AI companies are compelled to think about memory supply years ahead, not just when the next server arrives.
As models grow larger and workloads become more demanding, securing enough high-performance memory is becoming a bigger operational and financial constraint. Bank of America calls that rising burden AI’s “memory tax.”
For Micron Technologies (MU), that means stronger pricing, longer customer commitments, and a potentially more durable earnings cycle than investors have historically associated with memory chips.
In a note shared with me, Bank of America has responded by sharply increasing its forecasts after another earnings stunner, and the reason is far from being a single bottom-line beat.
BofA argues that AI is allowing memory to capture more value, while Micron gains better visibility through take-or-pay customer agreements.
Why Bank of America sees a more durable Micron earnings cycle
What changed for Bank of America is not simply that Micron beat expectations again, as it has for the past four consecutive quarters.
The bank now sees better visibility into how long elevated pricing and profitability can last, consequently raising its earnings forecasts in the process.
BofA bumped its fiscal 2027 revenue estimate to $275.4 billion from $230.3 billion and fiscal 2028 sales to $317 billion from $244.1 billion.
More Micron:
- Michael Burry increases his bet against popular chip giant
- Bank of America doubles down on Micron stock after AI bombshell
- Micron stock jumps as investors look beyond GPUs in AI chip trade
EPS estimates jumped to $171.78 from $140.24 for 2027 and $197.90 from $145.45 for 2028. It also reiterated its $1,550 price target, while calling Micron one of its top AI picks.
The conviction comes from demand becoming harder to dismiss as a temporary memory-cycle spike.
Micron now has 26 strategic customer agreements, up from 16 last quarter, covering an estimated 35% of revenue through 2030. More than 75% of fiscal 2027 output is already committed across those agreements and other customer commitments.
CEO Sanjay Mehrotra went further, saying Micron has “no line of sight” to when supply and demand return to balance. He expects memory conditions in 2027 and 2028 to be tighter than in 2026, as larger AI models, longer context windows, and greater concurrency consume more memory.
Investors initially hesitated after earnings, with Micron falling roughly 4% before reversing to finish about 3% higher.
That shows that even with Micron killing it from an earnings outlook perspective, investors still must decide how much of this AI-driven pricing power is truly structural rather than cyclical.
Micron’s cash flow may be the clearest sign this cycle is different
What impresses me about Micron’s quarter isn’t simply its astonishing $54.2 billion in revenue or the record 87% gross margin. It is what those numbers are now producing underneath the income statement.
Micron generated $44 billion in operating cash flow and $33.2 billion in free cash flow during fiscal Q4, while ending the quarter with $73.5 billion in cash and investments and a $68.3 billion net cash position.
CFO Mark Murphy said the balance sheet has “never been stronger,” even as Micron prepares to increase spending on technology and capacity.
I think the more important detail is where some of that visibility is coming from.
Micron received $12.3 billion in customer cash deposits during the quarter, while its 26 strategic customer agreements now represent roughly $150 billion in remaining performance obligations under agreements with defined pricing frameworks.
These are take-or-pay contracts, giving Micron something memory manufacturers historically lacked: clearer demand visibility before committing billions to new fabrication capacity.
That helps explain why Bank of America sees potential for capital returns to expand dramatically, estimating buybacks could eventually reach $60 billion to $100 billion in fiscal 2028-2029, although management has not yet clarified the near-term repurchase level.
I would not treat customer deposits as free cash, because they are eventually credited back against purchases. But combined with record earnings, committed output and long-term contracts, they suggest Micron is funding expansion from a considerably stronger position than in past memory cycles.
Micron’s valuation still leaves room for the earnings story to do the work
For me, the most attractive part of Micron’s setup is that its valuation hasn’t grown nearly as quickly as its earnings power.
The stock trades at roughly 6.2 times forward non-GAAP earnings, versus a sector median of 23.7 times, a discount of nearly 74% based on Seeking Alpha data.
Moreover, Wall Street is starting to respond.
I covered recently that Rosenblatt sharpy revamped its target to $1,900 from $1,500, while D.A. Davidson lifted its target to $2,100 from $2,000 and Mizuho raised its target to $1,400 from $1,300.
Nevertheless, I wouldn’t interpret the low multiple as automatically making Micron cheap. Memory stocks historically trade at compressed earnings multiples near strong parts of the cycle because investors worry profits will eventually normalize.
That is exactly why Micron’s customer commitments, tighter supply outlook, and growing AI exposure matter. If those factors make earnings more durable than previous cycles, today’s multiple may be pricing Micron like the old business, even as its economics are becoming materially different.
Related: 5-star analyst resets Micron stock target after $37.7 billion windfall