Lenovo says the ‘RAMageddon’ is the new normal, outlines survival guide — at ISC 2026 an exec said ‘it will never be like it was last year’

At the International Supercomputing Conference (ISC) this week, Lenovo confirmed that the ongoing memory shortage, colloquially termed ‘RAMageddon,’ will become the new normal. The company presented a ‘RAMageddon Survival Guide,’ emphasizing that high demand from AI infrastructure and industry shifts mean memory prices and supply constraints are unlikely to ease soon. This announcement underscores the long-term challenges facing the tech industry amid persistent shortages.

Lenovo’s presentation at ISC highlighted that the recent low prices for DRAM and NAND in early 2025 are unlikely to return soon. The company explained that the economics of memory manufacturing have fundamentally changed, with demand driven by AI infrastructure absorbing much of the new capacity planned around 2028 and beyond. Lenovo referenced SK hynix’s announcement to triple its memory production capacity by 2034 as supporting evidence that manufacturers are less likely to invest heavily in expanding supply unless high margins are assured.

Furthermore, Lenovo pointed out that memory capacity is becoming a critical factor in server design. Upcoming dual-socket servers with 16 memory channels per processor could require around 1 TB of memory to utilize bandwidth fully, making memory costs a significant consideration. Industry forecasts from Micron and SK hynix also predict that supply constraints could persist until 2027 or even 2030, with long-term supply agreements worth approximately $100 billion indicating that hyperscalers are planning for sustained shortages.

Additional evidence of the industry’s outlook is Apple’s reported attempt to source DRAM from Chinese maker CXMT, a Pentagon-blacklisted company, illustrating how valuable additional supply has become as prices climb. Meanwhile, memory vendors enjoy strong pricing power, with profit margins at multi-year highs, reducing incentives for rapid supply increases. The shift toward high-margin HBM for AI accelerators has further tightened the supply of traditional DRAM, making system memory more expensive and encouraging GPU-based solutions that minimize DDR5 requirements.

Implications for the Tech Industry and Consumers

This development signals a sustained period of high memory prices and constrained supply, affecting server and PC markets worldwide. Organizations investing heavily in AI infrastructure will face ongoing cost increases, and hardware manufacturers will need to adapt their designs to manage higher memory costs. For consumers, this could mean continued price pressures on memory modules and related components, influencing overall PC and server pricing.

Lenovo’s emphasis on industry shifts and long-term demand underscores that the memory shortage is unlikely to resolve quickly, with supply-demand imbalances driven by AI growth and manufacturing capacity limits. This situation may reshape hardware procurement strategies and accelerate adoption of alternative memory architectures, such as GPU-attached HBM, to mitigate costs.

Background of the Memory Shortage and Industry Trends

The ongoing ‘RAMpocalypse’ began in early 2025, driven by supply chain disruptions, increased demand for AI infrastructure, and industry capacity constraints. Major memory manufacturers like Micron and SK hynix forecast supply shortages extending into the late 2020s and early 2030s, with long-term supply agreements reflecting these expectations. Historically, memory prices have been cyclical, but current industry dynamics—such as the shift toward high-margin enterprise products and AI acceleration—are creating a structural change in supply and demand patterns.

Lenovo’s comments at ISC align with broader industry forecasts, including Micron’s projection of constrained supply until 2027 and SK hynix’s warning of shortages until around 2030. These developments are compounded by the increasing importance of memory capacity in server design, with new architectures demanding higher memory per processor, further elevating costs and supply pressures.

“Memory prices were unusually low in early 2025, and it will be a long time before we see those levels again.”

— Lenovo representative at ISC

Unresolved Questions About Memory Market Recovery

It remains unclear whether the industry’s forecasts of persistent shortages will prove accurate, especially if new capacity expansions or technological innovations alter supply dynamics. The exact timeline for prices to stabilize or decrease is uncertain, as is the potential impact of geopolitical factors or unforeseen supply chain disruptions.

Additionally, the effectiveness of strategies like increased use of HBM or alternative architectures in mitigating costs is still being evaluated, and the pace of demand growth from AI sectors could accelerate or slow, influencing overall market conditions.

Upcoming Industry Developments and Market Responses

The industry will closely monitor capacity expansions, such as SK hynix’s planned growth, and their impact on supply. Hardware vendors are expected to adapt server designs to optimize memory usage, potentially shifting toward GPU-based architectures to reduce costs. Market analysts will track pricing trends and supply agreements to assess whether the ‘RAMageddon’ period will extend into the late 2020s or if new solutions will alleviate pressures sooner.

Further industry conferences and quarterly reports from memory manufacturers and OEMs will provide updated forecasts, offering clearer insights into the duration and severity of the current shortage.

Key Questions

Will memory prices ever return to pre-2025 levels?

It is uncertain. Industry forecasts suggest prices will remain elevated for several years due to sustained demand from AI infrastructure and capacity constraints, though cyclical patterns could eventually lead to price corrections.

How will the memory shortage impact server and PC costs?

Higher memory prices are likely to increase overall hardware costs, especially for servers requiring large memory capacities, potentially slowing deployment or increasing prices for end users.

What strategies are companies using to cope with the shortage?

Companies are adopting alternative architectures like GPU-attached HBM, optimizing memory usage, and securing long-term supply agreements to mitigate the impact of shortages.

Is the ‘RAMageddon’ situation expected to improve soon?

According to industry forecasts and Lenovo’s statements, the shortage is expected to persist into the late 2020s, with some improvement possible around 2028 or later, depending on capacity expansion and demand trends.

Source: Tom’s Hardware: For The Hardcore PC Enthusiast