Understanding the memory market in 2026

The memory market in 2026 is not a single story – it is two stories running simultaneously and understanding which one you are in determines whether you are thriving or struggling.

That is the framing offered by Nikolaos Florous, Global Product Marketing Director at Memphis Electronic, who spoke to Procurement Pro following his presentation at Hardware Pioneers MAX in London.

Florous describes the current landscape as two market cycles operating in superposition: the AI investment cycle, which is driving extraordinary demand for advanced memory products, and the regular memory volatility cycle, which has never gone away. Right now, the AI cycle is dominant – but for how long, and at what cost to the rest of the industry, are the critical questions facing procurement teams and engineers today.

The AI cycle: a trillion-dollar trajectory

The numbers underpinning the AI investment cycle are striking. Florous estimates that hyperscalers and the ‘Magnificent Seven’ will collectively invest somewhere between $600 and $700 billion in the industry this year, with that figure on course to reach $1 trillion by 2028.

“There is a lot of strong investment, and this will definitely drive the demand for advanced technology node memory products like HBM via architectures like DDR5, LPDDR5, and so on,” he said.

The practical consequence for the memory market is a sustained period of high demand and high prices for premium product lines – a dynamic that is likely to persist for at least the next two to three years, regardless of any broader economic headwinds.

Is this the dot-com bubble? Not quite

The pattern of investment circulating between NVIDIA, OpenAI, Microsoft, and back again has drawn comparisons to the dot-com era, when telecom companies invested in equipment makers who in turn bought more telecoms services to justify the expenditure. Florous acknowledges the resemblance – but argues there is one critical difference.

“This time, however, is a little bit different because we see that the memory suppliers are getting the ROI back in spades. Revenue is booming, profits are booming. If there was zero ROI, then I would have more concerns.”

That said, the ROI picture is less clear when it comes to AI adoption more broadly. Florous points to the slow pace at which enterprises are integrating AI solutions and cites Amazon’s widely reported failure to replace a cohort of staff with AI agents as evidence that the technology still has significant limitations in practice. The more likely near-term model, he suggests, is human-AI collaboration rather than replacement.

A second wave of AI investment may come when intelligence moves from the Cloud to Physical Edge AI devices – a development Florous expects to generate another significant demand surge, though one that lies slightly further out on the horizon.

Market correction: not if, but when – and how much

Florous is clear that some form of normalisation is coming after 2028, when the underlying memory cyclicality will reassert itself. What he is equally clear about is that “normalisation” does not mean a return to the pricing environment of mid-2023.

“Are we going to see the prices we were experiencing back in the middle of 2023? This is likely not going to happen. There will be some normalisation – we are talking about maybe 30 or 40% priced down after 2028 for certain product lines – but they will not reach the levels we had back then. Prices will stay still quite significantly high.”

For procurement teams, this means planning on the basis of sustained elevated pricing rather than expecting a cyclical reset to previous lows. The floor has shifted.

What engineers and procurement teams should do now

The advice Florous offers to design and procurement teams is perhaps the most actionable part of the picture: design for procurement, not just for performance.

“I understand it’s good to have the best MLC-based eMMC with industrial temperature – but this product basically doesn’t exist anymore. It will not be offered. So you have to take this into account, and maybe relax some of your CTQs, maybe see also alternative and additional technologies you can bring in to relieve the supply chain. Engineers have to discuss more with their sales and procurement departments – what’s available, what’s the right thing – because at the end of the day, you can have a perfect design but not be able to produce it anymore, which doesn’t make any sense.”

Florous also flags double bookings as a significant amplifier of market distortions. When customers inflate their forecasts out of fear of losing supply, suppliers respond by increasing prices further, creating a self-reinforcing cycle. His advice: stay disciplined and resist the urge to over-order as protection against scarcity.

He also reports seeing customers downgrade their product specifications in order to maintain production continuity – moving from DDR4 to DDR3, for instance, where availability is better. This pragmatic approach to specification flexibility is increasingly common among procurement teams trying to sustain output in a constrained environment.

The industries most at risk

Consumer electronics, industrial IoT, and automotive are the sectors Florous identifies as most exposed. The automotive case is particularly stark: long-term fixed-price bill-of-materials contracts that were agreed on multi-year terms are now being forced back to the negotiating table by suppliers who can no longer honour those prices in the current environment.

“The bill of materials in automotive contracts was spanning up to 10 years at fixed price. What happens now is suppliers are forcing the automotive OEMs to go again and sit at the table and renegotiate the price – which will explode the bill of material in unprecedented ways. These are the industries that are impacted most, and more likely to see their exits, insolvencies, mergers, acquisitions.”

The geopolitical dimension adds further uncertainty. Florous notes that volatility in consumer sentiment – driven by energy costs, conflict, and broader economic pressure – is already affecting device replacement cycles, with mobile handsets now being held for six or seven years rather than three. He estimates a roughly 50% reduction in hardware production volumes in 2026 compared to 2025 for certain consumer product categories.

Supplier strategy: know who you are dealing with

For smaller and mid-sized businesses, the supplier relationship itself has become a strategic variable. Florous is direct on the risks of competing with hyperscalers for the same product lines from the same major suppliers.

“If you are a small or medium business, it will be a suicide to deal with Samsung, who are focusing exclusively on hyperscalers – because what you are forcing yourself into is competing eventually with a hyperscaler from the same product. And you have no chances to compete, because of volume, because of bill of materials, because of the price these guys can pay because of AI investing.”

The message is straightforward: smaller businesses need suppliers whose business models are aligned with their scale and requirements, not suppliers whose primary attention is directed at the largest players in the world.

No short-term relief

Asked to summarise the outlook, Florous is unambiguous. Customers are in some cases paying 10x what they were paying three years ago for the same component – and that is not about to change. Meanwhile, the benefits of the current cycle are flowing in a specific direction.

“There is not going to be any short-term relief. The money flows to Korea, it flows to the US with NVIDIA. Nothing flows to Europe, I am afraid. Not yet. This is a bit disappointing, but it’s the reality.”

For engineers and procurement professionals navigating this environment, the practical priorities are clear: build flexibility into design specifications, manage supplier relationships strategically, resist the temptation to double book, and plan budgets on the assumption that elevated prices are structural rather than cyclical.

The memory market is not heading back to where it was.

This article originally appeared in the July’26 magazine issue of Electronic Specifier Design – see ES’s Magazine Archives for more featured publications.