Here's a number that should make you sit up: by 2027, memory chips are expected to gobble up more than two-thirds of what the world's biggest cloud companies spend on AI hardware. That's a massive shift. Yet Micron Technology Inc. (MU) trades 26% below the record it set in June. The gap between those two facts is where the market might be missing something.
According to market research firm TrendForce, dynamic random-access memory (DRAM) and NAND flash storage will together account for 68% of total capital expenditure at the nine largest cloud service providers in 2027. This year, that share is 47%. And the base isn't small.
TrendForce, in an Aug. 25 research note, put combined spending by those nine companies at:
- $272 billion in 2024
- $466 billion in 2025
- $922 billion this year
- $1.383 trillion in 2027
Apply the 68% share to that 2027 forecast, and you're looking at roughly $940 billion aimed at memory in a single year. That's not a typo. That's nearly a trillion dollars for the stuff that stores and feeds data to AI chips.
Why the Share Keeps Rising
The share is climbing because prices are climbing, not just because servers are multiplying. Server DRAM contract prices rose a cumulative 64% in the second half of 2025, and TrendForce expects roughly 270% more across 2026. Enterprise solid-state drive prices, the NAND side of the same trade, rose about 35% in the second half of 2025 and are projected to gain 235% this year. High-bandwidth memory, the stacked DRAM that sits beside an AI accelerator and feeds it data, could add another 70% to 140% in 2027.
This is where Micron, as well as peers like SK Hynix Inc. (SKHY) and Samsung, comes into play. SanDisk Corp. (SNDK) is another way to gain exposure to the storage side of the AI infrastructure buildout.
The Problem Hiding in the Same Forecast
But here's the twist: when memory absorbs 68% of a budget, it becomes the line item a buyer attacks. TrendForce says cloud providers could respond by cutting memory capacity per system, reworking RDIMM configurations, reducing the HBM built into future AI chips, or shifting toward custom AI chips with the model architecture hardwired into the silicon. In other words, the very price surge that's boosting memory makers could also trigger a demand response that cools things down.
Higher memory costs also hand suppliers such as Nvidia Corp. (NVDA) justification to raise their own prices, which forces cloud budgets higher again. It's a cycle that could feed on itself, at least for a while.
Where the Street Sits
So what do the analysts think? According to MarketDash Analyst Ratings, there's a consensus Buy on Micron from 27 analysts, with an average price target of $1,303. That sits roughly 40% above Tuesday's close of $932.97. Twenty-six of the 27 rate it Buy.
Recent flow is mixed rather than uniformly bullish. Mizuho and Citigroup both trimmed targets in August, while New Street Research upgraded the stock to Buy on Aug. 14. The Street high of $2,000, set by Cantor Fitzgerald and Barclays in late June, has not been matched since.
Micron has already more than tripled this year and crossed a $1 trillion market value. That's a lot of good news baked in, but the memory story might not be done yet.
| Date | Firm | Price Target | Action | Rating |
|---|
| Aug 25, 2026 | Mizuho | $1,375 → $1,300 | PT Cut | Outperform |
| Aug 14, 2026 | New Street Research | $1,250 | Upgrade | Neutral → Buy |
| Aug 7, 2026 | Citigroup | $1,400 → $1,150 | PT Cut | Buy |
| Jul 14, 2026 | Keybanc | $1,600 → $1,750 | PT Raise | Overweight |
| Jun 29, 2026 | Cantor Fitzgerald | $1,500 → $2,000 | PT Raise | Overweight |
| Jun 25, 2026 | Barclays | $1,175 → $2,000 | PT Raise | Overweight |
Source: MarketDash Analyst Ratings
