Once a relatively modest tech company known for its graphics processing units (GPUs) for computer games, Nvidia has metamorphosised into one of the world’s most valuable public companies. Its chips power much of the AI computations in data centres at companies such as Google and Microsoft, and of course OpenAI, the maker of ChatGPT.
The company has nearly doubled in value in the last 12 months, and Nvidia enthusiasts expect it to become the first to hit $5 trillion in market cap. Serious competition is some time away, but other challenges include the potential for US and China to ratchet up their antagonism, even as some experts hope that America under returning President Donald Trump might be more transactional.
Meanwhile, even as reports emerge of some heating issues with its next-generation chips, on its Blackwell architecture, Nvidia is already going beyond data centre customers, into robotics and other potentially lucrative avenues. Here are five takeaways from 2024.
Consider this. Just one stock, Nvidia, can reverse the direction of the entire S&P 500 index, widely seen as the world’s most important stock market index. For example, on September 11, Nvidia surged more than 8.1 percent on positive demand commentary from CEO Jensen Huang.
The rally helped drive the S&P 500 to its biggest intraday upswing in nearly two years, Reuters reported later. The index reversed a 1.6 percent loss to end the day up 1.1 percent. And Nvidia’s jump added more than $200 billion to its market value and accounted for 44 percent of the S&P 500’s surge that day, according to Reuters, which cited data from the Japanese bank Nomura.
Nvidia’s most optimistic believers expect the company will be the first to reach $5 trillion in market cap and go beyond. The flip side is, of course, investors worry that the opposite could happen as well, on adverse news, which can wipe out value across the broader index.
In fact, earlier in September itself, between opening on September 3 and closing on September 4, Nvidia lost more than 8.4 percent on the news that it had received a subpoena from the US Justice Department. There were some technical clarifications later. Since then, Nvidia has continued to rise as more news poured in, on unabated demand.
Replaced Intel on the blue-chip Dow Jones Industrial Average index
One event put the stamp of validation this year on Nvidia’s stratospheric rise. The S&P Dow Jones Indices made significant changes to the Dow Jones Industrial Average (DJIA) that became effective before the markets opened on November 8 in the US.
The most noteworthy change, by far, was that Nvidia replaced Intel in the blue-chip DJIA, reflecting the stunning rise of the GPU maker on the back of demand for its AI accelerators, and the historic end of an era for Intel, an American tech icon that’s yet to offer a serious competing product.
Intel’s vaunted AI chip Gaudi hasn’t seen the traction that the company had hoped for this year, missing a sales target. Intel’s CEO Pat Gelsinger has since been ousted for failing to turn the company around. The S&P change removed Intel from the index after a 25-year reign.
Some heating issues remain with Blackwell, but normal, says company
Nvidia’s next-generation AI accelerators, called Blackwell, which have already faced delays, are causing servers they go into to heat up, according to Reuters on November 17, citing a report from The Information. As a result of the heating issue, some customers are worried they won’t meet data centre deadlines, according to the report. Nvidia’s Blackwell GPUs are overheating when placed in server racks designed for up to 72 chips, according to the report.
Thus far, there have been multiple attempts to redesign the racks, according to Reuters, citing The Information. Nvidia is working with cloud services providers and the engineering changes are part of the normal process, according to a statement from the company to Reuters.
Initially set to ship in Q2 (May-July for Nvidia), Blackwell chips were delayed, impacting major customers like Meta, Google and Microsoft. Unveiled in March, the Blackwell GPUs are said to be as much as 2.2x times faster compared to the previous generation, called Hopper, according to various news reports.
Also read: How Apple fared in 2024: Five takeaways from the year in review
Getting caught up in US-China geopolitical war
Nvidia’s most powerful AI tech is already inaccessible to China as the US has moved to keep those out of its powerful rival’s hands. News reports came out on December 9, citing a regulator in China that the country was investigating Nvidia for possible violations of the nation’s anti-monopoly laws.
China’s State Administration for Market Regulation has opened an investigation into Nvidia, over its acquisition of Israel’s Mellanox, which made network solutions for data centres and servers, and some agreements made during the acquisition, CNBC reported on December 9.
Industry watchers widely believe this to be part of the move-counter-move escalation of the hi-tech trade denial between the US and China, because Nvidia acquired the Israeli technology company four years ago. As the US passed regulations limiting its advanced tech’s sale to China, Nvidia has created products that meet those rules.
Going beyond data centres, with Jetson Thor chip for humanoid robots
One investor concern with respect to Nvidia’s expensive chips is that we don’t know how long demand will sustain from the point of view of training very large AI models. The biggest tech companies including Microsoft, Meta, Google and Amazon, which are doing a lot of this work, accounted for nearly 40 percent of Nvidia’s sales, Bloomberg reported in February.
The view among industry experts is that as AI spend moves more towards “inferencing” (meaning using AI models to get useful output) those tasks won’t need such powerful chips. Less expensive chips will do. It’s not clear, for now, how far out that scenario is.
Meanwhile, Nvidia is already to moving to tap other opportunities, outside data centres. For example, it’s planning a mid-2025 release for its Jetson Thor computer that is specifically aimed at powering robots and humanoid robots, the Wall Street Journal and others reported on November 14.
Unlike smartphones, where a handful of companies dominate, robotics is a fragmented market that offers a large opportunity, WSJ reported, citing Deepu Talla, Nvidia’s vice president of robotics and edge computing, who spoke to reporters at the company’s AI conference in Tokyo recently.
Nvidia’s approach, unlike that of Tesla’s, for example, is to offer a platform for robots and not build one itself. In March, Nvidia announced Project GR00T, a general-purpose foundation model for humanoid robots, and the new computer, Jetson Thor, for humanoid robots based on the Nvidia Thor system-on-a-chip (SoC). It also announced upgrades to its Isaac robotics platform, including generative AI foundation models and tools for simulation and AI workflow infrastructure.