One of Wall Street’s top semiconductor analysts just made one of the most aggressive price-target revisions in this AI cycle.
Bank of America analyst Vivek Arya nearly doubled its price target on Micron Technology Inc. (NASDAQ:MU), lifting it from $500 to $950 on Wednesday.
The move arrived alongside higher targets on four other AI hardware names and a third upward revision in eleven months to the firm’s 2030 forecast for artificial intelligence infrastructure spending.
Bank of America now sees the addressable market for AI data center systems reaching $1.7 trillion by 2030.
That figure was $1.4 trillion in February. It was $1.2 trillion in December. It was $823 billion last June.
In addition to Micron, Arya raised price targets on the firm’s top AI hardware names:
- Nvidia Corp. (NASDAQ:NVDA) was lifted from $300 to $320
- Marvell Technology Inc. (NASDAQ:MRVL) from $125 to $200
- Advanced Micro Devices Inc. (NASDAQ:AMD) from $450 to $500, and
- Coherent Corp. (NYSE:COHR) from $365 to $400.
| Company | Current Price | BofA Old PT | BofA New PT | BofA Implied Upside | Wall Street Consensus PT | Cons. Upside/Downside | BofA Rating |
|---|---|---|---|---|---|---|---|
| Micron Technology Inc. | $766.58 | $500 | $950 | +23.9% | $561.88 | −26.7% | BUY |
| Marvell Technology Inc. | $164.50 | $125 | $200 | +21.6% | $121.29 | −26.3% | BUY |
| Advanced Micro Devices Inc. | $448.29 | $450 | $500 | +11.5% | $419.77 | −6.4% | BUY |
| Coherent Corp. | $374.01 | $365 | $400 | +6.9% | $241.53 | −35.4% | NEUTRAL |
| Nvidia Corp. | $220.78 | $300 | $320 | +44.9% | $276.34 | +25.2% | BUY |
Why Micron Got The Biggest Move
“We view memory supply elasticity as now structurally lower,” Arya wrote, pointing to capital, packaging, power, and geopolitical constraints that have made it harder for memory makers to expand capacity quickly.
The implication is uncomfortable for anyone who has been waiting for memory pricing to mean-revert the way it always has.
It might not.
Micron’s fiscal 2026 capital expenditure guidance of more than $25 billion is roughly double what it spent in fiscal 2025. Most of that money is going into building cleanrooms rather than buying the equipment that fills them.
The new Idaho fab won’t deliver its first output until mid-2027. The Singapore advanced packaging facility for high bandwidth memory starts contributing in early 2027.
Meaningful new supply is a 2028 story.
Demand Does What Demand Does When Supply Is Fixed
BofA’s supply-demand sufficiency ratio stays below 110% through 2028 in both DRAM and NAND, a level that historically signals pricing power sits with the sellers.
The firm’s HBM forecast now reaches $168 billion by 2030, up from $35 billion in 2025. Memory content per AI accelerator climbs from 187 gigabytes today to 464 gigabytes by 2030.
Nvidia’s upcoming Vera Rubin systems, expected in the second half of this year, are already specified at 288GB of HBM4 per chip.
The Rubin generation doesn’t ship as a Micron call to action. It ships as a Micron purchase order.
Other 4 Names Get Upward Revisions
Marvell got the second-largest percentage move, jumping to $200 from $125. The thesis is optical.
BofA raised its ethernet transceiver forecast by roughly $7 billion in 2027 and $10 billion in 2028, with the 800G and 1.6T generations carrying the upside. Marvell holds an estimated 60% to 70% share of the digital signal processors used in those modules.
A $10 billion TAM increase translates to roughly $1.2 billion or more of incremental opportunity for the company.
- AMD was lifted to $500 from $450, with Arya citing CPU strength, the July analyst day, and potential for additional gigawatt-scale data center wins.
- Coherent went to $400 from $365 on its position in 800G and 1.6T transceivers, where the company holds 20% to 30% of the global market. BofA also flagged that Coherent will start shipping ultra-high-performance lasers for Nvidia’s scale-out co-packaged optics in the first half of 2027.
- Nvidia, despite the smallest percentage move at $320 from $300, remains BofA’s top sector pick.
Arya raised fiscal 2028 and 2029 sales and earnings estimates by roughly 7%. The catalyst calendar is heavy: the May 20 earnings print, the Computex tradeshow in early June with a possible new CPU launch, and the second-half Vera Rubin launch.
The Money Behind The Numbers
None of this works if the spending isn’t real.
“AI demand is scaling faster than traditional enterprise software adoption curves,” Arya wrote, citing two pieces of evidence Wall Street has been arguing about for months.
Anthropic’s annualized revenue run rate has reached roughly $30 billion as of April, up from $9 billion in December and $14 billion in February. OpenAI is reportedly projecting revenue of $30 billion in 2026, $62 billion in 2027, and $283 billion by 2030.
The hyperscaler commitments are larger still. Microsoft Corp. (NASDAQ:MSFT) disclosed roughly $627 billion of commercial remaining performance obligations, up 99% year-over-year, including an incremental $250 billion contracted from OpenAI.
Amazon.com Inc. (NASDAQ:AMZN) reported $364 billion of AWS backlog, excluding the recent $100 billion-plus Anthropic deal.
Alphabet Inc. (NASDAQ:GOOGL) said Google Cloud backlog nearly doubled quarter-over-quarter to more than $460 billion. Oracle Corp. (NYSE:ORCL) disclosed $553 billion of RPO in its fiscal Q3, up 325% year-over-year.
The aggregate cloud capex pool is now expected to reach 95% to 100% of operating cash flow in 2026 through 2028.
Historically, that ratio has run between 35% and 50%. The closest precedent BofA could find was the U.S. telecom industry during the peak 4G/5G buildout, when capex hit 65% to 70% of operating cash flow.
This AI buildout is even bigger.
Image: Shutterstock
Login to comment