#BernsteinSaysMemoryBullMarketToLastUntil2027 — But the Era of Explosive Price Gains Is Over

Wall Street investment bank Bernstein has issued a bold forecast that is reshaping how investors view the semiconductor sector: the current memory bull market is expected to persist through 2027. This represents a fundamental departure from the semiconductor industry's historical pattern of sharp booms followed by equally brutal busts.

---

The Numbers That Tell the Story

The data speaks for itself. In the second quarter of 2026, average DRAM prices surged 74% quarter-over-quarter. Server DRAM climbed 60% to 67%, while Mobile DRAM skyrocketed nearly 80%. As of June, DRAM chips widely used in PCs and servers had accumulated a 660% increase year-to-date.

Spot market data confirms the supply crunch remains intense. Server DDR5 spot prices are trading significantly above contract prices — clear evidence that AI and cloud service providers continue absorbing newly added capacity. In June alone, PC DRAM spot prices rose 5.6% to 11.5%, while Server DRAM climbed 6.1% to 26.4%.

The NAND segment tells a more complex story. Wafer spot prices softened, declining 3% to 4%. However, strong pricing in mobile NAND and SSDs is expected to drive overall second-quarter NAND contract prices up by approximately 60%. SSD and mobile storage prices could rise 70% to 80%, offsetting weakness at the wafer level.

A Structural Shift, Not a Temporary Cycle

According to Bernstein, the industry is not experiencing a temporary recovery but entering a period of stronger long-term support from structural AI demand — more durable than traditional consumer markets. This marks a major shift in how investors should evaluate memory manufacturers and semiconductor companies.

Unlike previous memory booms that were heavily dependent on smartphones, PCs, and consumer electronics, this cycle is increasingly driven by AI infrastructure. Every large language model, generative AI application, recommendation engine, and autonomous system requires massive amounts of data processed in real time — and this performance depends not only on powerful GPUs but also on advanced memory that feeds these processors at extremely high speeds.

Three Forces Reshaping the Industry

First, AI demand is exploding. High Bandwidth Memory (HBM) has become an indispensable component in AI accelerators powering global data centers. With cloud providers, hyperscalers, and enterprise customers continuing to expand AI infrastructure, HBM demand significantly exceeds current industry supply. UBS projects HBM demand will grow 90% year-over-year in 2026 and another 77% in 2027.

Second, supply-side discipline is unprecedented. Having endured previous downturns, manufacturers remain highly cautious about capacity expansion. Micron's CEO has stated that supply "will not be able to catch up with demand after calendar 2027".

Third, long-term agreements are locking in supply and demand. Microsoft, Google, Amazon, and other hyperscalers have secured supply through 2028 with minimum pricing clauses. Micron has already secured approximately $100 billion in non-cancelable, take-or-pay multi-year contracts — effectively insulating its business from the boom-bust cycles that have long plagued the DRAM industry. The company's HBM supply is fully sold out through 2028.

Who's Winning — and Who's Not

Bernstein maintains positive ratings on Samsung, SK Hynix, Micron, and SanDisk, while remaining cautious on Kioxia. HBM production capacity is concentrated among the first three — companies with excessive NAND exposure will be more vulnerable in this AI-driven DRAM cycle.

The financial results are staggering. SK Hynix currently dominates the HBM market with approximately 58% global market share, far ahead of Micron's 21%. Bernstein expects SK Hynix to achieve 91% gross margins in Q2 2026, with operating margins of 70% to 80%. Samsung, the world's largest memory chipmaker, reported operating profit up approximately 18-fold year-over-year to a record high in Q2. The company's stock is up 158% this year, while SK Hynix has gained 273% and Micron 242%.

The Transition: From Explosive Growth to Sustainable Expansion

While the bull market is expected to continue through 2027, Bernstein emphasizes that the phase of rapid price increases has already passed. Third-quarter 2026 DRAM price growth is expected to slow significantly to 13% to 18% — down sharply from the 74% surge in Q2.

Consumer electronics demand is showing signs of weakness. PC, smartphone, and consumer clients are reducing configurations or adjusting procurement rhythms. Although demand destruction has not yet fully materialized, Bernstein expects it will ultimately occur.

From the second half of 2027 through 2028, as long-term agreements take effect and new capacity comes online, memory prices are expected to gradually normalize.

A New Paradigm for Valuing Semiconductor Companies

Bernstein's outlook carries profound implications for equity investors and industry participants. Memory chip stocks could continue to outperform the broader tech sector, driven by earnings growth exceeding market expectations. Multi-year contracts and sold-out capacity provide visibility that was previously unimaginable in this historically cyclical industry.

The AI era may be fundamentally changing the boom-bust pattern. Demand is increasingly tied to long-term digital transformation rather than short-term consumer upgrade cycles. Companies across healthcare, finance, manufacturing, education, cybersecurity, and scientific research are integrating AI into their operations — creating a broader and more diversified customer base for advanced memory technologies.

---

The bottom line: Bernstein's forecast of a memory bull market lasting until 2027 represents a historic inflection point. What was once the semiconductor industry's most notoriously cyclical sector is transforming into a strategic infrastructure provider for the global AI economy. The explosive price gains are behind us — but a more sustainable, structurally healthier expansion lies ahead.

#MemoryBullMarket #AI #Semiconductors #Bernstein