Broadcom targets $100 billion in AI chip sales by 2027, taking aim at Nvidia

Broadcom targets US$100B in AI chip sales by 2027 - what sales teams need to know

Broadcom CEO Hock Tan says the company has "line of sight" to more than US$100 billion in AI chip sales in 2027. He also said the supply chain is secured to support that growth. For enterprise buyers weighing Nvidia-first strategies, this is a clear signal: they'll have serious options.

Context matters. Broadcom projects US$10.7 billion in AI chip revenue this quarter, after reporting US$20 billion in AI sales last year. The company also issued a stronger-than-expected outlook for the fiscal second quarter at ~US$22 billion in revenue and approved a US$10 billion stock buyback.

Where Broadcom is winning

• Custom accelerators: An alternative to Nvidia for training and inference at scale, built around specific workloads.
• AI networking: Chips and interconnects to move data efficiently across clusters running large models.
• Hyperscaler traction: Work with OpenAI, Anthropic, Google (TPU collaboration), and Meta-plus a new processor generation slated for a half-dozen additional clients this year.

Key numbers you can use in conversations

• Target: US$100B in AI chip sales in 2027.
• Current quarter AI chip revenue: US$10.7B.
• Last year AI sales: US$20B.
• Fiscal Q2 revenue outlook: ~US$22B; buyback: US$10B.
• Fiscal Q1 (ended Feb. 1): sales US$19.3B; EPS (ex-items) US$2.05; AI revenue US$8.4B (more than doubled year over year).
• Capacity signals: OpenAI expected to ship Broadcom-based systems in volume next year at >1 GW capacity; Anthropic at ~1 GW this year and >3 GW next year; Meta roadmap "alive and well," scaling to multiple gigawatts in '27 and beyond.

Why this matters for sales teams

• Budget gravity is shifting: Enterprise spend is consolidating around AI infrastructure where performance per dollar and supply assurance win deals.
• Second-source demand is real: Many buyers want alternatives to reduce vendor risk. Broadcom's custom silicon plus networking gives you a credible "Plan B" or "Plan A for specific workloads."
• Capacity and delivery are front-line topics: Tan emphasized secured supply chain-use that to address availability concerns and project timelines.
• Attach and upsell chances: Networking, interconnects, and services around cluster design create multi-line opportunities.

Talking points that open doors

• Workload-fit over one-size-fits-all: Position custom accelerators as a way to hit target SLAs, energy budgets, and TCO for specific model families.
• Throughput and latency at scale: Tie networking semiconductors to faster training runs and more predictable inference under load.
• Roadmap stability: Quote the 2027 sales target and multi-gigawatt customer plans to show commitment and longevity.
• Reference momentum: OpenAI, Anthropic, Google TPU, and Meta activity signals maturity-not theory.

Objections you'll hear-and how to respond

• "Isn't there an AI bubble?" Acknowledge the concern, then anchor on operational demand: AI revenue more than doubled to US$8.4B last quarter, with stronger-than-expected guidance and visible shipments ahead.
• "We're standardized on Nvidia." Great-position Broadcom as complementary for specific workloads, cost targets, or as a strategic second source.
• "Custom chips lock us in." Reframe as co-design for performance and cost control. The upside is predictable capacity and long-term roadmap alignment.
• "Can they deliver?" Point to the CEO's supply-chain assurance, current-quarter AI revenue, and the gigawatt-scale deployments planned by named customers.

Who to prioritize

• Hyperscalers and AI-native platforms building next-gen training clusters.
• Enterprises with private model roadmaps in finance, telecom, healthcare, and retail seeking cost control and data governance.
• Service providers and integrators that package compute, networking, and managed AI ops.

30-day action plan

• Map accounts: Identify customers hit by long lead times or rising AI TCO; log their GPU/TPU mix and networking pain points.
• Build a capacity story: Use the gigawatt milestones to frame delivery confidence and scaling paths.
• Quantify outcomes: Model cost per token, time-to-train, and energy savings for one pilot workload per top account.
• Partner early: Align with cloud, OEM, and SI partners to package compute + networking + services into one proposal.

Resources

• Google Cloud TPU overview - context for TPU demand referenced in the announcements.
• Nvidia Data Center platform - baseline for competitive positioning.
• AI Learning Path for Technical Sales Representatives - sharpen your technical discovery, demos, and enterprise AI deal strategy.

Bottom line: Big buyers want performance, predictability, and supply. Broadcom is signaling it can deliver all three at scale. Use that momentum to start higher-value conversations this quarter.