AI growth depreciation poses a risk for tech, markets. Here's why.

Nvidia stock price has crossed above $200 per share on Tuesday after CEO Jensen Huang unveiled a wide array of partnerships and AI innovations at the chipmaker's GTC event (GPU Technology Conference).

Monachil Capital Partners Managing Partner and CIO Ali Meli sits down with Josh Lipton to discuss the valuation risks associated with AI, going in-depth into the rise in GPU leasing prices and depreciation in hyperscalers' depreciation in AI revenue growth.

To watch more expert insights and analysis on the latest market action, check out more Market Domination Overtime.

Video Transcript

00:00 Speaker A

Well stocks trading at new highs here, but the next big test will come with big tech earnings this week. Our next guest notes AI as a whole has contributed to market frothiness. For more, let's get to Mona Cho at Capital Partners, managing partner and CIO, Ali Meli. Ali, it is great to see you. Let's let's start there. Uh talk about AI because Nvidia of course has the big show in our nation's capital today. They made all kinds of new news, Ali. The stock popped about 5% in today's trade. You say here, I thought this was interesting. Talk about AI, Ali. You say very promising in boosting productivity, but you think valuations have gotten ahead of market reality and you say you see risks. So let let let's stop there. Where do you see the risks, Ali?

01:06 Ali Meli

The risk that I see is mainly in valuation and accounting. So, let's taking a step back. The average leasing price for an Nvidia H100 GPU in at the beginning of 2024 was $10 per hour. Now that number has dropped to $2 per hour. So, you see an 80% depreciation in equivalent of rent that you can pay or you can, if you were an owner of a GPU, this is how much you can earn for renting out your GPUs. On the other hand, if you look at the depreciation curve, a lot of the buyers of Nvidia chips use four year, five year depreciation. So, they only depreciate that by 20% per year. Now, where does that translate into corporate earnings? That means that you get exaggerated corporate earnings in the sense that you get the revenue growth up front, but the depreciation is back-ended. If you look at the hyperscalers in the past year, they have spended around $400 billion dollars on AI related expenditures. Their total revenue growth, whether that was from AI or other sources was only $200 billion dollars. So, if they were to actually depreciate the the AI expenditure and look it as a upfront expense, that would have been a loss. Now, because they amortize the cost that they are putting right now over a 5-year period, they are showing income growth on top of revenue growth. But the question is is that revenue growth sustainable? And if they actually were going to use and a depreciation curve that captures the economic reality of it's a new technology, the same way that like a used cell phone depreciates a lot in value. If you use the same depreciation logic, if you look at Moore's Law for example, if they were to depreciate their assets by 50, 60% over a 2-year period, I think a lot of their earnings would be significantly lower.

View Comments