Micron plunges 22% in a week, is it a buying opportunity for memory chips? Bank of America bets on supercycle extending to 2028

According to Zhitong Finance APP, after a significant pullback in global AI computing power-themed stocks, especially popular semiconductor stocks related to memory chips, Wall Street financial giants like Bank of America, Nomura, and leading market research platform TrendForce have provided new fundamental growth evidence supporting buying on dips: the high level of revenue and profitability in the memory chip industry may be more enduring than what current stock prices reflect.

As of the week ending November 6, Micron had fallen about 22% in total, and although there was a subsequent rebound, it was hit again by unexpected inflation and rate hike expectations, creating a stark contrast between the valuation adjustment of memory stocks and the industry's prosperity. Bank of America's latest judgment provides a fundamental basis for buying on dips: even under a price increase assumption lower than TrendForce’s and factoring in a price pullback for some products in 2027, the global memory industry’s revenue is still expected to reach $1.47 trillion and $1.58 trillion in 2027 and 2028, respectively, continuously driving record-breaking revenue and profit levels.

Supporting the bullish outlook of top Wall Street investment houses such as Bank of America is the AI agent application landscape, dominated by the RSI AI training paradigm, which, by expanding high-efficiency AI inference processes, continues to boost demand for HBM, server memory, and enterprise-level SSDs. At the same time, advanced processes, production capacity transitions, and yield ramp-ups are limiting supply releases, while long-term supply agreements further increase sales visibility. The potential "buy-the-dip investment value" brought about by this round of significant corrections in AI computing power and semiconductor stocks ultimately depends on whether the pessimistic discount for "profit peak-outs" is much greater than the actual degree of supply-demand deterioration: if memory industry leaders can continually deliver cash flow and shareholder returns, the post-crash undervaluation could become the new starting point for repositioning in the next memory super cycle.

Bank of America expects global memory industry revenue to rise from $970 billion in 2026 to $1.47 trillion in 2027 and $1.58 trillion in 2028, representing growth rates of about 51.5% and 7.5% in the latter two years. Even under a significantly lower price increase assumption than TrendForce, Bank of America's analysts still believe the memory industry could maintain record revenues and profits through 2028.

Meanwhile, after Micron and SanDisk experienced heavy losses last week followed by rebounds on Monday, they were again impacted by inflation and rate hike expectations on Tuesday. However, it is worth noting that this decline was noticeably milder—the VIX fear index even dropped from 19.92 to 19.46, which largely suggests the market is re-evaluating the Federal Reserve’s rate path pressure and has not repeated last Thursday's panic sell-off. These signs point to the market reassessing whether profits created by tight supply can continue to offset valuation pressures brought by rising interest rates and adjustments to AI capital expenditure forecasts.

Bank of America firmly expects the memory supercycle to continue to 2028

In their latest bullish report on memory chip theme companies, Bank of America’s strategist team referenced TrendForce senior analyst Avril Wu’s opinion that the current upcycle in the memory chip industry could last until 2027 and potentially avoid a downturn in 2028.

Wu stated at Bank of America's online seminar on November 10 that memory chip prices may continue to rise, predicting that DRAM and high-bandwidth memory prices will increase by 10% to 20% quarter-over-quarter in Q4 2026, and NAND flash prices will rise by 8% to 13%.

She also expects HBM prices to more than double in 2027, with long-term supply agreements possibly helping to reduce price volatility after 2028.

Bank of America generally agrees with this optimistic industry outlook but uses more conservative price assumptions. The bank expects DRAM prices to rise by 8% quarter-over-quarter in Q4, NAND prices to rise by 3%, and anticipates price corrections may begin in early 2027.

The strategists also predict that the HBM price increase in 2027 will be less than 50%, in contrast to TrendForce’s prediction of prices more than doubling.

Bank of America’s strategists further state that their latest forecasts show global memory industry revenue hitting $970 billion in 2026, $1.47 trillion in 2027, and $1.58 trillion in 2028, indicating the industry could maintain record revenue and profitability levels through 2028.

With Meta Muse and OpenAI's GPT-6 Astra igniting a huge wave of AI agent adoption, "AI FOMO trading" (where investors buy in or cover previously reduced positions for fear of missing out on AI-related asset appreciation) is expected to continue. From the perspective of infrastructure, Muse and Astra are expected to further expand AI agent applications, and the resulting massive expansion of AI inference workloads will simultaneously increase the demand for model inference, tool execution, and state management, while also intensifying competition among some software companies.

From a technical perspective, AI agents extend a single response into repeated inference, retrieval, tool execution, and state saving. GPUs and TPUs handle model computation, CPUs are responsible for tool and task orchestration, HBM accommodates data needing high-bandwidth access, server DRAM supports runtime environments, and SSDs and HDDs cater to high-performance access and large-capacity retention, respectively. Meanwhile, the network infrastructure, high-speed optical interconnects in data centers, power supply, and cooling determine the practical utilization rate of costly accelerators. Thus, accelerating cloud revenue can simultaneously support multiple hardware segments; the specific profits earned by each company still depend on performance advantages, supply bottlenecks, and bargaining power.

For the "eye of the storm"—the memory chip sector, recently battered by AI trading and tech stock sell-offs—optimism on the fundamentals remains strong. In addition to Bank of America's optimistic reports on profitability and memory price trends, KeyBanc's recent research in Asia shows that AI data centers are still absorbing large amounts of storage and other AI computing power resources, driving price increases through to 2027.

After researching the Asian market, KeyBanc forecasts that HBM blended prices could increase by 80%–100% in Q1 2027. For DRAM memory prices, KeyBanc predicts a Q4 2026 quarter-over-quarter increase of 15%–20% on top of a strong Q3 price base, with another 10%–15% rise in Q1 2027. For NAND, they project a quarter-over-quarter price increase of about 20% in Q4 2026 (again, off a strong Q3 base), with another 10%–15% in Q1 2027.

Memory rebound faces another rate hike shock: stock prices respond first to inflation and the risk of continued Fed hikes, but profits and orders are still accelerating

On November 5, Eastern U.S. time, the Nasdaq 100 Index fell 6.36%, marking its biggest single-day drop since March 2020, with constituent stocks losing a combined $2.3 trillion in market value—a new single-day record; the Philadelphia Semiconductor Index plummeted 12.25%. After the Democratic Party won control of both houses of Congress, investors worried about stricter AI regulations, data center approval, and power cost-sharing requirements, which could delay computing power projects and chip procurement. These policy concerns, combined with already high valuations and crowded positions, have amplified the market’s reassessment of AI investment returns’ timing.

The memory sector suffered even deeper pullbacks. Between the close on October 30 and November 6, Micron and SanDisk fell by 22.01% and 27.44% respectively, while Samsung Electronics and SK Hynix slid about 16.13% and 20.17%, both underperforming the Nasdaq 100’s and Philadelphia Semiconductor Index’s declines of roughly 4.91% and 12.15% in the same period.

On November 9, global tech stocks generally rebounded, with Nasdaq 100 and Philadelphia Semiconductor Index rising by 3.40% and 6.85% respectively. Micron and SanDisk each gained about 6%. Funds rotated back to memory leaders, but this rally has not yet made up previous losses, showing that some Wall Street analysts' "cash flow safe haven" view is a judgment of operations and valuations, not a guarantee of short-term stock price resilience.

Tuesday's inflation data again raised the bar for any rebound. U.S. October overall and core CPI rose 0.4% and 0.3% month-over-month, both beating consensus estimates of 0.3% and 0.2% respectively, with core CPI up 0.3% for a third straight month. Waller said before the data’s release that if core CPI hits 0.3% or higher, a 50 basis point hike should be considered for December. After the data, the market priced in the probability of such a large hike at about 11.8%, but a 25-point move remained the mainstream expectation. By mid-2027, the probability of three 25bp hikes (including December this year) increased from two. Deutsche Bank also added a projected hike in January next year, in addition to those in December and March.

The logic behind continued Fed rate hike expectations is that employment and economic growth remain resilient, while service prices, energy cost transmission, and AI development jointly maintain inflationary pressure. For memory chip stocks, higher rates not only increase the discount rate on future profits but can also raise clients’ financing costs and lengthen data center construction cycles. On Tuesday, the Nasdaq 100 and Philadelphia Semiconductor Index fell about 0.60% and 0.35% respectively, with Micron and SanDisk down about 1.1% and 3.0%. This also shows that, though the market continues to absorb a rate repricing, it did not repeat last Thursday’s steep fall.

Can the memory supercycle weather the valuation winter? The key lies in how long high profit levels can be maintained

The consensus among Bank of America, TrendForce, KeyBanc, and others is that memory prosperity still has support, with divergences focused on the scale of price increases and timing of any pullback. TrendForce expects Q4 DRAM and HBM prices to rise by 10%–20% quarter-over-quarter and NAND by 8%–13%, while Bank of America's forecasts are only 8% for DRAM and 3% for NAND. For HBM in 2027, TrendForce expects prices to more than double, while Bank of America expects the increase to be under 50%.

Even a more conservative price path can sustain high revenue and profits, meaning Bank of America's bullish thesis relies on strong shipment growth, product upgrades, and order durability. This is also the most investment-relevant part of their “memory supercycle to 2028” view.

