Micron Slumps Despite Blowout Results as AI Capex Splurge Rattles Investors

TradingKey - On Wednesday, March 18, Eastern Time, Micron (MU) disclosed financial results and guidance that significantly beat expectations, yet its shares dropped as much as 6% in after-hours trading.

In the second quarter of fiscal 2026, Micron's quarterly sales nearly tripled to $23.9 billion. Earnings per share climbed to $12.20, far exceeding analyst expectations. Bloomberg data showed that analysts had previously expected average revenue of $19.7 billion and earnings per share of $9.

However, Micron also disclosed that capital expenditures for the current fiscal year will exceed $25 billion and forecasted that spending in fiscal 2027 will increase by more than $10 billion on top of that. Analysts believe the decline reflects market concerns over Micron's excessive capital spending.

No Choice But to Build: Micron's Cost of Chasing the Rivals

CEO Sanjay Mehrotra stated that construction-related costs within the fiscal 2027 capital expenditures will increase by over $10 billion.

Micron is building two new mega-fabs in Idaho and New York to boost its memory capacity in the U.S. The Idaho plant is expected to begin production in mid-2027, while the New York facility is slated to start wafer production in the second half of 2028.

It's not just Micron; its competitors Samsung and SK Hynix are also expanding capacity. Samsung currently has the Pyeongtaek P4 and P5 plants under construction, along with the Taylor plant in Texas. SK Hynix is building the Cheongju M15X plant and the Yongin semiconductor cluster, while also upgrading capacity at its M16 factory in Icheon.

Currently, Micron's HBM capacity is only one-third of Samsung's, and its capacity through the end of 2026 is already sold out. Without expanding capacity, it cannot increase order volume to gain market share. For Micron, which has the lowest market share, maintaining the status quo is not an option; in the capacity race, failing to expand is equivalent to handing future orders to competitors.

Micron’s Record Profits Fail to Mask Market Jitters Over $35 Billion Capex

Nevertheless, the higher-than-expected capital expenditure resulting from capacity expansion has a worrying side: will it weigh on Micron's gross margins and free cash flow in the future? Micron's capacity scale is significantly smaller than that of Samsung and SK Hynix, meaning its fixed asset depreciation costs per chip are higher, exerting a greater impact on gross margins.

Particularly as a cyclical industry, semiconductors will eventually face a decline in market demand. The wafer fabs Micron is currently building will become heavy fixed-asset depreciation on the balance sheet. Furthermore, compared to Samsung and SK Hynix building locally, the cost of Micron building plants in the U.S. is much higher.

A more serious issue is that across the industry, every historical crash in memory prices has been caused by overcapacity, and the production start dates for the new plants of Samsung, SK Hynix, and Micron are highly synchronized, nearly all concentrating around 2027.

Of course, a more optimistic view is that the memory demand currently driven by AI infrastructure needs will be larger in scale, longer in duration, and faster in iteration. This time, capacity expansion will not necessarily lead to a supply-demand imbalance, and the three giants are ramping up investments precisely because they are confident in this expectation.

The Nvidia Factor: Micron’s Bid to Break the SK Hynix Monopoly

Focusing on the more immediate future, the market is more concerned that Micron might fall behind in its competition with Samsung and SK Hynix. Micron recently announced that HBM4 entered mass production in the first quarter of 2026, dispelling rumors that it was excluded from NVIDIA's (NVDA) Vera Rubin supply chain. Last month, shares surged after CFO Mark Murphy stated the company had achieved high-volume mass production of HBM4.

However, the extent to which NVIDIA will rely on Micron's HBM4 supply remains a question. Currently, NVIDIA's HBM4 suppliers also include SK Hynix and Samsung. The former, due to its long-term partnership and deep integration with NVIDIA and TSMC (TSM) , has firmly established itself as NVIDIA's primary supplier and was the first to pass NVIDIA's qualification; rumors previously suggested it secured 70% of NVIDIA's HBM orders.

NVIDIA's procurement decisions for the Vera Rubin product line's memory will directly affect, or even determine, Micron's HBM market share; if NVIDIA shifts to its competitors, Micron will suffer a major blow.

From NVIDIA's perspective, however, the reason for including all three HBM giants in its supply chain is to offset SK Hynix's dominance and break its monopoly pricing, thereby gaining bargaining power. Industry rumors previously suggested that NVIDIA dispatched engineering teams to be stationed at Micron for real-time debugging to speed up product qualification. Given NVIDIA's urgent need for HBM chips, Micron's products offer both power consumption advantages and a check on competitors' premiums; NVIDIA is unlikely to abandon Micron for its rivals.

Year-to-date, Micron shares have risen a cumulative 62%, making it the top-performing constituent of the Philadelphia Semiconductor Index and the only company among the top ten largest U.S. tech firms by market cap to post a gain this year. Micron's current decline is also partly due to profit-taking by investors.

From a fundamental standpoint, Micron's Q2 gross margin reached 74.4%, indicating the company's strong earnings recovery capability. With HBM supply falling short of demand and orders backlogged through 2027, Micron's biggest issue isn't a lack of orders, but a lack of capacity. As AI infrastructure demand remains high, the tens of billions Micron is investing in capacity expansion today will only yield significant financial returns in the future.