Generative AI first burst onto the scene with the launch of ChatGPT on November 30, 2022. Up until then, hardly anyone had heard of the monumental capabilities that even the first version of ChatGPT boasted. It was the beginning of a revolution that began a race among tech companies to build AI infrastructure.
As all companies rushed to spend enormous amounts of cash on buying AI equipment, nobody asked the question: how did ChatGPT build its platform when AI infrastructure wasn’t even a thing? The answer lies in Nvidia’s Hopper GPU which was released in 2022. Considered one of the most powerful GPUs, the unique selling point of the GPU wasn’t AI. It was just the most advanced GPU Nvidia had built by that time.
In other words, Nvidia still hasn’t released a GPU that was ‘built for AI’. The upcoming Blackwell architecture will be the first true AI GPU. The company put a lot of resources behind this version, ensuring that it would fulfill the needs of all the major tech companies looking to accelerate the training of their AI models.
If the Blackwell GPUs are truly the first AI GPUs, surely the AI spending is only about to begin. Any company that is serious about gaining a lead in developing AI technology will need to be on these GPUs from day one. Big Tech, which is the biggest AI spender by a long shot, simply has no other choice but to buy the new advanced GPUs to stay in the race.
On the flip side, there are issues with scaling the Blackwell GPUs, the likes of which we hear almost every week. It is quite unlike Nvidia to face execution issues when launching a new GPU. But there are good reasons for these problems, if one may call them that.
When companies build their data centers, they need a whole lot of equipment in addition to the GPUs to make it run successfully. Blackwell GPUs are so advanced that some companies will need to change their whole data centers to accommodate these. In other words, the supporting equipment that needs to go in a data center to support Blackwell GPUs still isn’t perfect. For example, which liquid cooling racks work best with the GPU? How do data centers without Direct Liquid Cooling deal with these new GPUs? What about the software? There are simply too many questions even for Nvidia to answer. And that’s not a bad problem to have, considering they’re the only company capable of providing this much computing power.
Investors worried about Nvidia’s supply chain issues or problems with Blackwell GPUs need not worry. We are in unchartered territories and even Nvidia can be allowed to have some slip-ups. However, Blackwell GPUs are without a doubt the most powerful GPUs in the world. After a quarter or two, companies will have figured it out. Then they’ll need more. That’s where Nvidia can fix its margin issues, bring in monetization, and we have the next bull run!
Nvidia is 5th on our latest list of the 31 Most Popular Stocks Among Hedge Funds. As per our database, 193 hedge fund portfolios held NVDA at the end of the third quarter which was 179 in the previous quarter. While we acknowledge the potential of NVDA as a leading investment, our conviction lies in the belief that some AI stocks hold greater promise for delivering higher returns, and doing so within a shorter timeframe. If you are looking for an AI stock that is as promising as NVDA but that trades at less than 5 times its earnings, check out our report about the cheapest AI stock.
READ NEXT: $30 Trillion Opportunity: 15 Best Humanoid Robot Stocks to Buy According to Morgan Stanley and Jim Cramer Says NVIDIA ‘Has Become A Wasteland’.
Disclosure: None. This article was originally published at Insider Monkey.
Hedge Fund Resource Center