AI "sells shovel people"! Sales of global chip companies will exceed US$400 billion in 2025, and Goldman Sachs predicts that Nvidia’s revenue will reach US$383 billion next year

# Source: Wall Street Insights

## By Bao Yilong

Analysts surveyed by FactSet expect the combined sales of five companies—NVIDIA, Intel, Broadcom, AMD, and Qualcomm—to exceed $538 billion next year. This figure does not include sales from Google's TPU business or Amazon's custom chip segment. However, the industry's growth faces significant constraints, including shortages of key data center components, power supply challenges, and concerns over financing sustainability.

Driven by the explosive growth of artificial intelligence, the combined sales of global chip companies surpassed $400 billion in 2025, setting a historical record for the chip industry. This number is expected to climb further in 2026.

According to Goldman Sachs' estimates, NVIDIA alone is projected to achieve $383 billion in GPU and other hardware sales in 2026, representing a 78% year-on-year increase.

Analysts surveyed by FactSet expect the combined sales of five companies—NVIDIA, Intel, Broadcom, AMD, and Qualcomm—to exceed $538 billion next year. This figure does not include sales from Google's TPU business or Amazon's custom chip segment.

Last week, NVIDIA signed a $20 billion licensing agreement with Groq, a chip startup focused on AI inference acceleration. This marks a shift in the AI race battlefield from the training phase to the inference phase, as tech giants now compete to deliver the fastest and most cost-effective inference capabilities.

These hardware designers are playing the role of "shovel sellers" in this digital gold rush. Nevertheless, the industry's growth faces significant constraints, including shortages of data center components such as transformers and gas turbines, power supply difficulties, and doubts about whether AI companies can sustainably secure sufficient financing to maintain their chip procurement pace.

### Intensifying Competitive Landscape

NVIDIA achieved more than doubled year-on-year revenue growth in 2025, but the competitive landscape is undergoing changes.

Data center operators, AI labs, and enterprise customers have robust demand for NVIDIA's advanced H200 and B200 graphics processors. However, Google's increasingly sophisticated custom TPUs and Amazon's Trainium and Inferentia chips are also vying for customers.

NVIDIA's $20 billion licensing deal with Groq last week reflects the industry's shifting focus from AI training to inference. Inference refers to the process by which a trained AI model responds to prompts and generates answers.

Software developers such as OpenAI are collaborating with custom chip designers like Broadcom to develop their own chips. AMD, a chipmaker with a 50-year history, plans to launch its first GPU in 2026 that will truly challenge NVIDIA's AI processors.

Microsoft stated in October that it will double the floor space of its data centers over the next two years, meaning chip manufacturers may see higher revenue in 2026.

### Escalating Supply Chain Bottlenecks

2026 may bring unprecedented challenges. Shortages of components such as transformers and gas turbines are hindering data center construction, and operators are struggling to secure the massive power required to run computing clusters.

Another major challenge is the global shortage of components needed for AI data center servers.

Products facing tight supply include ultra-thin silicon substrate layers required for certain chips, as well as memory chips that transmit data to AI processors and help store computing results.

With the acceleration of data center construction and rising demand for inference, the need for high-bandwidth memory (HBM) chips has surged. Sumit Sadana, Chief Commercial Officer of Micron Technology, said:

> We are far from meeting customer demand, and this situation will persist for some time.

Micron is one of the largest manufacturers of HBM chips used in AI applications, and its stock price has risen 229% year-to-date.

Micron and its competitors such as Samsung and SK Hynix have emerged as the main beneficiaries of the supply crunch, enabling them to raise product prices and increase capital expenditures to expand manufacturing operations. However, building large-scale clean rooms and manufacturing facilities to meet the demand of major chip companies takes time.

### Doubts Over Financing Sustainability

There are serious concerns about the financing sustainability behind data center construction. It remains unclear whether major clients such as OpenAI can quickly raise sufficient funds to maintain their rapid pace of chip procurement.

Investors have grown accustomed to extraordinary quarter-on-quarter revenue growth, and any signs of slowdown are prone to triggering panic. This autumn, investors sold off a large number of AI stocks, including those of major chip designers, amid fears that the financing driving the procurement of AI infrastructure products may not be as solid as previously thought.

A significant portion of large-scale data center construction is driven by OpenAI, which has signed multi-billion-dollar computing power agreements with companies such as Amazon, Microsoft, and Oracle. Hyperscalers like Microsoft have committed to expanding data center construction in 2026, but some analysts believe this boom may slow down in 2027.

Gil Luria, Analyst at DA Davidson, said:

> 2026 has the potential to be the peak. If we haven't heard that OpenAI has raised $100 billion by the end of March, the market may start to hit the brakes.

As more chip companies launch AI products, concerns over margin pressures are also mounting.

Broadcom's stock price still fell after it reported record quarterly revenue in December, partly due to investor concerns that sales growth in its high-margin product lines will slow. However, some industry insiders hold a more optimistic view, believing that demand will remain strong.

Brad Gastwirth, Global Head of Research at Circular Technologies, a computing hardware distributor based in Massachusetts, said:

> I don't think this is the peak. The race for artificial general intelligence (AGI) is still driving enormous demand for computing power across all types of customers.

### Risk Warning and Disclaimer

The market is risky, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situations, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are consistent with their specific circumstances. Any investment made based on this article shall be at the user's own risk.