Nvidia ,NVDA 2.62%, Artificial Intelligence (AI) has been the biggest beneficiary of tremendous demand for chips. This is not surprising, as its graphics processing units (GPU) parallel computing capacity makes them ideal for AI model training and estimates.

It is worth noting that NVidia is running with the GPU market with an estimated market share of 92% at the end of last year. It is ahead of second place Advanced micro equipment, ,AMD 1.68%, Part of just 4%. However, it would not be surprising to see AMD away from the market share from Nvidia for its upcoming chips.

Let’s see why this can happen.

AMD is aggressively closing the difference

AMD recently unveiled its AI Chip Roadmap at an investor program. Chipmaker used the event to start its MI350 series of Data Center GPU, “4X generation-on-generation AI compute promised up to 35X leap in the completion and hinning performance.” This new line of chips is built on the 3-nenometer (NM) process node. It gives an advantage on the latest generation of Nvidia’s Blackwell processor, allegedly manufactured using the 4Nm node.

Chips manufactured using a small procedure node are theoretically more powerful and power-skilled, as they pack more number of transistors in a small area. As a result, electrons need to travel in a small area to perform computing tasks, and they generate less heat. Therefore, the latest generation of AI AMD’s AI Accelerator may be better than the offerings of Nvidia, at least on paper.

AMD explains that its MI350 series processor has strong specifications compared to Nvidia’s Blackwell Prasad, which packing in 1.6 times more memory. It also suggests that the Mi355x processor is 1.2 times faster than the blackwell chips of NVIDIA. Deepsek R1 And Lama 3.1.

Even more importantly, AMD has targeted Nvidia Big Time to hurt its Mi 400 Series of Accelerator, which are scheduled to be launched next year. The company is going to increase high-bandwidth memory (HBM) capacity from 288GB to Mi400 on Mi350 series chips. What is more, the AMD MI400 processor is going to more than doubled the memory bandwidth, 19.6 terabytes per second.

AMD states that the MI400 AI will mark a significant jump in compute performance, which will be much higher than what has been seen in the upgradation of the previous generation. Meanwhile, the current generation of the next generation of Nvidia’s AI GPU is expected to pack the same memory compared to the Blackwell processor, which means that AMD can enjoy a specific advantage on paper.

A large memory stack will allow the chips of AMD to move more data, while high bandwidth will enable at least theoretically rapid transmission. This potential advantage can increase the strong sales of AI GPU of AMD, helping the company occupy a large part of the market. Now, AMD does not need to overtake Nvidia to supercharges its development. This bus needs to occupy the dual digits of AI GPU market over the next three years.

Assuming that AMD can capture 10% of the AI ​​GPU market by 2028, its revenue from this segment may be a hit by $ 50 billion (an estimated $ 500 billion annual revenue basis that the AI ​​accelerator market is expected to be a hit by 2028). This may be enough to help insert more upside down than Nvidia.

Why AMD sounds priced for more reverse

AMD total revenue concluded 2024 with less than just $ 26 billion. The company says it sold more than $ 5 billion data centers GPU last year, meaning that non-AI data centers GPU businesses contributed $ 21 billion. Assuming that it does not increase the rest of its sections in the next three years (and produces $ 21 billion in the annual revenue from its non-AI GPU business in 2028), but AI manages to hit $ 50 billion in GPU revenue, its annual top line can be $ 71 billion after three years.

The stocks are currently trading on a sale of 8.4 times, which is almost in line with the average sales of the US technology sector 8.1. Assuming that AMD trades with many in the technology sector after three years, its market capitalization can hit $ 575 billion based on the expectation of being generated in 2028. This indicates a possible benefit of 150% from existing levels.

On the other hand, Nvidia is trading on an expensive Price-to-sell ratio Of 26. However, it would not be astonishing to see Nvidia trading at a discount, with the company’s hopes of slowing down in the next few years. For example, AMD’s upcoming chips cannot be rejected the possibility of extending more than Nvidia. So AMD looks like a solid AI Stock To buy now.