IDC: Global DRAM Revenue to Surge 177% in 2026

Memory Semiconductor Market Set for Historic Revenue Explosion

IDC projects global DRAM revenue will skyrocket 177% year-over-year to $418.6 billion in 2026, while NAND flash revenue is expected to climb 138.5% to $174.1 billion, according to a blog post published on April 29. The research firm sees total semiconductor industry revenue reaching a staggering $1.29 trillion in 2026, representing a 52.8% year-over-year increase — numbers that underscore the transformative impact of artificial intelligence on the chip industry.

The forecast positions memory semiconductors as the single largest growth driver in the broader chip market. IDC expects the entire memory sector to generate $594.7 billion in revenue by 2026, up an extraordinary 163% from the $226 billion projected for 2025.

Key Takeaways at a Glance

• DRAM revenue forecast to hit $418.6 billion in 2026, a 177% year-over-year surge
• NAND flash revenue expected to reach $174.1 billion, growing 138.5% YoY
• Total semiconductor revenue projected at $1.29 trillion in 2026 (+52.8%)
• Memory sector total to reach $594.7 billion in 2026, up 163% from 2025
• 2030 outlook: Global semiconductor industry revenue forecast at $1.75 trillion
• HBM supply: No substantial new high-bandwidth memory supply expected until late 2026

AI Transforms Memory From Cyclical Product to Strategic Asset

The semiconductor memory market has historically been one of the most volatile segments in the tech industry. Boom-and-bust cycles defined the DRAM and NAND markets for decades, with manufacturers regularly swinging between massive profits and painful losses.

That pattern is now fundamentally changing. IDC explicitly notes that memory semiconductors are transitioning from cyclical products to 'critical strategic assets' as AI workloads accelerate. This reclassification carries enormous implications for how investors, manufacturers, and enterprise buyers approach the memory market.

The shift is primarily driven by the insatiable memory demands of large language models, generative AI applications, and the data center infrastructure required to support them. Training a single frontier AI model can require terabytes of high-speed memory, and inference workloads at scale demand similarly massive memory footprints.

Unlike previous memory market upswings — which were often driven by smartphone upgrade cycles or PC refresh waves — this growth is structural rather than cyclical. AI workloads are not seasonal; they compound. Every new model, every new AI-powered service, and every enterprise AI deployment adds persistent demand for memory chips.

DRAM Dominates the Growth Story

DRAM accounts for the lion's share of the memory revenue explosion, with its projected $418.6 billion in 2026 revenue representing roughly 70% of total memory semiconductor income. This dominance reflects the critical role that DRAM plays in AI training and inference.

Modern AI accelerators from NVIDIA, AMD, and Google rely heavily on high-bandwidth memory (HBM), which is essentially a specialized, vertically stacked form of DRAM. HBM chips command significantly higher average selling prices than conventional DRAM modules, which helps explain the outsized revenue growth.

IDC's observation that no substantial new HBM supply will enter the market until late 2026 is particularly significant. This supply constraint suggests that:

• HBM pricing will remain elevated throughout most of 2026
• Memory manufacturers like Samsung, SK Hynix, and Micron will enjoy strong margins
• AI chip customers may face allocation challenges and extended lead times
• The supply-demand imbalance could push some AI deployments to explore alternative architectures

SK Hynix currently leads the HBM market, supplying the majority of HBM3E chips used in NVIDIA's H100 and H200 GPUs. Samsung and Micron are aggressively ramping their own HBM production, but IDC's forecast suggests meaningful supply relief remains months away.

NAND Flash Rides the AI Data Wave

While DRAM captures the headlines, NAND flash memory is experiencing its own AI-driven renaissance. The projected 138.5% revenue growth to $174.1 billion reflects surging demand for high-capacity, high-performance storage in AI data centers.

AI workloads generate and consume enormous volumes of data. Training datasets for frontier models now routinely span petabytes. Inference systems need fast local storage for model weights, caching layers, and user data. This creates sustained demand for enterprise-grade SSDs built on advanced NAND technology.

