NEWS
A Memory Crunch to Remind Us of 2022
|
Access to Random-Access Memory (RAM), Graphics Processing Units (GPUs), and Central Processing Units (CPUs) is being challenged, as the global semiconductor market has turned toward supporting Artificial Intelligence (AI) infrastructure. Commitments to AI have placed pressure on all other aspects of the chip market, including consumers, utilities, and industrial companies. Industrial automation suppliers Beckhoff, Emerson Electric, and Siemens are experiencing higher Bills of Materials (BOMs) and lead times, driving margin pressures and slowing new product delivery. Strained access is hampering the ability for suppliers to provide their latest automation innovations: Industrial AI and Software-Defined Automation (SDA).
IMPACT
New RAM Requirements Might Make Manufacturers Forget SDA
|
Growth in SDA is expected to continue to grow at a 20% Compound Annual Growth Rate (CAGR) from 2026 to 2036, but getting these systems online will become more expensive than previously planned, influencing greenfield expansion plans. Compared to the on-chip CPU memory of Programmable Logic Controllers (PLCs), virtual PLCs (vPLC) require 3X to 32X as much memory (up to 320 Megabytes (MB) for large flagship PLCs versus up to 2 Gigabytes (GB) for the high-performance engineering for vPLCs). To accommodate these higher memory requirements, customers will need access to more powerful Industrial Personal Computers (IPCs), servers, or other sources of compute to run the vPLC.
Stronger IPCs are necessary to build the flexible systems that SDA aims to achieve, with software decoupled from hardware. But passing on higher BOMs to customers will hinder adoption, if reliable and more affordable alternatives are available. Higher IPC costs also interrupt a tenet of SDA—the shift of PLC costs from a high overhead Capital Expenditure (CAPEX) to an Operational Expenditure (OPEX).
RECOMMENDATIONS
Long-Term Thinking Might Solve Short-Term Memory Problems
|
Keeping costs low to drive SDA adoption will be vital in the coming years. However, price pressures are not guaranteed to ease until after new chip fabs are fully operational, because AI hardware cycles continue to move faster Therefore, industrial automation suppliers should make a concentrated effort to ensure that the cost of hardware capable of running multiple vPLC instances is low enough to drive SDA adoption by taking the following steps:
Do:
- Learn the Lesson of 2022, Again: Long-term supplier contracts provide steady pricing in the face of unstable supply chains, driven this time by strangled access to supply from semiconductor fabs due to AI and data centers. Long-term agreements prevent exposure to the cyclic nature of semiconductor supply stocks that are already racking the industry.
- Consider Eating a Hardware Loss: Companies going all in on SDA are already planning to transition some revenue to recurring, license-based automation. Deprioritizing hardware enables suppliers to consider not passing margin pressures on, instead understanding that capturing lifetime value from the software will be more valuable over time.
Don’t:
- Reduce Hardware Quality: Industrial hardware is built on the promise that it will last anywhere from 15 years for IPCs to 25 years for PLCs. A reduction in material quality is going to risk the most critical feature of automation systems—long life spans. Breaking this trust is a surefire way to become the biggest loser in the transition to SDA, and as open systems become more prevalent, best-of-breed hardware is becoming a requirement.
