Nvidia GPU Shortage: Exploring Factors and Industry Responses

Nvidia GPU Shortage: Exploring Factors and Industry Responses

With rising demand for AI tools, we have seen a significant rise in the shortage of compute GPUs for them. The Chairman of TSMC (Taiwan Semiconductor Manufacturing Company) has reported that one of the major causes of shortages is the limited packaging capacity of CoWoS (chip-on-wafer-on-substrate). In this article, we are gonna have a look at the main causes of Nvidia’s GPU shortage caused by a shortage of CoWoS.

What causes Nvidia’s GPU Shortage?

High demand in GPU is also a main cause of shortage but the shortage of CoWoS packaging is another topic to discuss. Mark Liu has reported that it’s not mainly a shortage in AI topics but further in our CoWoS capacity. TSMC is one of the major manufacturers of AI processors but with the high demand for AI processors like Nvidia’s A100 and H200 compute GPUs, they are unable to fulfill that need because of a lack of hardware resources.

Current Supply

At this time, TSMC has covered more than 80% of the raised demand but the rest of the 20% gap couldn’t be completed due to a lack of CoWoS capacity constraint. It’s been more than one and a half years to cover the rest of the demand. It will be possible soon as TSMC is trying to expand its packaging facility.

How Tech Giants are Solving the CoWoS Challenge

A sudden hike in AI-generative tools that requires a good GPU computing power to perform tasks, needs a good graphics card to do that. This is one of the major causes of GPU shortage in the market. The company has ensured to double its CoWoS packaging capacity by the end of 2024.  They’ve dedicated a $2.9 billion investment to build their new facility in Taiwan. This move can fill the gap of shortage. Because of this, we may see a drop in the prices of GPUs.

There are other companies like TSMC that emphasize advanced chip packaging, Intel and Samsung are also spotlighting this bottleneck area. Intel is projected to quadruple its advanced packaging capacity by the year 2025. Also, other outsourced semiconductor assembly and test (OSAT) companies are trying to do the same.

Conclusion

In demand of AI generative tools, we have seen a huge hike in demand for GPU. To fulfill this hike in demand, companies like TSMC are doubling their CoWoS packaging capacity. They’ve raised a significant amount of investment to do that in the next couple of years. Most probably it will fill the demand of AI and HPC sectors.

Haseeb Sagheer
I'm Haseeb Sagheer, a tech writer with a BS in Software Engineering. With a passion for both technology and gaming, I've spent seven years in the gaming community as a player and writer. My expertise lies in simplifying complex tech ideas for engaging content. Let's explore the tech world together.