If forecasting component demand feels harder than it used to, you’re not alone. What used to be a relatively stable process, based on historical demand and predictable lead times, has become far less certain over the past 12–18 months. A big part of that comes down to AI explains Chris Withers, sales director at Zel Components.
IDC forecasts global semiconductor revenue to rise 52.8 per cent to $1.29 trillion in 2026, driven largely by accelerating demand for AI and memory technologies. That’s a huge jump in demand, but it’s not evenly distributed.
The report points to a broader rebalancing of supply chains, with pressure shifting toward mature-node production and supporting technologies across global manufacturing bases. This includes China, as demand from AI infrastructure changes how capacity is allocated.
For example, memory has become a bottleneck because as AI data centres expand, they’re absorbing a growing share of global dynamic random-access memory (DRAM) and NOT AND (NAND) output.
In fact, it’s estimated that the shortage is likely to continue until around 2027, with the top U.S. and South Korean suppliers raising DRAM production at a pace that will meet only about 60 per cent of demand.
However, this ripple effect hits availability across the wider electronics market. Memory prices have climbed, with increases of up to 90 per cent in early 2026 compared to late 2025, according to Counterpoint research.
For procurement teams, this means that demand at the product level might look steady, but availability at the component level can change quickly, and without warning.
When demand doesn’t behave as expected
Passive components are also starting to tighten, particularly those used in power-heavy and high-performance applications. High-capacitance MLCCs, polymer capacitors and tantalum capacitors are all seeing increased demand.
This is largely because of the role they play in supporting AI and high-density computing systems, so as a result, pricing has started to move. For instance, Panasonic has informed distributors and direct customers of 15-30 per cent increases on select tantalum capacitor models.
For manufacturers working on display modules, embedded systems or power-dense designs, this creates a knock-on effect. Memory constraints can influence system architecture, while capacitor availability affects board stability and overall design choices.
Simultaneously, not everyone is impacted equally. Larger OEMs tend to have the advantage of forward ordering, which gives them a buffer against volatility. SMEs, on the other hand, are likely to feel the impact of spot pricing, allocation issues and sudden lead time changes.
Moving from lean at all costs
In this environment, strict lean inventory strategies can start to show their limits. Holding minimal stock works well when supply is reliable, but it can leave teams exposed when it isn’t.
That doesn’t mean companies are abandoning efficiency, but there is greater emphasis on flexibility, with alternative distribution methods. Having approved pin-for-pin equivalents means teams have options when a primary supplier runs into issues, whether that’s pricing, availability or lead times.
There’s also more interest in distributed inventory models. Instead of relying on a single source or channel, companies are spreading risk across franchised distributors, independent stock and global inventory pools.
In addition, research highlights growing use of AI-driven tools in supply chains to improve forecasting and identify potential disruptions earlier in the process.
Planning for the unknown
This is because AI-driven demand has introduced a new level of unpredictability into the supply chain, and it’s not something traditional forecasting models were built for.
With analysis suggesting that memory constraints, in particular, are likely to persist well into 2027, the ongoing imbalance between demand and supply is making planning far more difficult across the electronics ecosystem.
Instead of trying to forecast perfectly, teams are having to respond more quickly when things change. That means designing with flexibility in mind, qualifying alternative components earlier and avoiding reliance on single sources where possible.
After all, it’s not only forecasting that’s harder. It’s the conditions it depends on — from memory availability through to capacitor supply — that are moving faster than the models built to predict them.
For more information on pin-for-pin alternatives and ways to reduce distribution risk across current designs, visit the Zel Components website.
Click here for more news: https://designsolutionsmag.co.uk/category/news/
