Next Nvidia? It's tough!

How difficult is it to challenge NVIDIA?

As the bet around artificial intelligence heats up, some chip companies have been "selected" by Wall Street, believing that they may benefit from this wave of enthusiasm. However, analysts generally believe that the gains these companies receive will not come from the share NVIDIA currently holds.

Patrick Moorhead, an analyst at Moor Insights & Strategy, wrote that in this round of chip wars around artificial intelligence, "NVIDIA is free and clear":

No one wants an industry with only one leader.

But for now, the booming market for generative AI chips belongs only to NVIDIA.

Earlier this week, AMD released the AI GPU MI300X chip, which is considered a catch-up to NVIDIA.

However, although AMD claims that its chip will surpass NVIDIA's flagship product H100 in multiple indicators, it has not revealed whether the chip has potential customers and said it will only start increasing the production of new chips in the last quarter of this year.

Stacy Rasgon, an analyst at Bernstein, said that when AMD's new chips are generally on the market by the first half of next year, NVIDIA's H100 has been on the market for 18 months, thus leading significantly. AMD is "far behind."

They (AMD) may get the remaining share of the (artificial intelligence market) - although perhaps this is enough.

Richard Windsor, founder of Radio Free Mobile and former Nomura Securities analyst, pointed out that so far, NVIDIA has captured about 85% of the artificial intelligence chip market, and AMD may be one of the "most important challengers" so far, but the problem is "how many developers really want to use it":

I think AMD does have the potential to gain some market share. But ultimately, NVIDIA has a huge lead in this area because it started earlier than anyone else.

Will AI boost demand for other processors?

After NVIDIA became the first chipmaker to break through a trillion-dollar valuation last month, the market values of AMD, Broadcom, and Marvell also soared, growing by more than 20%.

Although these companies have also been optimistic, it is clear that it is difficult to compete with NVIDIA on the most advanced artificial intelligence chips, so they are beginning to hope that the artificial intelligence market will boost demand for other types of processors. Intel Vice President Kavavitha Prasad said that many customers who want to train models with their enterprise data also want to keep their information nearby instead of risking it with companies that provide large AI models. She said that with all the computing work involved in preparing data input for training, this will create a lot of work for Intel's CPUs and AI accelerators.

However, the Financial Times pointed out that the rapid changes in demand for data centers due to the surge in services such as ChatGPT will make it difficult for chip manufacturers to predict how their market will develop.

Rasgon believes that as data center customers invest their spending in AI accelerators, CPU sales may even decline in the next few years.