I read the content in the post below about the AI bubble warning theory due to defaults in collateralized loans of private equity funds.
Since around last fall, NVIDIA's stock price has been struggling. Recently, despite tremendous performance, the stock price is falling.
I believe that one of the reasons the stock price is falling is due to rising memory prices.
Big tech companies' CAPEX investment is said to be 600 billion dollars this year. In a situation where there is skepticism about whether this is actually possible, given concerns of a bubble and all that,
Significant increases in memory prices inevitably lead to reduced purchases of NVIDIA GPUs.
Within the limited budget of 600 billion dollars, they need to build data centers, lay down power infrastructure themselves (a few days ago, President Trump gathered big tech companies and told them that from now on, they should figure out laying down the power infrastructure for AI data centers themselves), and they also need to buy GPUs and memory. If memory prices spike here, what would need to be cut? The only thing that can be reduced is the quantity of GPUs.
This year, several securities firms predict Samsung Electronics' operating profit at 200 trillion (135 billion dollars) and SK Hynix's profit at 150 trillion (100 billion dollars). Combined, their operating profit alone is 235 billion dollars. Expecting the profit margin to exceed 50%, the memory sales of both companies alone should reach 500 billion dollars. I see the proportion of memory going to data centers and AI in the memory market to be at least 60-70% this year, then AI data center memory sales alone would be 500 billion dollars * 65% = 325 billion dollars.
Doesn't that seem strange?
Big tech companies' CAPEX investment is 600 billion dollars this year, and half of it goes to memory..? Then what about the cost of building data centers? What about the cost of buying GPUs?
In other words, unless money is falling from the sky, they have no choice but to significantly reduce the quantity of GPU purchases...
Originally, when memory booms, set prices rise, causing the set market to shrink significantly. Then set manufacturers and system semiconductor companies suffer. Then as demand decreases, the memory market also enters a downcycle... and this cycle repeats.
This time, I think GPU might be the trigger.
Besides AI, there are also views that the mobile and PC markets, due to sharp price increases, will see the entire market shrink by at least 10% to 30% this year. Since consumer wallet size is fixed, it's basic economic knowledge that when price (P) increases, quantity (Q) decreases.
Samsung's profit is 500 trillion... Samsung and SK Hynix combined are opening an era of 1000 trillion operating profit... This is the time to be careful of people who just write numbers in Excel and go all out. Because the market is not just Excel numbers.
When one side is good, the other side has to pay an equivalent price.
Anyway, I've thought that with these memory price increases, as we go into the latter half of the year, NVIDIA's GPUs might not sell as well as expected, and I've somewhat imagined that smart money might catch this first and move proactively.
You know the above is fiction, right? ㅎㅎ
This was Pazz.