The NAND market's recovery is especially notable given its difficult 2023, when oversupply drove prices to historic lows and forced manufacturers to cut production. The AI boom has effectively rescued the NAND industry from what could have been a prolonged downturn.

Key NAND growth drivers include:

• Enterprise SSD demand for AI training clusters and inference servers
• Data lake storage expansion for AI training datasets
• Edge AI devices requiring high-capacity local storage
• Smartphone AI features driving higher storage tiers in premium devices
• Automotive AI applications requiring reliable flash storage

The Road to $1.75 Trillion by 2030

IDC's longer-term forecast paints an even more dramatic picture. The firm expects global semiconductor revenue to reach $1.75 trillion by 2030, nearly tripling from recent levels. This trajectory implies sustained double-digit compound annual growth rates across the industry.

The 2027 memory forecast is also revealing. IDC projects total memory revenue will reach $790.4 billion in 2027, representing a further 33% increase over 2026. While this growth rate is notably slower than the 163% jump expected in 2026, it still represents extraordinary expansion by historical standards.

This deceleration pattern — explosive initial growth followed by sustained but more moderate expansion — is consistent with what many analysts call the 'AI infrastructure buildout phase.' The initial wave involves massive capital expenditure as hyperscalers and enterprises race to build AI capabilities. Subsequent years see continued growth driven by expanding use cases, but at a more normalized pace.

Compared to previous technology cycles, the AI-driven semiconductor boom is remarkable in both its speed and magnitude. The smartphone revolution, for context, took roughly a decade to add $200 billion in annual semiconductor revenue. AI appears poised to add multiples of that in just a few years.

What This Means for the Industry

The implications of IDC's forecast extend well beyond memory chip manufacturers. The projected revenue explosion will reshape supply chains, capital allocation decisions, and competitive dynamics across the technology sector.

For memory manufacturers, the outlook justifies aggressive capital expenditure on new fabrication facilities and advanced packaging capacity. Samsung's planned investments in HBM and advanced DRAM, SK Hynix's expansion of its Icheon and Cheongju facilities, and Micron's $15 billion Idaho fab project all look increasingly well-timed.

For AI chip designers like NVIDIA and AMD, the memory supply constraint represents both a risk and an opportunity. Limited HBM availability could cap the number of AI accelerators they can ship, but it also reinforces the premium pricing power of their GPU platforms.

For enterprise buyers, the forecast suggests that memory costs will remain elevated through at least 2026. Organizations planning large-scale AI deployments should factor sustained high memory pricing into their total cost of ownership calculations.

For investors, the memory semiconductor sector's transformation from a cyclical commodity business to a strategic AI infrastructure play could warrant a fundamental re-rating of memory stock valuations. Traditional cyclical discount multiples may no longer be appropriate.

Looking Ahead: Supply Constraints and Market Dynamics

The most critical near-term variable is the HBM supply timeline. IDC's assessment that meaningful new supply won't arrive until late 2026 sets up a potentially challenging year for AI infrastructure deployment.

NVIDIA's next-generation Blackwell and Rubin GPU platforms will require even more HBM per chip than current-generation products. If memory supply fails to keep pace with accelerator production, the bottleneck could shift from GPU manufacturing to memory availability.

The memory industry's response to this demand surge will likely include accelerated transitions to more advanced process nodes, expanded adoption of advanced packaging technologies like hybrid bonding, and potential new entrants exploring alternative memory architectures.

One wildcard is the geopolitical dimension. With memory manufacturing concentrated primarily in South Korea (Samsung, SK Hynix) and the United States (Micron), trade tensions, export controls, and industrial policy decisions could significantly impact supply dynamics. China's domestic memory industry, led by CXMT and YMTC, faces ongoing technology restrictions that limit its ability to compete in advanced memory segments.

As the semiconductor industry hurtles toward IDC's projected $1.29 trillion revenue milestone in 2026, memory chips stand at the center of the AI revolution. The days of memory as a commodity afterthought are definitively over. In the age of AI, memory is power — and the market is pricing that reality in with unprecedented urgency.