HBM Supply Chain Feels Pressure: Is a Shortage Coming Soon?

Supply chains are prone to challenges, particularly when it comes to acquiring technology, but it seems the more and more businesses attempt to bolster the fast-paced evolution of the industries, the harder it becomes to keep up with digitalization as a component HBM manufacturer. 

The demand on parts makers has never been greater. As digital solutions form the backbone of society–as shown recently through the global impacts of the Microsoft outage—innovators in the upstream technology supply chain are playing a continuous game of catch-up. The latest to dawn on the global market, as reported by SK Hynix, is an impending disruption to the supply to high-band-width memory chips. 

It is possible for chip makers to rebound after experiencing an extremely high demand for their parts. Currently there is an abundance of NAND (used for digital cameras and USB flash drives) and DRAM (processing) chips, representing the rebounding effect of chip makers in the memory market. But, with increasing use of artificial intelligence (AI), and the demands on higher power in more compact units, the pressure is on chip suppliers to innovate as there are little-to-no signs of their customers slowing down. 

An executive at SK Hynix stated they expect to see HBMs and DRAMs soak up 61% share of the global memory chip market by 2028. The question is: will these unfavorable odds act for or against HBM suppliers? 

SK Hynix Reports an Impending Shortage of HBM

The answer to every technological challenge is “more bandwidth”, but the solution is not clean cut, particularly when we look at the scale of the challenge. As major tech companies continue to ramp up their processing power in more compact packages, the demands on chip providers become greater and it would just so happen that data-center providers, consumer product companies, and other industries heavily dependent of these components are accelerating their output much faster than suppliers can handle—or, at least, with minimal margin for delays. 

As a result, SK Hynix expects a shortage of HBM chips until 2026. Although there are other players in the industry, such as Micron Technology with a minimal share of the market (4-6%) and Samsung Electronics, the second-lead to SK Hynix, TrendForce suggests that the HBM supply will soon disappear, leaving the industry in potential disarray. 

There’s no blame to be shed, so it seems, as the shortage is primarily a result of booming and disruptive technologies—which is a positive thing in itself. Demands on faster GPUs and CPUs will inevitably continue, only HBM suppliers become major factors in the rate of innovation. 

However, simply stating that products are sold-out in advance for 2025 doesn’t necessarily suggest a shortage. It’s important to look at the market and whether customers’ needs have been met for the next 18 months. 

Current HBM Supply Chain Challenges

This component shortage is much more complex than the previous one. In the past, a lack of semiconductors was a direct result of further upstream delays—a reduction in parts availability earlier in the supply chain. The recently anticipated shortfall is due to more than resources, but in fact the rate of change that technology firms want to see in order to embed more advanced software in their products. 

For example, increasing the use of AI demands two things: larger datasets and faster memory processing power. While the former is a matter of expanding data centers’ capacity, the ability to process data at record speeds is what makes such a compelling case of AI. 

Big names like Apple, Microsoft, and Google have already set out their plans to use AI in their smart devices, which has already caused a shift in their approach—by developing their own chips with AI in mind. 

Potential Impacts of a HBM Shortage

There are both short-term (the next couple of years) and long-term impacts of an HBM shortage, particularly as companies—both suppliers and customers—will have to navigate the challenge. 

Supplier Change-Up: With a short supply comes a need to switch suppliers or acquire components from multiple sources. The leading provider of GPU units for AI NVIDIA has already made the decision to gather HBMs from other companies, making it less dependent on SK Hynix despite its majority share of the supply chain. Moreover, as companies expect a shortage, they are likely ramping up their procurement efforts to ensure that they can leverage new sources with small-batch suppliers to bolster the impact. As a result, companies like Samsung and SK Hynix are dependent on their ability to weather the storm, and we could see an entirely new division of the industry post-2026. 

Companies Become Their Own Suppliers: It has already been reported that Huawei is developing its own HBM chips. While this was meant as a response to the US sanctions against the use of US-made components, Huawei may be onto something. Much like the changes happening at the business-to-consumer (B2C) level, companies with the power to make their parts could begin to invest in more complex, compact parts for the future. 

Increased Interim Costs: Much like we’ve seen this the coronavirus pandemic dwindled, the cost implications of a shortage can extend far beyond the inflection point. While companies not only have to overcome the immediate price battle in the wake of the HBM shortage, the trailing effects could be long-lasting. Previously, semiconductors have been known to cost 20% more as a result of supply-demand imbalance, which could also be a factor that drives companies to expand their development efforts—a driving force behind the internalization of chip development among big tech businesses. 

The risk to the leading HBM suppliers in this scenario is that major tech companies become even wiser to the idea of producing their own products to plug the gaps—not just to respond to shortages, but also understanding the return on investment (ROI) in their own development of components. 

Slowing Innovation and Production: With AI implemented in various new industries, such as the automotive and energy sectors, the rate at which we could see more intelligence in these areas could be stunted by an inability to get components. SK Hynix already holds the lion’s share of the HBM segment, and states its parts are sold out for the next year, so questions will likely be answered in the next year or so. 

An Alternative Thought on HBM Shortage

Are we too quick to assess whether a shortage is on the horizon? Although the leading HBM supplier stated it sold out for 2025, this could simply be a hindrance to new and smaller technology companies. Although there is potential for prices to remain, the fact that many customers may not be able to acquire advanced chips until 2026 could scupper the efforts of emerging startups in the AI space. 

NVIDIA is one of the main customers of SK Hynix, which could prove a blessing for the company as its CEO grows wary of the increased commitment from big tech companies to make their own chips. The company acquiring more granular components for finished semiconductors would help maintain the company’s leading position, but new and emerging technology development projects will be hot on its tail.