- Solutions
ENTERPRISE SOLUTIONS
Infuse new product development with real-time intelligenceEnable the continuous optimization of direct materials sourcingOptimize quote responses to increase margins.DIGITAL CUSTOMER ENGAGEMENT
Drive your procurement strategy with predictive commodity forecasts.Gain visibility into design and sourcing activity on a global scale.Reach a worldwide network of electronics industry professionals.SOLUTIONS FOR
Smarter decisions start with a better BOMRethink your approach to strategic sourcingExecute powerful strategies faster than ever - Industries
Compare your last six months of component costs to market and contracted pricing.
- Platform
- Why Supplyframe
- Resources
The Commodity IQ Price Index for memory surged again in May, as booming demand and constrained supply of artificial intelligence (AI) memory chips boosted the reading to a four-and-a-half-year high. The index rose by 5% sequentially in May to more than three times the index baseline, marking the highest level since October 2019.
The stratospheric reading baseline indicates that pricing is rising at a robust pace.The price index has risen every month since February and has soared by 42% since the beginning of the year.
High-Bandwidth Memory is Driving the AI Boom
The strength and duration of the price runup – combined with the high price index – are being driven primarily by booming demand for high-bandwidth memory (HBM), a speedy type of DRAM suitable for use in AI data center server applications.
Specified by NVIDIA for use with its AI graphics processing units (GPUs), HBM employs stacking technology to combine multiple DRAM dice in a single package. This approach allows HBM to achieve higher bandwidth and a smaller form factor than alternative high-performance DRAM types.
Riding the generative AI boom triggered by the emergence of ChatGPT, HBM sales are expected to account for more than 20% of the total DRAM market in 2024, up from less than 10% in 2023, according to market watchers.
SK Hynix is the early leader in HBM, accounting for as much as half of market revenue, according to researchers. However, Samsung and Micron are striving to catch up, with the companies expanding production dramatically, while increasing output of newer part types, like HBM3 and HBM3e.
As a result of the booming demand and limited supply base, SK Hynix and Micron have sold out of HBM for all of 2024 and are nearly sold out for 2025. The imbalance in supply and demand is causing HBM average selling prices to soar, costing as much as five times as much as DDR5 DRAM.
However, as suppliers shift production capacity away from alternative DRAM types to HBM, prices of other types of memory could increase as well. Most non-HBM DRAM types are experiencing flat pricing conditions.
For HBM buyers, the current situation provides few good alternatives.
Are There any Alternatives for HBM Buyers?
In the highly commoditized world of semiconductor memory, buyers often can find alternative parts or sources that allow them to meet their price and delivery requirements. However, with HBM serving as the designated memory for NVIDIA GPUs, employing an alternative DRAM technology is not an option in the AI market. Furthermore, with HBM manufacturing heavily concentrated at one company, SK Hynix, buyers don’t have the opportunity to play different suppliers off against each other.
Under these circumstances, HBM buyers should prepare for an extended period of inflated prices and limited availability, possibly lasting through the end of 2025. (NEW P) However, with the HBM boom potentially impacting the pricing and availability of other memory types, such as DDR5, buyers should take steps to address the situation, by preparing to build safety stocks and lock in pricing as much as possible.
It should be noted that if the AI boom goes bust, supply conditions could change dramatically in favor of buyers. The purchasing community should monitor the demand situation for all types of AI chips and be on the lookout for any signs of order cancellations or delays. Such news could signify a sudden end to AI-driven chip demand surge, similar to the dot-com bust of 2000.