Since generative Artificial Intelligence began to make headlines in 2022, Nvidia has become the central supplier of the hardware powering the boom. The company reached a valuation of US? trillion (S?.5 trillion) in late October and is on course to report more net income in 2025 than its two main rivals will chalk up in sales, combined. A wave of data centre spending has reinforced the view that demand for Artificial Intelligence infrastructure is still rising, even as some in the industry question whether the pace of investment is running ahead of profitable use cases.
Nvidia’s current flagship products are its Blackwell Artificial Intelligence accelerators, which follow the Hopper line and are adapted from graphics processing units originally built for video games. Blackwell delivers 2½ times Hopper’s performance in training Artificial Intelligence, according to Nvidia. The chips are sold in multiple formats, including the GB200 super-chip, which combines two Blackwell GPUs with one Grace central processing unit. Nvidia has also tuned its products for inference, a capability that is becoming more important as more users rely on Artificial Intelligence for a widening range of tasks.
The company’s edge comes from more than raw chip performance. Nvidia has steadily updated its hardware and the software that supports it, and it has built cluster systems that let customers deploy large volumes of chips quickly. That combination has helped it control about 90 per cent of the market for data centre GPUs, according to IDC. Advanced Micro Devices and Intel are trying to catch up, while major cloud groups including Amazon.com, Alphabet’s Google and Microsoft are developing in-house chips to reduce reliance on Nvidia. AMD has shown some progress through supply deals with OpenAI and Oracle.
Demand for Nvidia’s products remains strong. In late October, Nvidia projected revenue of about US? billion from its data centre unit over the next five quarters. That forced even the most optimistic analysts to raise their estimates and helped to add US? billion to Nvidia’s market value in one week. Microsoft, Amazon, Meta and Google have announced plans to spend hundreds of billions of dollars collectively on Artificial Intelligence and the data centres to support it, while OpenAI has continued buying computing power for future deployment.
Geopolitics has become one of the biggest threats to Nvidia’s momentum. In April, Nvidia said it was taking a US?.5 billion inventory writedown caused by a US ban on the supply of its H20 chip to companies in China. Although Washington later allowed H20 sales to resume, Beijing responded by telling Chinese companies to avoid Nvidia’s products. Jensen Huang has argued in Washington that broader sales to China would support US national security by preventing rivals such as Huawei from filling the gap, but no agreement has yet restored full access to the Chinese market.