Nvidia’s Rapid Rise: A Deep Dive into Its Growing Market Dependence

Nvidia (NASDAQ: NVDA) started 2023 with a market cap of $360 billion. Nearly two years later, it has skyrocketed over $3.4 trillion. While Nvidia is known for its graphics processing units (GPUs) in personal computers and automotive sectors, the data center segment has primarily fueled this remarkable growth.

In the industry, Nvidia’s data center GPUs stand out as the most powerful for developing and deploying artificial intelligence (AI) models. However, the company is facing challenges in keeping up with demand from AI startups and major tech corporations, which could pose risks down the line.

Nvidia’s financial results for its fiscal 2025 second quarter (ending July 28) indicate an increasing dependency on a select few customers for revenue generation. Understanding this situation is crucial as it may lead to vulnerabilities in the future.

Who Can Afford GPUs?

A study by McKinsey and Company reveals that 72% of organizations globally use AI in at least one business function. This trend is growing, but many companies lack the resources or expertise to create their own AI infrastructure. For context, Nvidia’s leading GPUs can reach up to $40,000, often requiring thousands to effectively train an AI model.

Major tech companies such as Microsoft (NASDAQ: MSFT), Amazon (NASDAQ: AMZN), and Alphabet (NASDAQ: GOOG) (NASDAQ: GOOGL) purchase hundreds of thousands of GPUs, consolidating them in centralized data centers. This allows smaller businesses to rent computing power for a fraction of the cost of building systems themselves.

In efforts to widen accessibility, cloud companies like DigitalOcean are providing AI support to even small businesses by offering clusters of one to eight Nvidia H100 GPUs for basic AI tasks.

Luckily, affordability is improving. Nvidia’s new Blackwell-based GB200 GPU systems can perform AI inference 30 times faster than older H100 models. Selling for the same range of $30,000 to $40,000, the GB200 represents a leap in cost-effectiveness.

This advancement means that cutting-edge trillion-parameter large language models (LLMs), mainly developed by well-capitalized tech giants and elite AI startups like OpenAI and Anthropic, may soon be accessible to a wider range of developers. However, it remains uncertain when GPU prices will drop sufficiently for average businesses to establish their own AI setups.

Emerging Challenges for Nvidia

Nvidia’s current sales heavily rely on a handful of tech giants and top AI startups, leading to a concentrated revenue model.

In its fiscal 2025 second quarter, Nvidia recorded $30 billion in total revenue, a staggering 122% increase from the previous year. Of this, the data center segment brought in $26.3 billion, reflecting a remarkable 154% growth.

Notably, Nvidia’s 10-Q filing discloses that four unnamed customers contributed nearly half of its $30 billion revenue:

Data source: Nvidia.

Only customers contributing 10% or more of revenue are mentioned, so other significant GPU buyers may exist who fell below this reporting criterion.

Notably, Customers A and B together represented 25% of Nvidia’s revenue in Q2, a rise from 24% just three months prior in Q1 of fiscal 2025. Thus, revenue concentration appears to be increasing.

This trend brings specific concerns. Customer A alone spent $7.8 billion with Nvidia over the past two quarters, and only a limited number of companies can maintain such extensive investments in chip technology. If one or two of Nvidia’s key clients reduce their spending, the company faces a risk of losing substantial revenue that may not be easily compensated.

Image source: Nvidia.

Identifying Nvidia’s Major Customers

While Microsoft regularly purchases Nvidia GPUs, an analyst indicates that the tech firm is the top buyer of Blackwell hardware expected to ship at the end of this year, suggesting Microsoft could be Customer A.

Other potential major customers include a mix of Amazon, Alphabet, Meta Platforms, Oracle, Tesla, and OpenAI. According to their public filings, their spending on AI infrastructure is noted as follows:

• Microsoft allocated $55.7 billion to capital expenditures (capex) in fiscal 2024 (ending June 30), focusing predominantly on GPUs and data centers. Further increased spending is planned for fiscal 2025.
• Amazon’s capex may exceed $60 billion in calendar 2024 to support its AI growth.
• Meta Platforms anticipates spending up to $40 billion on AI infrastructure in 2024, with additional investments in 2025 for enhancing its Llama AI models.
• Alphabet is projected to allocate about $50 billion to capex this year.
• Oracle dedicated $6.9 billion to AI capex in its fiscal 2024, with plans to double it in fiscal 2025.
• Tesla aims to surpass $11 billion in AI infrastructure investments this year to utilize 50,000 Nvidia GPUs in its self-driving software.

With this robust pipeline, Nvidia’s revenue outlook seems strong for the upcoming year. However, uncertainty looms regarding the sustainability of such high levels of spending from these companies in the long term.

Nvidia CEO Jensen Huang projects that data center operators will invest $1 trillion into AI infrastructure over the next five years. If accurate, this could support Nvidia’s continued growth into the late 2020s. However, new competitors are entering the market, which could challenge Nvidia’s position.

Advanced Micro Devices has released its own AI data center GPUs, poised to compete with Nvidia’s offerings.