Broadcom Inc. Q2 FY2025 earnings call

• Total revenue in fiscal Q2 2025 was a record $15 billion, up 20% year-on-year, all organic. Q2 consolidated adjusted EBITDA was $10 billion, up 35% year-on-year.
• Semiconductor revenue in Q2 was $8.4 billion, with AI semiconductor revenue over $4.4 billion, up 46% year-on-year. Custom AI accelerators grew double digits, and AI networking grew over 170% year-on-year, with AI networking representing 40% of AI revenue. The Tomahawk 6 switch was announced, enabling better performance in AI clusters.
• Non-AI semiconductors revenue in Q2 was $4 billion, down 5% year-on-year, but had bright spots like broadband, enterprise networking, and server storage up sequentially.
• Infrastructure software revenue in Q2 was $6.6 billion, up 25% year-on-year, with over 87% of 10,000 largest customers adopting VCF. Q3 infrastructure software revenue expected to be ~$6.7 billion, up 16% year-on-year.
• Free cash flow in Q2 was $6.4 billion, representing 43% of revenue. Ended Q2 with $9.5 billion in cash and $69.4 billion in gross principal debt, and repaid $1.6 billion post-quarter end.

Semiconductors: Q2 revenue was $8.4 billion, with growth accelerating to 17% year-on-year. AI semiconductor revenue was over $4.4 billion, up 46% year-on-year, representing 56% of total revenue. Non-AI semiconductors revenue was $4 billion, down 5% year-on-year. Infrastructure Software: Q2 revenue was $6.6 billion, up 25% year-on-year, representing 44% of total revenue.

• Q3 consolidated revenue forecasted to be approximately $15.8 billion, up 21% year-on-year.
• AI semiconductor revenue forecasted to be $5.1 billion, up 60% year-on-year (10th consecutive quarter of growth).
• Infrastructure software revenue expected to be approximately $6.7 billion, up 16% year-on-year.
• Q3 consolidated gross margin expected to be down approximately 130 basis points sequentially, primarily due to higher mix of XPUs within AI revenue.
• Q3 adjusted EBITDA expected to be at least 66%.

Q: Ross Seymore asked about inference commentary and confidence in growth rate matching next fiscal year.

A: Hock E. Tan said it's a combination of XPUs and networking, with more inference seen now.

Q: Harlan Sur asked about smoothing growth rates and AI business trajectory.

A: Hock E. Tan said the rate seen in fiscal '25 is expected to sustain into '26 based on visibility.

Q: Ben Reitzes asked about AI networking strength and Tomahawk's role.

A: Hock E. Tan said AI networking is strong, with Tomahawk 6 having strong interest though not in big orders yet.

Q: Blayne Curtis asked about scale-up networking.

A: Hock E. Tan said scale-up is rapidly converting to Ethernet.

Q: Stacy Rasgon asked about AI revenue math and SAM.

A: Hock E. Tan said the trend is to see AI revenue ramp up in '26 as seen, but no update on SAM.

Q: Vivek Arya asked about XPU business near term and longer-term milestones.

A: Hock E. Tan said XPU trajectory is as expected, and no updates on '27 numbers yet.

Q: C.J. Muse asked about inference workloads for custom silicon.

A: Hock E. Tan said custom accelerators add value in optimizing algorithms for LLMs.

Q: Karl Ackerman asked about co-packaged optics in scale-up networks.

A: Hock E. Tan said scale-up will move to optical interconnects, with co-packaged optics a path but not the only one.

Q: Joshua Buchalter asked about gross margins and custom business.

A: Kirsten Spears said XPU margins are slightly lower than other parts of the business.

Q: Timothy Arcuri asked about scale-up ecosystems and AI networking growth.

A: Hock E. Tan said Ethernet will prevail in scale-up as it has in traditional networking.

Q: Christopher Rolland asked about AI demand drivers and market share shift.

A: Hock E. Tan said inference is driven by monetizing AI platforms after training.

Q: Vijay Rakesh asked about fiscal '26 growth and NVLink.

A: Hock E. Tan said growth in '26 is expected to sustain, and Broadcom doesn't participate in NVLink.

Q: Aaron Rakers asked about capital return, debt, and M&A strategy.

A: Hock E. Tan said capital is used for dividends, reducing debt, and opportunistic buybacks, with M&A considered for significant deals.

Q: Srini Pajjuri asked about 2026 expectations and networking mix.

A: Hock E. Tan said no comment on prospects, and networking mix in AI expected to be closer in range in '26.

Q: Joe Moore asked about export controls on AI.

A: Hock E. Tan said it's hard to give comfort as export controls are dynamic.

Q: William Stein asked about VMware subscription model conversion.

A: Hock E. Tan said about 2/3 of VMware contracts are in renewal process, with another year plus expected for conversion.