Micron Joins the US$ 1 Trillion Club as UBS Triples Price Target to US$ 1,625 with HBM Sold Out until 2026
Memory joins the ranks of Nvidia, Apple, and Microsoft. HBM4 capacity is sold out, and CEO Sanjay Mehrotra states it meets between 50% and 65% of key customer demand.
Micron Technology crossed US$ 1 trillion in market capitalisation at the close of 26 May and extended the rally on 27 May, with an additional rise of over 8% in pre-market trading. The stock accumulated a 73% gain for the month and approximately 999% over the past 12 months, according to data from CompaniesMarketCap. It became the fifth company to surpass the US$ 1 trillion mark, alongside Nvidia, Apple, Microsoft, and Alphabet.
The immediate catalyst came from UBS. Analyst Timothy Arcuri tripled the price target from US$ 535 to US$ 1,625, a rise of 204%, marking a new high for the stock on Wall Street. The thesis is simple: the High-Bandwidth Memory market, the memory component attached to AI GPUs, has transitioned from a cyclical product sold in the spot market to a long-term supply contract with partially fixed pricing.
Arcuri revised the annual price increase hypothesis for HBM to 50%, from the previous 35%. The target price is derived from a multiple of 15 times the estimated earnings for 2029, discounted by one year. To achieve this, Arcuri projects that the entire memory sector will double, approaching US$ 2 trillion in combined capitalisation.
Sold-Out Capacity, Unmet Demand
In the latest earnings call, CEO Sanjay Mehrotra confirmed that all of Micron's HBM4 capacity for 2026 has already been allocated. The company is delivering between 50% and 65% of what key customers request for the medium term, according to the call. Micron is one of only three global manufacturers capable of producing HBM at scale, alongside SK Hynix and Samsung.
The direct effect is structural. Hyperscalers such as AWS, Microsoft, Google, and Meta are competing for allocation. Meta has already confirmed AI capex between US$ 115 billion and US$ 135 billion in 2026, nearly double that of 2025. Each US$ 1 billion in AI server capex carries, according to industry estimates, between US$ 200 million and US$ 300 million in high bandwidth memory. In contracted volume, this means a multi-year queue for any buyer without a contract in place by 2024 or 2025.
The second effect concerns server pricing. Dell, which reports on 28 May, has already guided a backlog of US$ 43 billion in AI servers for the beginning of FY27 and AI revenue of approximately US$ 50 billion for the year. Zscaler's CFO, Kevin Rubin, noted on 27 May the anticipation of data centre equipment purchases for the fourth fiscal quarter specifically to lock in prices before further increases in memory, storage, and processors. The reduction in Zscaler's free cash flow guidance from 26.5%-27% to 22.8%-23.3% is the first visible translation, in a software company, of the cost shock that Micron is capturing.
The Brazilian Perspective: Data Centre Capex and HBM Queue
The HBM shortage directly impacts the expansion schedules of Brazilian data centres. AWS, Microsoft, and Oracle have announced combined investments exceeding US$ 16 billion between 2024 and 2025 for new zones and expansions in São Paulo. Each rack equipped with Blackwell or MI300 requires HBM proportional to the sold capacity. Without guaranteed allocation, some of the timelines announced in Brasília will require silent revisions in 2026 and 2027.
On the local demand side, Itaú, Bradesco, and Nubank have cited increased technology budgets aimed at inference in their latest earnings calls. The cost of serving a large model within the country, however, depends on the same HBM that is being allocated to hyperscalers in the United States. For a CIO in São Paulo planning to run sovereign inference in 2026 and 2027, the picture that Micron presented this week indicates that the window to close contracts with tier-1 server suppliers with available HBM within 12 months is closing.
Consultancy CI&T, with clients in Brazilian banks, has already repositioned its offering for deploying models in April to include a preliminary benchmark of hardware capacity with the chosen supplier. This is the type of discussion that, two years ago, belonged in the infrastructure sector and now requires board-level consideration.
What Micron Needs to Deliver
UBS's US$ 1,625 thesis requires two confirmations over the next four quarters: that Micron maintains price discipline in multi-year contracts, avoiding a repeat of the deflationary cycle of 2022-2023; and that HBM4 engineering delivers competitive yields against SK Hynix, which has historically led the category. Failure in either of these will result in a lower multiple. Success, and Micron will begin to discuss, in 2027, the second trillion.