Don’t Forget About the Memory

For most of its history, DRAM (working memory) and NAND (storage memory) were the textbook commodity-cycle businesses in technology, meaning that key players in the industry manufactured a largely undifferentiated bit, competed primarily on cost-per-bit, and were punished for it. High fixed costs, low pricing power, and a boom-bust rhythm made memory a great technology met with not so great economics. However, AI has structurally broken that pattern and birthed one of the biggest winners of this revolution.

Because AI is no longer just an accelerator-rack story, demand has become more durable and broad-based than the single end-market spikes of past cycles. On the supply side, the dynamics are inverted – every new process node is more capital- and cleanroom-intensive while HBM’s (high-bandwidth memory) rising trade ratio actively cannibalizes commodity DRAM supply. Moreover, greenfield fabs – the only real lever for a step up in supply – are gated by multi-year permitting and construction timelines, skilled-labor shortages, and energy. All of this has de-commoditized pricing with rapid and extreme intensity. Of course, the skeptic’s counter is the temptation to over-build into a high-margin window, which would hurt both pricing and margins down the line.

Micron – one of the three key players in the industry, aside from Korean giants Samsung and SK Hynix – occupies a peculiar and rather useful position in the AI reporting calendar as its fiscal year ends in August, so its quarters land roughly a month ahead of the broader peer group. That timing, combined with the fact that memory now sits on the critical path to every AI system, makes Micron one of the first clean reads each quarter on whether AI infrastructure spending is holding, accelerating or rolling over.

In that sense, Micron’s earnings have graduated into the same reference class as a small number of supply-chain bellwethers whose prints move the entire market’s sentiment, such as Nvidia (demand barometer), TSMC (the leading-edge foundry read), and ASML (equipment and order-book read).

Micron’s fiscal Q3 results (USD billion, %)

Source: Micron Technology, InterCapital Research

The graph above is enough to highlight how important this once-commoditized industry has become in the AI landscape. Beyond the 74% sequential and 346% YoY growth in the top line, the real shock was the record 85% gross margin, a level that, for a business that was historically lucky to clear 40-50% at the top of a cycle, looks like a completely different industry. More importantly, the quarter was driven by pricing, not volume, as DRAM revenues increased on low-single-digit bit growth and low-60s percent ASP (average selling price) expansion, while NAND growth was driven by mid-single-digit bit growth and ASP up in the mid-80s percent range.

Set the top line of USD 41.6 billion and adjusted EPS of USD 25.1 against Street consensus of roughly USD 35.5 billion in revenue and USD 20.5 in EPS, and it is clear why Micron rallied last Thursday, jumping more than 15%.

Micron’s business unit financial results (USD billion, %)

Source: Micron Technology, InterCapital Research

Turning to the individual business units, the expansion runs across every end market. Data centers, comprising the Cloud Memory and Core Data Center business units, crossed USD 25 billion in revenue in the quarter, as agentic AI is structurally reshaping data center infrastructure, expanding beyond accelerator-only racks to include CPU racks for the agent control plane and program execution, and storage racks for rapidly expanding context store.

The Mobile and Client business unit is also expected to record further growth despite unit volume declines, reflecting resilient demand for high-end devices at higher prices across categories. On-device AI platforms enable improved tokenomics, greater privacy, and lower latency, and are expected to be the key driver of memory demand growth in PCs and smartphones in the future.

As for the Automotive and Embedded business unit, two key verticals are highlighted – autonomous vehicles and robotics. Depending on the level of autonomy, autonomous vehicles carry over five times the memory and storage content of an average vehicle, and that content should rise further as the mix shifts toward higher levels of autonomy. On the other hand, continued advances in simulation, foundation models, and physical AI in robotics create a growing, content-rich opportunity for high-bandwidth, low-power memory and storage that powers real-time perception, inference and control. Notably, humanoid robots carry 10 times the amount of memory as an average L2+ vehicle, which should set off another multi-decade memory demand cycle beginning in the latter part of this decade.

Micron’s fiscal Q4 guidance

wdt_ID wdt_created_by wdt_created_at wdt_last_edited_by wdt_last_edited_at GAAP outlook Non-GAAP outlook
1 Revenue USD 50bn +/- USD 1bn USD 50bn +/- USD 1bn
2 Gross margin ~86% ~86%
3 Diluted EPS USD 30.73 +/- USD 1.00 USD 31.00 +/- USD 1.00

Source: Micron Technology, InterCapital Research

Aside from another mind-blowing Q4 guidance, the actual headline for the outlook is business model transformation via Strategic Customer Agreements (SCAs). Micron announced 16 SCAs spanning data center, consumer, and automotive segments, representing roughly 20% of DRAM volume and about a third of NAND volume for the period. The full plan is expected to bring approximately half or more of company revenue under SCAs. Management is explicit that these are intended to fundamentally change how the business behaves through a cycle, not just for Micron but for the whole industry as others are likely to follow.

Multi-year take-or-pay agreements have historically been an oddity in the memory industry, and binding commitments to purchase specific volumes are a telling sign of just how scarce supply has become. The largest SCAs carry a ceiling at the current calendar-Q2 market price and a floor through the full term, while others are signed at fixed prices, and critically, the floor price supports gross margins well above peak quarterly margins in any past cycle. The SCAs also come with cash commitments that will begin to land on the balance sheet more meaningfully from next quarter.

Micron’s share price movement (USD, January 2025 – June 2026)

Source: Bloomberg, InterCapital Research

All in all, on the surface and beneath it, Micron’s results crushed once again, sending the durable, structural signal the market needed: demand is not fading but locking in commitments at record profitability for the supplier. From here, investors will be watching the moderation in the rate of price increases, HBM share and ramp economics, SCA penetration and structure, and CAPEX discipline. If you haven’t already read the full earnings report, I am leaving the link to Micron’s investor relations page so you can dig into one of the craziest reports out there right now.

Marin Orel
Published
Category : Blog

Want to invest? Do not know how and where? Contact us and we will solve everything for you.