Broadcom beats estimates, doubles AI chip sales to $8.2B, sets Q1 outlook at $19.1B

Broadcom's AI surge: what sales teams should act on now

Broadcom just posted a quarter that beat on earnings and revenue, and guided higher on the back of AI demand. The headline: AI chip sales are set to double year over year to $8.2 billion this quarter, driven by custom accelerators and AI networking semiconductors.

For sellers, this is a signal. Budgets are flowing into custom silicon, data center upgrades, and AI network infrastructure. That means new projects, fresh champions, and bigger deal sizes across hyperscalers, large enterprises, and AI-first startups.

The scorecard

• Earnings per share: $1.95 adjusted vs. $1.86 expected
• Revenue: $18.02 billion vs. $17.49 billion expected
• Q1 revenue guide: about $19.1 billion (28% YoY), above the $18.3 billion consensus
• AI chip sales outlook: $8.2 billion this quarter, doubling YoY
• Net income: up 97% to $8.51 billion ($1.74 per share)
• Segments: Semiconductor Solutions $11.07 billion (+22%); Infrastructure Software $6.94 billion (+26%)
• Dividend: $0.65 per share, up from $0.59
• Stock reaction: initial pop, then down more than 2% after hours; up ~75% year-to-date 2025 after doubling last year

What's driving the growth

• Custom AI accelerators (XPUs) for top buyers that want control over performance, cost, and supply.
• AI networking chips enabling higher-bandwidth clusters and faster interconnects.
• A $73 billion backlog across custom chips, switches, and data center parts over the next 18 months.

Customer momentum to watch

• Anthropic is using the latest Google TPU "Ironwood."
• Broadcom now counts five custom-chip customers; a new fifth customer placed a $1 billion order for delivery in late 2026.
• Large orders continue: previously disclosed $10 billion orders tied to custom chips and TPU demand.

Why this matters for sellers

• Budgets are expanding for AI infrastructure, not just GPUs-think accelerators, optical, switches, storage, power, and cooling.
• Procurement cycles are moving faster for anything that reduces training cost, speeds inference, or eases supply constraints.
• Custom silicon programs create multi-year, multi-vendor ecosystems. That's a lot of cross-sell surface area.

Account angles and signals

• Hyperscalers and cloud platforms: Custom accelerator roadmaps, new data center builds, and AI networking upgrades.
• AI-native companies: Training clusters, inference farms, and colocation expansions.
• Enterprises with rising AI workloads: Network fabric refreshes, storage upgrades, and power/cooling retrofits.
• Partners and OEMs: Integrations around switching, optical modules, and accelerator-ready servers.

Talking points for your next call

• Cost-to-train and cost-per-inference: Map your offer to throughput gains and lower total cost of ownership.
• Supply and time-to-deploy: Show how you help teams go live faster or de-risk timelines.
• Scalability: Prove your solution won't bottleneck custom chips and high-bandwidth fabrics.
• Operational simplicity: Highlight automation, observability, and support-busy infra teams care about fewer moving parts.

Objections you'll hear, and brief responses

• "We're standardizing on a single vendor." - Position as complementary: reduce vendor lock-in, improve resilience, and optimize for their chosen stack.
• "Budgets are locked." - Point to the shift from general IT to AI infra spend; reframe around ROI per watt, per rack, or per model.
• "We'll wait for next-gen chips." - Emphasize modular upgrades now (networking, storage, cooling) that compound benefits later.

Actions to take this quarter

• Prioritize accounts with custom accelerator programs, AI networking projects, or new colocation contracts.
• Build a short business case template around throughput, latency, and cost-per-token or cost-per-epoch.
• Partner with data center operators and OEMs to bundle end-to-end solutions.
• Track public backlog and order disclosures to time outreach around expansions and deliveries.

Context and resources

• Company investor materials often publish backlog and segment details: Broadcom Investor Relations
• Market data provider referenced in estimates: LSEG
• Want quick upskilling on AI buyer language and use cases for sales roles? Explore AI courses by job

Bottom line

Broadcom's numbers confirm the shift: AI spend is widening from GPUs to full-stack infrastructure. If you sell into data centers, cloud, or AI platforms, your pitch should tie directly to efficiency, speed, and scale-because that's where the checks are being written.