Broadcom Reports AI Revenues Hit $10.8 Billion as Demand Accelerates
Broadcom's second-quarter earnings call made a straightforward case: AI infrastructure demand is growing faster than the company previously expected. CEO Hock Tan reported AI semiconductor revenues of $10.8 billion, up 143% year over year and ahead of the company's own forecast.
The numbers suggest this is not a one-quarter spike. Broadcom took bookings of more than $30 billion in AI semiconductors during the quarter against $10.8 billion shipped. That gap signals customers are committing to longer-term purchases rather than placing spot orders.
The company beat analyst expectations on both earnings and revenue. Non-GAAP earnings reached $2.44 per share on revenues of $22.19 billion, topping consensus estimates by 1.82% and 0.68% respectively.
Third Quarter Guidance Points to Steeper Growth
Broadcom projected third-quarter revenues of about $29.4 billion, up 84% year over year. AI semiconductor revenues alone are expected to reach $16 billion, a significant jump from the second quarter's $10.8 billion.
The company also forecast infrastructure software revenues of $8.9 billion, up 31% year over year. CFO Kirsten Spears said strength in both operating segments supports the higher guidance.
Long-Term Contracts Extend Visibility Through 2028
Tan went beyond quarterly guidance to address multiyear demand. Broadcom expects AI semiconductor revenues to exceed $100 billion in fiscal 2027 and said growth should continue into fiscal 2028 based on work with six core customers.
The company named Google, Anthropic, OpenAI, and Meta as key partners. Management tied these relationships to long-term agreements and multi-gigawatt deployment plans that extend through 2028 and 2029.
Tan said visibility now runs through 2028, compared with roughly 2027 three months earlier. Customers are placing orders earlier to secure compute capacity and power infrastructure.
Margin Pressure Reflects Product Mix, Not Deterioration
Gross margin fell to 77.1% in the second quarter, down 230 basis points year over year. The decline reflects semiconductors becoming a larger share of revenues, not weakness in the underlying business.
Spears said the expected third-quarter gross margin of about 74% also reflects revenue mix. AI semiconductors carry lower margins than software, while networking remains richer. She emphasized this is a temporary mix effect, not a structural reset.
Operating margin actually rose 200 basis points to 67.3% as the company held operating expenses relatively flat despite higher revenues.
Networking Now Accounts for 40% of AI Revenues
Broadcom's AI networking business represented almost 40% of AI revenues in the quarter. Tan described around 30% as a more typical share over time, suggesting the product mix may shift as custom accelerator programs scale.
The software business added a second leg to growth. Software revenues rose 9% in the second quarter, with annual recurring revenue growth holding at 17%. Tan linked stronger third-quarter software expectations to VMware Cloud Foundation 9.1 and on-premises cloud demand for enterprise AI inferencing.
Supply Appears Adequate Through 2027
When asked about wafer and high-bandwidth memory supply, Tan said Broadcom is comfortable with capacity for 2026 and 2027. The company is already planning for 2028 and 2029, suggesting capacity planning is keeping pace with demand.
What This Means for Management
Broadcom's earnings call signals that AI infrastructure buildout is moving from speculation to contracted reality. Managers overseeing data center strategy or technology infrastructure should note that major cloud providers and AI companies are locking in multiyear commitments for compute and networking capacity.
The visibility into 2028 and 2029 also suggests the AI infrastructure cycle will extend longer than many expected. This affects capital planning, vendor relationships, and competitive positioning.
For leaders evaluating AI infrastructure investments, Broadcom's customer mix and deployment timelines offer a window into how the largest AI companies are building out their infrastructure. The emphasis on custom accelerators and networking suggests these capabilities are becoming as important as raw compute power.
Learn more about AI for Executives & Strategy to understand how infrastructure decisions shape AI competitive advantage.
Your membership also unlocks:
