Three days ago MU closed at $848.95. It is now at $976.94, up $128.17 or 15.10% on the day. Bank of America added Micron to the firm's US 1 List, its highest-conviction investment ideas across all sectors. Analyst Vivek Arya raised his price target to $1,550 from $1,500, implying roughly 59% upside from the current price. Morgan Stanley simultaneously raised its target to $1,200 and projected DRAM prices rising another 25% next quarter. SKHY is up 6% and SKHX, the 2x leveraged ETF, is up 12%. The sector declared structurally broken last week is the sector institutions were aggressively building into today.
The timing is not coincidental. The US 1 List addition comes precisely after a 25% pullback from the June all-time high of $1,255, after hedge funds completed one of the largest momentum unwinds on record, and after the sell-side consensus EPS for fiscal 2027 held at approximately $154 throughout the entire correction without a single major downward revision. MU has now recovered $155 from its recent intraday low. Maximum sentiment pessimism against unchanged fundamental models is the environment in which institutional conviction calls are most credible.
The Open Source AI Argument That Inverts the Bear Case
The most analytically significant element of BofA's note is not the price target. It is the framework Arya used to address the Kimi K3 concern that triggered Friday's selloff. The standard bear argument is that cheap, open-source Chinese AI models reduce the amount of compute hyperscalers need to buy, which reduces demand for the expensive chips and memory inside those clusters.
BofA's rebuttal is specific and data-supported. Open-source models do not reduce HBM demand. They multiply it. Kimi K3 has 2.8 trillion parameters and requires approximately 1.4 terabytes of HBM per inference instance. Because open-source models are free to deploy, they proliferate across far more users and use cases than proprietary models that carry per-query pricing. The same AI capability, once commoditized through open-source release, generates more aggregate inference demand than it did when priced as a premium service. More inference instances running simultaneously means more HBM clusters in active operation.
Every DeepSeek moment, every Kimi K3, every open-weight model that matches frontier proprietary performance at lower cost, expands the total addressable market for AI inference infrastructure by making the capability accessible to users who previously could not afford it. The aggregate memory demand from a million small inference deployments exceeds the demand from a thousand enterprise deployments at equivalent capability. BofA forecasts the HBM market growing to $246 billion by 2030 and frames open-source proliferation as a structural demand driver, not a headwind.
The Fundamental Numbers the Selloff Did Not Change
Last quarter, Micron generated $41.46 billion in revenue, up 346% year over year. GAAP gross margin came in at 84.6%, above Microsoft's operating margin in the mid-40s. The Q4 guidance of approximately $50 billion in revenue with approximately 86% gross margins implies the trajectory is still accelerating, not plateauing. Micron has signed 16 multi-year strategic customer agreements covering approximately 20% of DRAM volume and one-third of NAND volume through 2030. Fourteen of those deals represent approximately $100 billion in revenue performance obligations at minimum prices. Even at the floor pricing in those agreements, Micron earns gross margins well above any prior peak cycle level.
Arya noted that Micron has beaten EPS estimates by an average of 24% for eight consecutive quarters. That is a systematic pattern where the business consistently outperforms the models analysts are using to forecast it. At $976.94, the fiscal 2027 consensus EPS of $154 implies a forward PE of approximately 6.3 times, still less than a third of the S&P 500's roughly 20 times multiple.
The CXMT Bear Case and Why BofA Dismissed It
BofA directly addressed the ChangXin Memory Technologies threat. CXMT is expanding DRAM capacity aggressively, raising $8.6 billion in a domestic IPO, and represents the most credible medium-term Chinese semiconductor challenge to US memory pricing.
Arya's response distinguishes between the types of memory CXMT can and cannot produce. CXMT currently manufactures LPDDR4 and DDR4, conventional DRAM for smartphones and PCs. It does not produce HBM3E or HBM4, the stacked high-bandwidth memory required for AI accelerators. Producing HBM requires a different process architecture, advanced through-silicon via packaging, and yield expertise CXMT has not demonstrated at commercial scale. US equipment export restrictions add further uncertainty about whether CXMT can access the advanced lithography tools needed to close the gap.
With Micron's revenue mix increasingly weighted toward HBM, conventional DRAM price declines from CXMT would have to be large enough to offset HBM's growing revenue share to materially compress blended margins. As HBM scales as a percentage of total revenue, that scenario becomes progressively harder to construct.
The Buyback Catalyst Nobody Is Pricing Yet
BofA highlighted a specific upcoming catalyst that receives little attention relative to the earnings discussion. Micron's CHIPS Act restrictions on share repurchases expire around December 9, 2026. UBS analyst Timothy Arcuri projects the company will generate more than $400 billion in cumulative free cash flow through 2028. If allocated toward buybacks once the restriction lifts, UBS calculates Micron could repurchase more than 40% of outstanding shares by end of 2028 at current prices. The math of 40% share count reduction is mechanical: EPS per remaining share grows proportionally without any revenue growth assumption. This catalyst begins less than five months from now.
SK Hynix Reports July 29: The Next Confirmation Event
Today's move is driven by analyst conviction, not yet by SK Hynix's own Q2 results. The actual Q2 earnings report is scheduled for July 29, eight days from now. In Q1, SK Hynix delivered a 49.71% EPS surprise and a 6.06% revenue surprise against consensus. Q1 revenue was 52.58 trillion won, up 198% year over year, with the company holding approximately 58% of global HBM market share by revenue according to Counterpoint Research. Management noted on the Q1 earnings call that "as this supply-demand imbalance persists, customers are prioritizing procurement over price," a statement that directly validates the pricing durability thesis BofA is building its $1,550 target on.
If SK Hynix's July 29 Q2 report confirms HBM pricing held, contracted volumes expanded, and 2027 demand visibility is intact, it provides the hard financial data behind the analyst calls that moved MU $128 today. The $155 recovery from recent intraday lows to $976.94 reflects institutional capital that moved before that confirmation. Whether it holds at these levels depends on SK Hynix on July 29, Alphabet earnings tomorrow, and whether the Hormuz situation provides any geopolitical relief. Those catalysts, arriving in sequence over the next nine days, will determine whether today's institutional call proves precisely timed or simply early.
Explore more exclusive insights at nextfin.ai.
