Record SK Hynix Earnings Show How the AI Boom is Reshaping Electronics Supply Chains

The year 2024 will go down as one of the most schismatic periods in the history of the electronics supply chain, with the combination of the ongoing AI boom and weak demand from most other sectors producing radically different conditions depending on the market.

Most devices and end markets are set to generate either tepid demand growth or outright sales decline this year, leading to mainly favorable buyer conditions. However, the insatiable demand for
data-center-based AI processing triggered by the success of ChatGPT has spurred overwhelming sales growth for a select set of commodities, spurring rising prices and reduced availability.

The Impact on High-Bandwidth Memory (HBM)

Outside of Nvidia’s GPU lines, the most dramatic impact of the AI phenomenon has been in the memory market, where pricing, lead time, and availability conditions for high-bandwidth memory (HBM) have been extremely challenging this year. Specified for use with Nvidia’s GPUs in AI-enabled data centers, HBM has experienced a dramatic rise in demand.

A dramatic example of this trend appeared in late October, when SK Hynix reported fiscal Q3 revenue that represented an all-time high for the company, exceeding the previous record set in Q2. SK Hynix largely credited massive growth in HBM sales, noting a 70% sequential rise in revenue. The company also noted record profitability during the quarter.

The strong performance reflects SK Hynix’s success in winning an early lead in the HBM market, supplanting dominant memory supplier Samsung. In May, according to one market watcher, SK Hynix accounted for more than 52% of global HBM market revenue, compared to 42.4% for Samsung.

Given that Samsung is known for staying at the forefront of semiconductor memory market trends, the company’s early concession of HBM leadership to SK Hynix is surprising. HBM has emerged as the fastest-growing sector of the memory market, commanding high pricing because of its required use with Nvidia’s data center AI GPUs.

According to one market watcher, global HBM revenue is set to rise by more than a factor of three from 2024 through 2029.

How Buyers Can Stay Ahead of the Curve

The rise of SK Hynix at the expense of Samsung illustrates the bifurcated semiconductor market in 2024, with the companies that can capitalize on the AI phenomenon outpacing their rivals.

This is apparent in recent semiconductor market trends, which show that growth is concentrated among AI-empowered suppliers, leaving other players behind. Despite record revenue during the quarter, more than half of all global semiconductor suppliers had lower revenue than one year ago in Q2.

With few options to find alternative parts or suppliers for HBM, purchasers should focus on developing long-term partnerships with memory suppliers to ensure stable supply. Moreover, as AI capabilities expand from the data center to edge devices like PCs and smartphones, shortages of critical components could affect more application markets.

As a result, buyers should identify essential AI parts and work with the most AI-savvy suppliers to cultivate long-term agreements.