Key Takeaways
- Edge computing allows faster responses on factory floors by processing data locally rather than sending it to cloud servers.
- A shift in memory allocation for AI products over standard industrial controllers is driving up costs and limiting availability of conventional DRAM.
- Designing for edge computing deployments now emphasizes memory management, which impacts procurement strategies and specifications.
Understanding the Shift in Edge Computing
Edge computing has emerged as a solution for factories aiming to enhance operational efficiency by processing data locally, preventing delays associated with relying on cloud services. This technology empowers machines to respond in milliseconds to sensor readings and camera footage, maintaining production continuity even if internet connectivity falters, while also reducing bandwidth expenses.
The impetus behind this year’s changes stems not from the manufacturing sector but from developments in memory technology. DRAM (dynamic random-access memory) is crucial for computers, enabling the software to function effectively. AI models used for monitoring products require significant amounts of DRAM, given the need for real-time processing. The primary suppliers of this memory are Samsung, SK Hynix, and Micron, whose production lines serve both the for standard industrial applications and higher-margin AI products.
As demand for high-bandwidth memory (HBM) for AI accelerators increases, conventional DRAM production is deprioritized by suppliers, which in turn results in constraints in availability and rising prices. According to TrendForce, the second quarter of 2026 is expected to see a 58% to 63% increase in conventional DRAM prices, alongside a 70% to 75% rise in NAND flash costs. They reported a similar but decelerating price rise in the third quarter.
The challenge mainly concerns older technology like DDR4 memory, which is commonly utilized in industrial applications due to its reliability over a decade-long service life. Although consumer electronics have transitioned to DDR5, many factory systems remain rooted in DDR4 or even older DDR3 technology. The limited capacity for legacy memory production exacerbates supply challenges, especially for companies like VersaLogic that construct industrial single-board computers.
Variants of memory complexity across the plant reveal a disparity in the impact of these shortages. Basic controllers generally require less memory and remain stable, while the newest AI components, which demand significant memory capacity, are the most affected. As a result, those investing in cutting-edge AI technologies are experiencing the most severe consequences of supply chain constraints.
Organizations need to rethink how they specify edge computing systems. Memory management must evolve from being a mere component of procurement to become an integral part of the design phase. By adopting more efficient models that require less DRAM, it’s possible to lower costs and reduce the amount of memory necessary. Additionally, establishing procurement terms early in the planning process is crucial, especially as larger buyers secure agreements ahead of time.
The necessity for hardware to support memory from multiple suppliers is also gaining attention, as diversifying supplier options can mitigate the effects of prolonged lead times.
Despite the challenges presented by shifting memory dynamics, the strategic case for implementing edge computing solutions remains robust. Organizations that adeptly navigate these procurement complexities will be better positioned to implement effective edge computing nodes within their manufacturing operations this year.
The content above is a summary. For more details, see the source article.
