Demand exceeds supply. The production cost is even twice as high
The artificial intelligence craze is increasingly disrupting subsequent links in the global supply chain. The problem concerns not only DRAM memory or advanced packaging techniques, but primarily the production of semiconductors themselves. The demand generated by data centers and AI accelerator manufacturers has grown so large that even TSMC is having serious difficulty keeping up with orders.
The rush is driven by profits running into the billions of dollars
The latest analysis prepared by JPMorgan shows that customers are now willing to pay up to 100% more to have their systems produced in an accelerated mode. Delivering the finished chip to the market becomes a priority, and manufacturing costs become a secondary consideration. In practice, this means launching the so-called hot runs, i.e. emergency production runs carried out outside the standard queue.
Although the report does not indicate specific companies, it is easy to guess that the largest players of the AI market, including NVIDIA and AMD, are among the customers fighting for priority. For them, a fast delivery rate is crucial, especially in the context of short new product launch cycles and enormous pressure from the server market.
This is particularly visible in the case of the Greens, who in recent years have accustomed the industry to introducing subsequent generations of accelerators at intervals counted in months. To maintain this rhythm, the company must rely on the most flexible approach of its production partners. This situation shows how much the AI boom is changing the rules of the semiconductor game and how far customers are willing to go today just to obtain production capacity on time.
