Cathie Wood — whose Ark Investments recently divested a bulk on its NVIDIA stake ahead of the stock’s massive rally last week — thinks the chipmaker’s shares are too costly. This isn’t the first time the chipmaker has been over-valuated though. In late 2021 NVIDIA’s share price had experienced a major parabolic trend based on hype. With the advent of ‘internet influence’ and ‘follower culture’ a person, company, brand, or ideal doesn’t really have to have any inherent value to be valuable or monetizable. Social influence can flip elections, get great jobs, market goods, create bank runs, and drive up the price of securities. In April NVIDIA’s market cap was $666.66 Bn and it now sits at $982 Bn. Almost a 50% increase. All of this while AMD and NVIDIA won’t even be able to reap the full benefits of the $52 billion windfall from the U.S. government for the CHIPS Act based on the fact that they are fabless and contract with outside firms like TSMC and Samsung to produce their chips.

There is talk about NVIDIA’s artificial intelligence chip, designed specifically to make building AI systems such as ChatGPT faster and cheaper. However, when asked what an AI chip is, Hannah Dohmen, a research analyst with the Center for Security and Emerging Technology stated “There really isn’t a completely agreed upon definition of AI chips.” NVIDIA is a meme stock. People are impressionable. The company has a lot of credibility for its GPUs in the video game world and investors are showing it love based on trends. While they didn’t invent the visual display processor, they did coin theirs as the GPU (Graphics Processing Unit). The CUDA (Compute Unified Device Architecture) platform is a software framework developed by NVIDIA to expand the capabilities of GPU acceleration. It allows developers to access the raw computing power of CUDA GPUs to process data faster than with traditional GPUs. Which just sounds like another term to describe a multi-core architecture. However, it is a fact that the 256 Grace Hopper superchips, which are stitched together in NVIDIA’s own DGX GH200 supercomputer, are combined to process 144 terabytes (TB) at a time. Which just sounds like they’re combining hardware like the case of the Spectrum X platform. So, there are two sides on the coin to consider.

How difficult will it be for other chip designers to compete with NVDIA? Probably not very difficult at all. They have, however, garnered the support of speculative investors for the time being.