By some accounts, Nvidia currently controls between 80% and 98% of the AI hardware market, and there's no reason to expect that dominance to be challenged any time soon. However, Nvidia's stock price took a tumble recently, leading some analysts to begin questioning whether Nvidia is as bulletproof as it once seemed. The main issues facing the company are its ability to make enough GPUs to satisfy global demand and increasing power consumption, two factors leaving big openings for its competitors.
Korean media is now reporting that some analysts are concerned about the long-term outlook for Nvidia in the AI space, a notion unheard of just a few months ago. The primary catalyst for these doubts is a recent 10% drop in its stock price, which was the biggest in the company's history and occurred on April 19. Compounding the negativity is that Nvidia relies solely on TSMC to manufacture its GPUs, and TSMC simply can't keep up with the insatiable demand for these chips. Although the company is opening new fabs in Japan, Taiwan, and the United States, it will be years before they're operational, creating a "bottleneck" for Nvidia and the AI industry.
In the report from Korea, an official with the largest Internet company in Korea states that going with Nvidia for a build-out will make it difficult to realize a profit due to the costs involved, both for the GPUs and the related power consumption. For these reasons, this official says his firm is looking at options from Intel and Samsung to help reduce costs. As Wccftech points out, one analyst states that if power consumption continues to increase, 2027 global data center power consumption will be equivalent to that of the Netherlands, Argentina, and Sweden combined.
Power consumption alone could leave the door open to a more efficient competitor coming along and eating Nvidia's lunch, as this problem is not going away. At the same time, if demand for AI hardware is not expected to let up anytime soon and power consumption keeps increasing, it's feasible the industry won't be able to meet the demands of the AI market without low-power processors being involved at some point.
Of course, Nvidia may also be working on a low-power version of a future architecture, as it will do everything possible to maintain its position in the market. Nvidia is also aware that companies are getting fed up with long wait times for its A100 and H100 GPUs and have begun looking elsewhere. That's one of the reasons Nvidia recently formed a new unit to go after the custom silicon market, with the hopes of neutralizing it before it grows too big to control.
