The semiconductor firm has shown an uncanny knack for betting on the right technology at the right time, and is now doubling down on artificial intelligence.
Boyden's perspectives on the news and trends that are transforming industries
In the past five years, Nvidia’s market value has increased from $31 billion to $505 billion. Its revenues grew by 233% in the same period, and its operating profit more than doubled. Last year Nvidia’s market capitalization reached $258 billion, surpassing Intel’s – despite the latter’s much higher revenues and the fact that it both designs and fabricates chips, while Nvidia focuses only on design.
Nvidia has some distinct advantages, notably its high-performance semiconductors. The company is perhaps best known for its specialized graphics processing units (GPUs), used widely in gaming and artificial intelligence (AI). Its processors power countless data centres, including those that host the biggest players in cloud computing, such as Amazon, Google, Microsoft and Alibaba. Nvidia’s chips are also deeply embedded in the IT industry and many areas of scientific research.
CEO Jen-Hsun Huang, who co-founded Nvidia in 1993, is not one to rest on his laurels, however. Nvidia has moved to the forefront of machine learning in the past decade, with GPU-accelerated computing. And the company is doubling down on data centres and AI. In September Nvidia struck a $40 billion deal to acquire British semiconductor firm Arm. The acquisition has yet to survive the scrutiny of regulators, but should it move forward, Nvidia would focus Arm’s considerable design capabilities on central processing units (CPUs) for data centres and AI.
Around half of Nvidia’s revenues still come from gaming chips, but its AI business is growing rapidly. This consists of specialized chips as well as software that allows programmers to modify them. Many of these systems go to data centres, which account for 36% of Nvidia’s revenues. By many accounts AI is disrupting the data centre, and Nvidia has evolved with this trend. Its AI hardware-software system works with the machine learning algorithms collected in the libraries of tech giants like Google and Facebook.
Nvidia’s prospects for the next decade are good: Adoption of accelerated computing is expected to widen, as it will enable companies to gain more processing power without adding more CPUs. And as more companies use artificial intelligence, spending on servers could shift to Nvidia’s accelerated computing model, says Stacy Rasgon of Bernstein. Of this, he added, half could go on accelerated chips, a market dominated by Nvidia’s GPUs. Nvidia sees the global market for accelerated computing growing to more than $100 billion a year, according to The Economist.
As with any hot technology, competition is growing, with many striving to make chips even more specialized for AI. British firm Graphcore, for example, is touting its “intelligence processing unit”. Intel acquired an AI chip startup in 2019. Advanced Micro Devices (AMD) is finalising its purchase of Xilinx, which makes an accelerator chip. Competition is also coming from Nvidia’s biggest clients as heavyweights like Google, AWS and others get into designing chips of their own. But its rivals do not have Nvidia’s software ecosystem, or its agile focus, which has rarely failed it.
