“Everyone wants to build and train these huge models, and the most efficient way to do that is with CUDA software and Nvidia hardware,” Bob O’Donnell, president and principal analyst of TECHnalysis Research, told Yahoo Finance.

Nvidia is expected to continue to power the majority of the AI ​​industry in 2025. The company’s Blackwell chip, the successor to its popular Hopper line of processors needed for AI applications, is in production – and its customers like Amazon are already adding new cooling options to their data centers to handle the intense heat the processors generate generate.

“I don’t know what the current backlog is (for Nvidia’s chips), but if it’s not a year, it’s almost a year,” O’Donnell said. “So most of everything they’re probably going to make next year is pretty much sold out already.”

With hyperscalers demanding higher or at least equal capex in 2025 than in 2024, you can expect some of that to ultimately go toward buying Blackwell chips.

While Nvidia will retain control of the AI ​​crown, there is no shortage of challengers looking to take the throne. AMD and Intel (INTC) are the leading contenders among chipmakers and both have products on the market. AMD’s MI300X chip series is based on Nvidia’s H100 Hopper chips, while Intel has the Gaudi 3 processor.

However, AMD is better positioned to steal market share from Nvidia as Intel continues to struggle with its turnaround efforts and search for a new CEO. But even AMD has a hard time breaking Nvidia’s lead.

“What AMD needs to do is make software really usable, build the systems where there is more demand… among developers, and that could ultimately lead to more revenue,” Newman said. “Because these cloud providers will sell what their customers demand.”

However, it’s not just AMD and Intel. Nvidia’s customers are increasingly developing and pushing their own AI chips. Google has its Broadcom-based Tensor Processing Unit chips (TPUs), while Amazon (AMZN) has its Trainium 2 processor and Microsoft (MSFT) has its Maia 100 accelerator.

There is also concern that the move to “inference AI models” will reduce the need for powerful Nvidia chips.

Technology companies develop AI models by training them on massive amounts of data, also known as the training process. Training requires incredibly powerful chips and a lot of energy. The inference or actual implementation of these AI models is less resource and energy intensive. As inferencing makes up a larger portion of AI workloads, the thinking goes, companies will move away from having to buy as many Nvidia chips.

Huang said he was prepared for this, stating at various events that Nvidia’s chips were just as good at reasoning as they were at training.

Even if Nvidia’s market share declines, that doesn’t necessarily mean its business will perform worse than before.

“This is definitely about meeting all expectations,” Newman said. “Even with much stronger competition, which I think they certainly will have, that doesn’t mean they will fail. That means people are building a bigger pie.”

Email Daniel Howley at [email protected]. Follow him on Twitter at @DanielHowley.

Click here for the latest technology news that will impact the stock market

Read the latest financial and business news from Yahoo Finance