Nvidia’s AI GPU shortage could last until 2025

The global surge in demand for artificial intelligence (AI) capabilities has led to an unprecedented interest in specialized hardware solutions, particularly GPUs. Nvidia, a leading name in this space, is witnessing a skyrocketing demand for its AI-focused GPU and CPU products. However, this burgeoning market is encountering supply chain issues, primarily due to overwhelming demand. According to Taiwan Semiconductor Manufacturing Company (TSMC), the primary manufacturer for many of these advanced chips, the intense need for AI hardware is causing significant disruptions in the supply chain. Consequently, we’re entering an era marked by a shortage of AI-specific GPUs.

With a dominant position in the AI hardware market, Nvidia has become the go-to supplier for companies looking to augment or transform their existing systems. Yet the scarcity of these crucial components may persist for an extended period, affecting businesses worldwide. This imbalance between supply and demand highlights the challenges the industry faces as it seeks to scale AI applications rapidly.

Artificial intelligence GPU shortage has begun

The AI boom has led to a spike in demand for specialized hardware, notably GPUs, putting immense pressure on supply chains. Taiwan Semiconductor Manufacturing Company (TSMC), a key player in the chipmaking industry, has sounded the alarm, warning that the shortage of AI-specific GPUs could last for up to 18 months. The company’s chairman, Mark Liu, attributes this to the meteoric rise in the development of GenAI (Generative Artificial Intelligence), which demands robust computing power.

TSMC acknowledges that it was caught off-guard by the sudden bottleneck in CoWoS packaging, a critical step in AI GPU production. Orders for CoWoS packaging services have surged to three times the usual volume. Despite plans to double production capacity by 2024, TSMC believes that the scarcity may persist at least until 2025.

Given the urgency, Nvidia is diversifying its supply chain and has engaged with Intel and Samsung for HBM and GPU packaging services. This strategic move aims to alleviate some of the immediate supply-chain pressures while TSMC ramps up its capabilities.

Both Nvidia and AMD have projected that their AI-based revenues could catapult into the hundreds of billions of dollars range. However, these forecasts may be overly optimistic, given the existing production constraints. For Nvidia, however, the outlook appears more favorable due to their timely introduction of the H100 GPU. The company aims to sell between 1.5 to 2 million H100 units by 2024.

While both companies are grappling with production shortages, the end of the tunnel is in sight. Once these supply chain issues are resolved, the demand for AI GPUs is likely to be met, opening the floodgates for unprecedented growth in the AI industry.