Shares of Broadcom climbed about 5% on Thursday after CEO Hock Tan outlined a bullish outlook for the company’s role in the artificial intelligence boom. Speaking to analysts, Tan said Broadcom expects its artificial intelligence chip revenue in 2027 to be “significantly in excess of $100 billion,” driven by surging demand for custom-designed silicon chips used in large-scale AI data centers.

The forecast exceeded many expectations on Wall Street and reinforced investor confidence that Broadcom’s AI-driven growth could remain strong for years.

Custom AI Chips Drive Growth

Broadcom has become a key player in designing specialized chips for hyperscale data centers, allowing major technology companies to optimize AI performance while controlling infrastructure costs.

According to Tan, Broadcom is nearing 10 gigawatts of AI computing capacity across six major customers, indicating a large pipeline of future demand.

Analysts estimate the company could generate between $12 billion and $15 billion in revenue per gigawatt of AI capacity by 2027, potentially pushing AI-related revenue beyond $120 billion.

Broadcom’s networking technology and custom silicon design help hyperscale cloud providers reduce inference costs when running large language models.

Supply Chain Secured Through 2028

The semiconductor industry has recently faced shortages in high-bandwidth memory needed for AI computing. However, Tan said Broadcom has already secured memory capacity and advanced wafer supply through 2028, reducing the risk of production bottlenecks.

The company has also improved manufacturing yields and costs, helping maintain profit margins even as AI infrastructure expands.

Tan said the economics of Broadcom’s AI business will remain consistent with the company’s broader semiconductor operations.

Competition With Nvidia Still Intense

The AI chip market is dominated by Nvidia, whose graphics processing units power many leading AI models. Tan noted that competition with Nvidia is unlikely to weaken demand for Broadcom’s chips. Instead, he said the intense competition among large language model developers increases the need for high-performance hardware.

Companies building AI models “cannot afford to have a chip that is just good enough,” Tan said, emphasizing the importance of cutting-edge computing performance.

Ripple Effects Across AI Supply Chain

Broadcom’s strong outlook also lifted shares of companies involved in AI infrastructure. Credo Technology Group rose about 10%, while Amphenol gained roughly 4% as investors bet on continued demand for copper connectivity used in AI server networking.

Meanwhile, optical networking companies such as Lumentum and Coherent saw their shares decline, reflecting investor expectations that copper connectivity solutions may dominate certain AI data center deployments.