Joining me on today's call are Hock Tan, President and CEO; Kirsten Spears, Chief Financial Officer; Charlie Kawwas, President, Semiconductor Solutions Group; and Ram Velaga, President, Infrastructure Software Group. In our fiscal Q1 2026, total revenue reached a record $19.3 billion, and that's up 29% year-on-year and exceeding our guidance on the back of better than expected growth in AI semiconductors. This top-line strength translated into exceptional profitability with Q1 consolidated adjusted EBITDA hitting a record $13.1 billion, which is 68% of revenue. Looking ahead to next quarter, Q2 2026, we're guiding for consolidated revenue of approximately $22 billion, which represents 47% year-on-year growth.

In Q1, revenue was a record $12.5 billion, as year-on-year growth accelerated to 52%. This robust growth was driven by AI semiconductor revenue, which grew 106% year-on-year to $8.4 billion, way above our outlook. Driving this is AI revenue growth, which will accelerate very sharply to 140% year-on-year to $10.7 billion. For Google, we continue our trajectory of growth in 2026 with strong demand for the seventh generation Ironwood TPU.

In 2027 and beyond, we expect to see even stronger demand from next generations of TPU. For 2027, this demand is expected to surge in excess of 3 GW of compute. We have the track record to deliver these XPUs at high volumes at an accelerated time to market with very high yields. Consistent now with the strong outlook for our XPUs, demand for AI networking is accelerating.

What went well
  • Record total revenue of $19.3 billion, up 29% year-on-year and exceeding guidance on better-than-expected AI semiconductors.
  • Record consolidated adjusted EBITDA of $13.1 billion at 68% of revenue, above the 67% guidance.
  • Record semiconductor revenue of $12.5 billion, up 52% year-on-year, driven by AI.
  • AI semiconductor revenue grew 106% year-on-year to $8.4 billion, well above outlook; custom accelerator business grew 140% year-on-year.
  • Added a sixth XPU customer (OpenAI), expected to deploy over 1 GW of first-generation XPU compute in 2027.
  • AI networking revenue grew 60% year-on-year to one-third of total AI revenue, gaining share on Tomahawk 6 and 200G SerDes.
  • Line of sight to over $100 billion of AI (chip) revenue in 2027, with supply chain secured for 2026 through 2028.
  • Semiconductor operating margin of 60%, up 260 basis points year-on-year on strong operating leverage.
  • Free cash flow of $8 billion; returned $10.9 billion to shareholders ($3.1 billion dividends, $7.8 billion buybacks) and authorized an additional $10 billion repurchase.
What went wrong
  • Non-AI semiconductor revenue was flat year-over-year at $4.1 billion, with a seasonal decline in wireless.
  • Infrastructure software revenue grew only 1% year-on-year to $6.8 billion on seasonal renewals (though VMware itself grew 13%).
  • Consolidated gross margin was 77% and guided flat sequentially, with no expansion.
  • Non-GAAP tax rate elevated at ~16.5% on global minimum tax and geographic income mix.
  • Inventory days rose to 68 (from 58 in Q4) as the company pre-positioned components for AI growth.

Guidance Changes

MetricPeriodCurrent guidance
Consolidated revenueQ2 FY2026$22.0B, up 47% YoY
Semiconductor revenueQ2 FY2026~$14.8B, up 76% YoY
AI semiconductor revenueQ2 FY2026$10.7B, up ~140% YoY
Infrastructure software revenueQ2 FY2026~$7.2B, up 9% YoY
Non-AI semiconductor revenueQ2 FY2026~$4.1B, up 4% YoY
Consolidated gross marginQ2 FY2026~77%, flat sequentially
Adjusted EBITDA marginQ2 FY2026~68% of revenue
Non-GAAP tax rateQ2 / FY2026~16.5%
AI (chip) revenueFY2027Line of sight to in excess of $100 billion, supply secured

Performance Breakdown

MetricYoYNote
Total revenue +29% Better-than-expected AI semiconductor growth and operating leverage.
AI semiconductor revenue +106% Ramp of custom AI accelerators across all five (now six) customers plus AI networking.
Semiconductor revenue +52% AI-driven acceleration to a record $12.5 billion.
Custom accelerator (XPU) revenue +140% Strong demand including Google's 7th-gen Ironwood TPU and other customer ramps.
AI networking revenue +60% Tomahawk 6 (100 Tbps) and 200G SerDes capturing hyperscaler demand; one-third of AI revenue.
Infrastructure software revenue +1% Seasonal renewals; VMware revenue itself up 13% with ARR up 19%.
Operating income +31% Record $12.8 billion; operating margin 66.4%, up 50 bps YoY.
Semiconductor operating margin +260 bps Strong operating leverage; margin 60%.

