Micron Closes Up 12% as Bank of America Report Boosts Memory Sector
Claire Weston
Micron Technology closed up more than 12% on Tuesday after Bank of America reiterated its Buy rating and $1,550 price target — the core thesis being that every download of a Chinese open-source model creates fresh demand for memory hardware, shifting AI storage's growth engine from closed-source to open-source.
Why did a single research note move the stock 12%?
BofA analyst Vivek Arya's team published a bullish report, reiterating a Buy rating with a $1,550 price target — the direct catalyst for the rally.
The thesis is not about near-term earnings. It is structural: a wave of Chinese open-source large models is generating batch-level new demand for memory hardware.
This means → the market priced the note as "long-term AI-storage logic reconfirmed," not just a routine rating reiteration.
How do open-source models become a growth engine for memory chips?
Closed-source models (like ChatGPT) are deployed centrally by the provider; users call them via API. Open-source models publish their weights — every download means the customer must build its own hardware stack to run it.
In plain terms = closed-source is "eating at a restaurant"; open-source is "taking the recipe home and cooking yourself" — you need your own pots and pans, meaning HBM (high-bandwidth memory), DRAM, and NAND.
Moonshot AI's Kimi K3 has 2.8 trillion parameters; a single inference instance needs roughly 1.4 TB of HBM. Even OpenAI's much smaller open-source model oss-120b — about one-twenty-third the size — still requires around 63 GB of model-weight memory.
Chinese models are far cheaper — won't that shrink hardware demand?
The market worry: ultra-low Chinese API prices signal collapsing hardware costs. BofA pushed back explicitly: API pricing reflects a business-model choice, not a hardware-cost decline.
The price gap has three sources: ① architectural efficiency gains; ② China's roughly 1.5–2× cost advantage in power, labor, and land; ③ subsidies from cloud providers like Alibaba, Tencent, and Baidu.
This means → even though Tencent Hunyuan's API costs just $0.06 per million tokens versus Anthropic Claude Opus 4.8 at $15 — a 250× gap — any customer deploying locally still needs every last HBM chip.
Could CXMT take a bite out of Micron's business?
BofA's verdict: no material threat in the near term. CXMT currently holds a low-single-digit to roughly 10% share of global DRAM wafer capacity, focused on consumer-grade and standard DRAM.
The critical gap is at the high end: CXMT has not entered HBM3E or HBM4 — precisely the product lines where Micron's margins are highest.
U.S. OEMs still face significant uncertainty over whether they can get government approval to buy CXMT products. This reflects geopolitical factors still providing a protective layer for Micron.
What does the buyback window opening mean?
Micron's CHIPS Act subsidies came with stock-buyback restrictions, expected to expire around December 2026.
Once lifted, free cash flow could reach $12–13 billion per year; at a 40% capital-return policy, that implies roughly $5–6 billion in annual buybacks.
In plain terms = that buyback run-rate equals 5–6% of Micron's current ~$1 trillion market cap — a sustained source of share-price support.
After a 12% pop, what comes next?
Tuesday's 12% single-day gain was the market's concentrated pricing of three reinforcing narratives: open-source models driving memory demand, limited CXMT threat, and the buyback window opening.
This means → part of the logic has already been "pre-spent." Whether the rally holds depends on whether AI memory demand shows up in actual earnings data.
Put simply = the research note gave the direction; the earnings report will give the answer.
Content is for reference only, not financial advice.