Nvidia, the world’s most valuable company by market capitalization, continues to rely heavily on a small group of customers for a significant portion of its revenue, Fortune reports.
The AI chip supplier recently disclosed in its quarterly 10-Q filing that a few key accounts have contributed tens of billions of dollars in sales, with three anonymous customers each purchasing between $10 billion and $11 billion worth of products over the first nine months of 2024.
These customers, referred to as “Customer A,” “Customer B,” and “Customer C,” collectively made up more than a third of Nvidia’s total sales in the fiscal third quarter, which amounted to $35.1 billion. Each customer accounted for around 12% of Nvidia’s turnover, indicating that the company has likely allocated a large share of its limited chip supply to these accounts.
Nvidia’s founder and CEO, Jensen Huang, has acknowledged the company’s ongoing supply constraints. The company relies on Taiwan’s TSMC for the fabrication of its advanced AI microchips, rather than producing them in-house. This limited production capacity has contributed to the high demand for Nvidia’s products, particularly in the data center market, which is expected to grow significantly in the coming years. Bloomberg Intelligence’s Mandeep Singh believes that the data center training market could reach $1 trillion without any significant decline in spending, a projection that bodes well for Nvidia’s future revenue.
While Nvidia has not publicly identified these customers, speculation suggests that they could include major companies such as Microsoft, Meta, and possibly Super Micro, although the company has declined to comment on these assumptions. The anonymity of these key accounts is likely to remain a trade secret, as both Nvidia and its customers seek to protect competitive advantages.
One important factor affecting Nvidia’s future is the shift from AI training to inference, the latter being the process of running generative AI models once they have been trained. Unlike training chips, which require cutting-edge technology, inference chips do not need to be as advanced, opening up the market to more competition. Companies such as AMD and even Tesla, which produces its own custom silicon, are positioning themselves to capture a larger share of the inference market.
Looking ahead, Nvidia faces potential risks from this shift. While its dominance in AI training chips remains strong, the increased competition in the inference market could impact Nvidia’s market share. Furthermore, some of the cloud computing hyperscalers that have been major customers for Nvidia’s training chips, such as Alphabet, could eventually reduce their orders as they develop their own AI solutions.
