The artificial intelligence (AI) wars are heating up. Software companies are battling it out for consumers' attention, but hardware giants are also jostling for position in what could become a multitrillion-dollar revolution for global data centers. 

Advanced Micro Devices (AMD 1.14%) held its data center and AI technology event on June 13, where it revealed its long-awaited MI300 chip designed for AI workloads. Its goal is to snatch some of the dominant market share in this segment currently held by its key competitor, Nvidia (NVDA -1.99%)

The new MI300 chip is already slated to power the world's fastest supercomputer, which is expected to come online later this year, but AMD might be set to score an even more lucrative customer. Here's why investors will want to add AMD stock to their portfolios right now.

The MI300 could be a game changer

Nvidia currently has an estimated 90% market share in the data center chips designed for AI. The company's CEO, Jensen Huang, says there's about $1 trillion worth of existing data center infrastructure that needs to be upgraded to support accelerated computing and AI workloads, so it's no surprise AMD is trying to capture a slice of that opportunity. 

The tech world has eagerly anticipated the specifics surrounding AMD's new MI300 chip, and now that the cat is out of the bag, there's a lot for investors to look forward to. There will be two variations of the MI300. First, the MI300A is a 3D stacked chip package that combines both CPUs and GPUs to create what AMD calls the first-ever advanced processing unit (APU) for data centers.

It's the package of choice for the upcoming El Capitan supercomputer at the Lawrence Livermore National Laboratory (LLNL), which will be the organization's first exascale platform with two exaflops of performance. It's widely expected to be the world's fastest supercomputer. Here's how the LLNL describes its capabilities: If all 7.7 billion people on Earth completed one calculation per second, it would collectively take them eight years to do what El Capitan can do in one second

Then there's the MI300X, which is a GPU-only chip specifically designed for training large language models up to 80 billion parameters, which AMD says is an industry record. Customers can also get the MI300X on what AMD calls the Instinct platform, which combines eight of the chips into an industry-standard configuration, allowing them to plug straight into existing data center infrastructure with minimal adjustments. This could rapidly speed up deployment and also drive adoption. Speaking of which...

The world's largest cloud provider might soon be a MI300 customer

Amazon (AMZN 0.58%) Web Services (AWS) is Amazon's cloud services platform. It's the largest in the industry by revenue and also by number of solutions. Here's an interesting fact to keep in mind: AWS is one of Nvidia's oldest data center customers.

But according to a report by Reuters, AWS is interested in AMD's new MI300 chips. While the cloud giant hasn't commented publicly, its vice president of elastic compute cloud, Dave Brown, said it's considering using them. He commented that AWS typically benefits from AMD's designs that plug directly into existing systems, which is a small insight into how some of the largest cloud providers might be thinking about future upgrades. 

It's important to note that AWS recently signed a new deal with Nvidia for its new EC2 P5 cloud infrastructure that will help its business customers seamlessly scale up from 10,000 Nvidia H100 GPUs to 20,000, which will enable them to train AI models significantly more quickly. But considering AMD's new MI300X has 2.4 times more high bandwidth density compared to the H100, it's no surprise it has the attention of AWS potentially for other projects. 

Over time, AMD has a real opportunity to dent Nvidia's market share in this segment. 

AMD stock is a buy right now

Nvidia recently achieved a $1 trillion valuation on the back of its most recent earnings report, where it blew away Wall Street's expectations and increased its revenue guidance for the current quarter by a whopping $4 billion (relative to the Street's estimate). AMD's MI300 hardware is slated for release later this year, and if the company manages to sign a few high-profile customers like AWS, it could also see a powerful boost to its financial results. 

AMD stock has doubled this year, but it's still down 18% from its all-time high, which is in stark contrast to Nvidia stock, which is reaching new heights almost on a daily basis right now. If AMD manages to snatch some market share in the AI data center space over the next few quarters, that should really boost its stock. 

AMD is currently valued at $204 billion, but I think it has a pathway to joining Nvidia in the $1 trillion club within the next decade, so long-term investors could stand to benefit the most.