Broadcom Inc (NASDAQ: AVGO) reported a 221% year-over-year surge in AI semiconductor revenue to $16.7 billion for fiscal Q3. CEO Hock Tan claimed the company’s custom chip for OpenAI outperforms Nvidia’s Grace Blackwell at half the cost.

Total Q3 revenue of $29.59 billion beat estimates of $29.36 billion, with adjusted EPS of $3.32 topping expectations of $3.24. Despite the beat, shares fell after hours on softer-than-expected Q4 guidance.

Competitive Edge Against Nvidia

Tan highlighted “Jalapeno,” OpenAI’s first-generation custom accelerator built by Broadcom. He stated that Jalapeno “outperforms the Grace Blackwell Ultra in performance per watt, latency, throughput, and power” for inference workloads.

Pointing to the economic advantage of specialized hardware, Tan noted that a chip optimized for specific large language models will “outperform any GPU” and can do so “at half the cost of a GPU.”

Customer and Deployment Outlook

Broadcom outlined significant growth trajectories for its custom AI chip business, stating that Anthropic is on track to become the company's largest XPU customer in 2027 and sustain that position in 2028.

Regarding OpenAI's infrastructure, the company noted that the Jalapeno project is on track for the planned deployment of 1.3 gigawatts in 2027. Management indicated there is line of sight for OpenAI to deploy over 5 gigawatts of Jalapeno and its successor generation of XPU.

Google and Anthropic Expansion

During the earnings call, Tan announced that Broadcom has started production shipments of Google’s next-generation TPU v8i, following high-volume deliveries of its Ironwood TPU v7 to both Google and Anthropic. The new TPU v8i features more memory and bandwidth than Ironwood and is optimized for AI inference workloads.

Broadcom recently expanded its long-term agreement with Google to develop and supply future generations of TPUs and AI networking products. Under this agreement, Broadcom expects to deliver "multi-tens of billions of dollars" worth of TPUs annually over the next several years. Tan stated that demand in 2028 and 2029 will be met through successive generations of increasingly sophisticated TPUs developed with Google.

Anthropic represents another major growth vector. Broadcom is deploying 1 gigawatt of Ironwood capacity for Anthropic in 2026. The company expects Anthropic to deploy another 5 gigawatts of TPU v8i in 2027, with an additional 10 gigawatts of potential deployments in 2028.

Financial Performance and Guidance

Broadcom’s AI semiconductor revenue is expected to accelerate further to $21.7 billion in the fourth quarter, representing a 236% year-over-year increase. For the full fiscal year 2026, the company projects AI revenue to reach $58 billion.

Looking ahead, Tan mapped out an aggressive growth trajectory, projecting AI semiconductor revenue to double to approximately $115 billion in fiscal 2027 and double again to $230 billion in 2028.

For the fourth quarter, Broadcom expects total revenue of approximately $34.8 billion, which is slightly below analysts’ estimate of $35.03 billion. This weaker-than-expected outlook pressured the stock in after-hours trading.

Shares closed at $367.24, down 0.66% on Wednesday, and fell another 0.82% to $364.23 in after-hours trading. Jim Cramer dubbed the print an “overtime save” as the market digested the historic surge in custom AI chip demand despite the soft guidance.

What the Numbers Show

The disclosure reveals a concentrated dependency on hyperscale AI customers for future revenue growth. With Anthropic projected to be the largest single customer by 2027 and OpenAI targeting multi-gigawatt deployments, Broadcom's XPU business expansion is heavily tied to the capital expenditure cycles of these entities. The simultaneous ramp of Google’s TPU v8i and Anthropic’s massive capacity additions indicates that Broadcom’s custom silicon strategy is successfully diversifying beyond a single client, reducing reliance on any one partner despite the high concentration risk inherent in the sector.

Broadcom Inc (NASDAQ: AVGO) leadership indicated it expects to ship approximately $350 billion worth of AI semiconductors to customers over the next two years. The company also stated its expectation to sustain operating margins even as the shifting product mix dilutes gross margins.

Guidance and Margin Outlook

During a conference call, Broadcom’s chief executive outlined the firm’s strategic outlook for artificial intelligence hardware. The management team highlighted a significant revenue pipeline, projecting total AI semiconductor shipments of $350 billion within a 24-month horizon.

Regarding profitability dynamics, the CEO addressed the impact of changing product compositions. While the mix of products is expected to dilute gross margins, the company maintains confidence in its ability to sustain overall operating margins. This suggests that operational efficiencies or other margin-positive factors are anticipated to offset the gross margin pressure from the evolving product portfolio.

• Projected AI semiconductor shipments: $350 billion
• Timeframe for shipments: Next two years
• Margin stance: Operating margins expected to be sustained despite gross margin dilution