $22.1B of revenue (+47% YoY) and $2.40 of adjusted EPS, in line with management’s own $22B guide. The stock sits near $423 after roughly doubling over twelve months, with Susquehanna raising its target to $490 last week. Look for details from Tan about how the $45B backlog will get paid for and whether the customer set will keep expanding. AVGO 0.00%↑ is the company that figured out how to be Nvidia-adjacent at a fraction of the visibility risk -- with less pomp and circumstance, their data will have a lot to tell us about the broader market.
All data sourced via Revelata’s deepKPI from American Eagle Outfitters’ 10-K, latest 10-Q, and quarterly 8-K press releases, with additional context from analyst reporting and peer SEC filings.