Broadcom’s AI revenue jumps 77%: Stock rebounds 14%

Broadcom CEO Hock Tan provided a positive second-quarter forecast on Thursday, alleviating investor concerns over AI chip demand amid a competitive landscape. Following a 6% decline in shares, Broadcom’s stock surged 14% in extended trading after the announcement.

The chipmaker anticipates revenue of approximately $14.90 billion, exceeding analyst estimates of $14.76 billion, according to LSEG data. Broadcom is experiencing substantial demand for its custom artificial intelligence chips as cloud computing companies seek alternatives to expensive Nvidia processors to expand their AI infrastructure.

Tan stated that revenue for AI semiconductors is projected at $4.4 billion in the second quarter, driven by hyperscale customers investing in custom AI chips for their data centers. Analysts expect Broadcom to benefit further as large tech firms shift from off-the-shelf chips to in-house processors due to the increasing complexity of AI task requirements.

Broadcom and TSMC explore Intel acquisition options

Broadcom has engaged with four new hyperscale customers for the development of custom chips, in addition to the existing three customers already utilizing its processors. These new clients have not been factored into the projected revenue opportunity of $60 billion to $90 billion by 2027, as revealed by Tan during the post-earnings call.

OpenAI is finalizing its first custom chip design with Broadcom to decrease its reliance on Nvidia, as reported by Reuters. “The company is enabling hyperscalers and others who want to build custom AI accelerators to control their designs and costs,” said Anshel Sag, an analyst at Moor Insights & Strategy.

Broadcom is also assessing Intel’s advanced manufacturing process, known as 18A, having tested portions of chips. Summit Insights analyst Kinngai Chan noted that Broadcom’s position in the AI market is more diversified with multiple AI ASIC customers, providing a competitive edge.

In the first quarter, Broadcom reported revenue of $14.92 billion, surpassing estimates of $14.61 billion. AI revenue rose over 77% to $4.1 billion due to strong demand for its custom-made accelerators. The infrastructure software segment’s revenue increased more than 47% to $6.70 billion, exceeding expectations of $6.49 billion. Adjusted earnings per share rose 45% to $1.60, exceeding the expected $1.49. Adjusted EBITDA grew 41% year-over-year to $10.08 billion, surpassing the FactSet consensus estimate of $9.66 billion.

After-hours trading saw Broadcom shares increase by more than 12%, erasing previous losses and pushing the stock back above $200. Despite earlier struggles following competition from the Chinese startup DeepSeek’s AI model, Broadcom shares had reached a record high of $250 on December 16.

The news of engaging with two new hyperscaler clients to develop custom accelerators was a key highlight in the earnings call that pushed shares higher. Current customers include known entities such as Alphabet, Meta Platforms, and TikTok’s parent company ByteDance. The announcement of new customer engagements is expected to mitigate concerns around AI spending and indicates a preference for custom silicon over standard graphics processing units from rivals like Nvidia.

Broadcom’s estimated serviceable addressable market (SAM) for AI is projected to be between $60 billion and $90 billion by fiscal 2027. Tan has indicated that this estimate may be conservative.

For its fiscal 2025 second quarter, Broadcom expects revenue to rise 19% year-over-year to $14.9 billion, surpassing analysts’ consensus estimate of $14.7 billion. AI revenue is projected to increase to $4.4 billion. The legacy semiconductor business is anticipated to show modest revenue growth, while software revenue is expected to grow 23% year-over-year to $6.5 billion. Management forecasts adjusted EBITDA to be approximately 66% of projected revenue, equating to $9.83 billion, which exceeds the FactSet consensus of $9.5 billion.

Broadcom has increased its research and development investments to develop the next generation of accelerators and expand its Tomahawk 5 high-bandwidth network switches. “These R&D investments are very aligned with the roadmap of our three hyperscale customers as they each raise towards 1 million XPU clusters by the end of 2027,” Tan explained.

Featured image credit: Max Bender/Unsplash