Earnings Call Themes & Trends

TopicPrevious mentionCurrent periodTrend
Custom XPU customer count5 customers6 customers with OpenAI added, deploying over 1 GW in 2027
Anthropic compute$10B + $11B TPU ordersOff to a good start on 1 GW of TPU compute in 2026; demand to surge above 3 GW in 2027
2027 AI revenue visibility$73B AI backlog / over $100B impliedExplicit line of sight to over $100 billion of AI chip revenue in 2027, supply secured through 2028
Customer-owned tooling (COT) threatCalled an overblown hypothesisReiterated; COT programs are ~2x less performant, no meaningful competition expected for many years
Gross-margin / rack dilutionWarned of AI/system-sale gross-margin dilution (Q4)Reversed - it is a chips business, margins solid and stable; dilution impact not substantial
AI networking mixOne-third of AI revenue (Q1)Guided to ~40% of AI revenue in Q2
Scale-up interconnectCopper direct-attach at 200G SerDesStaying on Direct Attach Copper through 400G SerDes (2028), avoiding costlier/higher-power optics; CPO not needed yet

Q&A Summary

Clarify the greater-than-$100 billion 2027 figure, and reconcile hyperscaler ROI pessimism with your share-gain outlook.
The $100B+ is all chip content (XPUs, switch chips, DSPs). Customers are the few LLM/platform players (some hyperscalers, some not); inference to monetize models is driving surprisingly strong, sustained compute demand.
Do customer-owned-tooling (COT) projects take meaningful XPU share, and how do you widen the 12-18 month lead?
COT players face huge technology hurdles (silicon design, SerDes, advanced packaging, networking); they cannot afford a merely good-enough chip competing with NVIDIA and peers. Broadcom's high-volume, high-yield time-to-market keeps it far ahead; no meaningful COT competition for years.
What drives AI networking to 40% of AI revenue and how does the mix evolve in the $100B+?
Tomahawk 6 (only 100 Tbps switch) plus 1.6 Tb DSPs are seeing huge demand, growing networking even faster than XPUs near-term; expect networking to settle around 33-40% of AI revenue.
How does disaggregating prefill/decode affect custom vs GPU silicon demand?
Workload specialization favors XPUs - purpose-built silicon (e.g., sparse cores for mixture-of-experts) outperforms general-purpose GPUs; customers increasingly design separate training and inference chips.
Guardrails on gross margin as racks ship - could it fall ~500 bps with a floor?
That view is mistaken; gross margin stays solidly at reported levels. AI product mix will not substantially affect it - yields and costs make AI margins consistent with the rest of semiconductors.
Is the 2027 gigawatt math (~8-9 GW, ~$20B/GW content) correct?
Gigawatts is the right lens; dollars per gigawatt vary by customer but are in that ballpark. By gigawatts, 2027 is getting close to 10 GW.
How did you secure supply visibility through 2028, and can you grow in 2028?
Early lock-up of T-glass, substrates and other components with strong partners, informed by customers sharing 2-4 year custom-silicon roadmaps; yes, growth expected in 2028.
How do you keep visibility and share across fragmented multi-supplier customers?
Only six customers, each treating custom silicon as a strategic multi-generation play (not optional); Broadcom has clear visibility into their roadmaps, with GPU/neocloud usage being transactional 'optionality.'
Are the non-TPU custom engagements mostly inference, and what are their GPU cost/performance advantages?
Most customers start with inference (less compute, cheaper, lower power than GPUs), then extend to training; maturing customers increasingly build two specialized chips per year (training and inference).

More on Broadcom Inc.

Reported 2026-03-04 · figures from the Broadcom Inc. Q1 2026 earnings call.

See how VectorShift works for your firm

Request Demo