Broadcom (AVGO) Reports Q4 2025 Earnings

Broadcom reported fourth‑quarter results that beat expectations for both earnings and revenue, and provided an aggressive outlook for the current quarter driven by surging artificial‑intelligence demand.

The shares rose in after‑hours trading before slipping slightly more than 2 percent in the regular session.

Key financial metrics versus LSEG consensus:

• Earnings per share: $1.95 adjusted vs. $1.86 estimated
• Revenue: $18.02 billion vs. $17.49 billion estimated

Broadcom forecast fiscal first‑quarter revenue of roughly $19.1 billion, a 28 percent year‑over‑year increase that outpaces the $18.3 billion average analyst target.

Chief Executive Hock Tan indicated that AI‑related chip sales are expected to double year‑over‑year to $8.2 billion this quarter, reflecting growth in both custom AI accelerators and AI‑networking semiconductors.

Net income surged 97 percent to $8.51 billion, or $1.74 per share, compared with $4.32 billion, or $0.90 per share, in the same period last year.

Alongside Nvidia, Broadcom has emerged as a major beneficiary of the AI boom. Its stock is up about 75 percent in 2025 after having doubled in 2024, propelled by custom chips that rival Nvidia’s GPUs, including variants based on Google’s tensor‑processing units (TPUs).

During the earnings call, the company disclosed that it added a fifth customer for its custom AI chips. Anthropic, previously identified only as a “mystery customer,” placed a $10 billion order for Google‑based TPUs in the quarter.

In June, Broadcom reported three confirmed customers and four prospects for its custom AI solutions. By September, a fourth mystery customer had placed another $10 billion order, and in October the firm announced a strategic collaboration with OpenAI to develop bespoke accelerators.

Anthropic is now running Google’s latest TPU, codenamed Ironwood, according to Tan.

Tan emphasized that many of Broadcom’s clients prefer to “control their own destiny” by pursuing multi‑year programs to develop proprietary AI accelerators—what the company refers to as XPUs.

The newly added fifth customer has committed $1 billion in orders slated for delivery in late‑2026.

Overall, Broadcom carries a $73 billion backlog of custom chips, switch fabrics, and other data‑center components tied to AI workloads over the next 18 months.

Revenue in the quarter grew 28 percent, anchored largely by a 74 percent jump in AI‑chip sales—equating to $8.2 billion of AI‑related revenue.

AI chip sales fall under the semiconductor solutions division, which posted $11.07 billion in revenue, up 22 percent year‑over‑year and exceeding StreetAccount’s $10.77 billion estimate.

The infrastructure‑software segment, which includes the company’s VMware portfolio, reported $6.94 billion in sales, a 26 percent increase that also beat market expectations.

Broadcom announced a quarterly dividend of $0.65 per share, up from $0.59, payable later this month.

Strategic Implications

Broadcom’s aggressive push into custom AI accelerators reflects a broader industry shift away from commodity GPU solutions toward vertically integrated, application‑specific silicon. By offering both the silicon and the networking fabric needed for large‑scale AI clusters, Broadcom positions itself as a one‑stop shop for hyperscale data‑center operators.

The company’s sizable backlog provides visibility into future cash flows and suggests that enterprise customers are willing to lock in volume purchases despite ongoing supply‑chain constraints. However, the reliance on a handful of large, often undisclosed, customers introduces concentration risk—if any of these relationships falter, revenue could be impacted.

From a technology standpoint, Broadcom’s XPU strategy aims to blend the high‑throughput characteristics of GPUs with the low‑latency, high‑efficiency traits of ASICs. This hybrid approach could yield better performance‑per‑watt metrics, a critical factor for data‑center operators facing rising energy costs.

Competitive dynamics remain intense. Nvidia continues to dominate the GPU market and is expanding its software stack (e.g., CUDA, DGX). Meanwhile, rival chipmakers such as AMD, Intel, and emerging China‑based firms are accelerating their AI‑chip roadmaps. Broadcom’s success will hinge on its ability to deliver differentiated performance, maintain pricing discipline, and scale its manufacturing footprint.

Financially, the company’s diversified portfolio—mixing high‑margin software licensing (VMware) with hardware sales—provides a buffer against cyclical downturns in any single segment. The increase in dividend payout signals confidence in cash generation, but investors should monitor capital‑expenditure plans as the XPU program scales.

Overall, Broadcom’s earnings beat, robust AI‑chip backlog, and strategic partnerships underscore its evolving role from a traditional semiconductor supplier to a full‑stack AI infrastructure provider.