Technical support comes from AI systems’ simultaneous demand for different storage tiers: HBM provides high-bandwidth data for GPUs and ASICs, server DDR5 supports primary memory needs, and enterprise SSDs handle data retention, retrieval, and caching. Meanwhile, limited advanced process and wafer resources restrict competition between HBM and traditional DRAM. Ramping up yields and packaging output for new products takes time. Notably, TrendForce’s previous forecast of a 121% rise in blended HBM average prices by 2027 also factors in a higher proportion of costlier HBM4 and the introduction of HBM4e; it should not be interpreted as all products of the same specification doubling in price.

Long-term agreements are increasing demand visibility. SK Hynix recently disclosed it has signed long-term agreements with 14 major clients, usually five years in length, covering over half its capacity for 2027–2028. Prices stipulated in these agreements adjust with the market within a set range, so the main benefit is enhanced capacity utilization, sales, and earnings predictability, rather than fully locking in future profits. For a capital-intensive memory chip firm, this helps reduce mismatches between capacity expansion and customer procurement pace.

Bank of America analyst Vivek Arya’s team recently explained the current opportunity from a stock price cycle perspective: as of November 5, the Philadelphia Semiconductor Index has fallen about 38% from its June high, exceeding the roughly 32% average pullback during the past four major downcycles; however, the 136-day adjustment is only about 54% of the historical average of 253 days. The team expects the selloff could end in early 2027, as stock prices typically lead sales downside by 6–9 months. A forward P/E of 13.7, about 22% below the S&P 500, suggests much pessimism is already priced in. However, they also estimate that if AI capital spending in 2027 is 20% below forecasts, relevant chip companies' earnings per share could face 35%–40% downside risk.

The Nomura CW Chung team puts more emphasis on cash flow: cloud vendors make upfront infrastructure investments, while memory suppliers accumulate cash through tight supply and price hikes. Based on their share price and profit forecasts, Samsung and SK Hynix have projected 2027 P/Es of about 3.2x and 2.8x, respectively. Continued tightness until 2028, more long-term deals, and increased shareholder returns explain why they believe markets are over-discounting. Citi analyst Atif Malik also reiterated his "Buy" on Micron with a $1,300 target on October 29, stating that investors are underestimating the degree of DRAM and HBM supply tightness for 2027–2028, and that customer long-term deals add to revenue visibility.

Morgan Stanley analyst Shawn Kim’s latest adjustment reveals another divergence: in a November 9 report, he raised 2027 operating profit forecasts for Samsung and SK Hynix but cut target prices by 19% and 38% respectively, while maintaining an “Overweight” rating. His concern is that if new capacity comes online at the end of 2027 just as data center deployments are delayed, long-term profit margins could be depressed. This shows the institutional debate has shifted from whether companies can profit now to how many years high profits can last.

Overall, considering supply and demand, long-term deals, and changes in valuation, the recent massive correction in memory chip stock prices and valuations has created a more fundamentally sound buy-on-dips window for leaders with strong HBM competitiveness, high-performance server-level DRAM/NAND portfolios, and cash accumulation abilities. The investment value largely depends on how much free cash flow a company can retain under more conservative price and margin assumptions. If demand materializes, product upgrades continue, and shareholder returns verify profit resilience, today's discount on "profit peak-out" could be the source for the next round of revaluation.

Based on closing prices as of November 10, recent verifiable institutional target prices still assign substantial upside to memory leaders. U.S. memory chip giant Micron (MU.US) closed at $724.77, with Bank of America maintaining a “Buy” and a $1,550 target as of November 9—implying about 113.9% upside; Citi maintained its “Buy” and $1,300 target on October 29, implying about 79.4% upside.

Both institutions emphasized the DRAM and HBM supply constraints in 2027–2028, as well as how long-term supply contracts support sales and profit visibility. For SanDisk (SNDK.US), Morgan Stanley on October 30 maintained an “Overweight” rating and a $1,750 target price, representing about 103.5% upside from the November 10 close of $859.91, expressing optimism that growing data center demand will improve its business structure, with valuation measured using metrics across cycles.

The latest ratings for Korea’s memory giants reflect institutions’ “cautiously optimistic on near-term profits, reassessing long-term margins” stance. In a November 9 report, Morgan Stanley cut the target price for Samsung Electronics common stock from 381,000 KRW to 309,000 KRW and SK Hynix from 2.6 million KRW to 1.6 million KRW, maintaining “Overweight” on both; based on November 10 closes (198,900 KRW and 1,212,000 KRW), the implied upside is 55.4% and 32.0%, respectively. The bank also raised its 2027 operating profit forecasts for both, with the target price cut mainly due to reassessing new supply, data center project cycles, and long-term profit margins around 2028.