How Micron Makes Money: DRAM, NAND, HBM and AI Explained

Micron sells something every AI system needs and almost nobody sees: the memory that keeps expensive processors busy. In fiscal Q3 2026, that business generated $41.46 billion in revenue. 

But rising AI demand explains only part of the story. Micron’s earnings also depend on which chips its factories make, what customers pay for them, and how much cash those factories consume.

Let’s break down how Micron makes money from DRAM, NAND, AI servers, smartphones, PCs and cars. Then we will follow the money through its business units, factory spending and customer contracts to understand what investors actually own.

What Does Micron Sell? DRAM, NAND and Memory Chips Explained

Micron is a semiconductor company specialising in memory and storage. It designs products, develops manufacturing technology and manufactures chips, using its own facilities and external partners for selected processes. It earns revenue primarily when control of products passes to customers.

Its customers include cloud companies, server and computer manufacturers, smartphone makers, automotive suppliers and distributors. Revenue comes mainly from physical products such as chips, memory modules and storage drives.

Three terms explain most of the product catalogue of Micron.

1. DRAM, or dynamic random access memory, holds the information a processor needs while working. It is fast, but loses its contents when power is switched off. DDR memory in computers and low-power memory in phones belong to this family.
2. NAND flash stores information even without power. It is used in smartphone storage and solid-state drives, or SSDs. An SSD combines NAND with a controller and software that manage the stored data.
3. NOR flash is commonly used to store code that helps devices start and operate reliably, including automotive and industrial equipment.

Think of a restaurant kitchen. The processor is the chef. DRAM is the preparation counter, keeping ingredients within reach. NAND is the pantry, holding supplies for later. A brilliant chef still works slowly if the counter is too small or ingredients arrive too late.

That is why faster processors can create demand for more capable memory.

Micron revenue by product technology

Source: Micron Q3 FY2026 earnings presentation. Shares calculated using unrounded company figures; displayed amounts are rounded.

The key accounting point is that high-bandwidth memory, or HBM, is a type of DRAM. Adding HBM revenue on top of total DRAM revenue would count the same business twice.

Likewise, an SSD is a finished product built around NAND. NAND revenue and SSD revenue are overlapping views of the business.

Micron Revenue by Business Segment: Cloud, Data Centers, Mobile and Automotive

Micron also reports revenue by customer market. These business units cut across the product categories above, so the two tables should never be added together.

The company reorganised its reporting in fiscal Q4 2025. Its current units are Cloud Memory, Core Data Center, Mobile and Client, and Automotive and Embedded. This is why older revenue breakdowns may show different segment names.

Source: Micron Q3 FY2026 earnings release. Shares and growth calculated from unrounded figures. A further $8 million came from All Other operations.

Cloud Memory makes money from hyperscalers and HBM

The Cloud Memory Business Unit serves large cloud customers and includes HBM for all data centre customers. It also supplies conventional server memory. Consequently, its $13.77 billion revenue should not be described as HBM revenue.

HBM stacks DRAM chips vertically and connects them through the silicon to move large amounts of data quickly. Micron’s HBM4 offers more than 2.8 terabytes per second of bandwidth per stack.

Bandwidth measures how quickly data moves. Capacity measures how much fits in memory. In the kitchen analogy, HBM improves how quickly ingredients reach the chef, rather than simply building a bigger pantry.

A useful detail: Micron’s 36GB HBM4 stack has the same capacity as its comparable HBM3E stack, but more than twice the bandwidth. Customers can therefore pay for better performance even when the number of gigabytes stays unchanged.

Core Data Center makes money from servers and enterprise storage

Core Data Center supplies memory to enterprise customers, equipment manufacturers and smaller cloud customers. It also contains storage products for all data centre customers, including SSDs.

This gives Micron exposure to more of the AI system. Data must be stored, retrieved and moved through ordinary server memory as well as HBM.

Its advanced data centre SSDs combine Micron NAND, controllers, firmware and manufacturing capabilities. This creates opportunities to earn money from a complete storage solution, with performance and reliability that a customer can test, rather than relying entirely on the underlying chip’s price.

Mobile and Client makes money from phones and PCs

This unit supplies memory and storage for smartphones and client computers. Revenue depends on device shipments, memory per device and pricing.

Those drivers can move in different directions. In Q3, Mobile and Client revenue rose 49% from the previous quarter even though bit shipments declined. Higher prices more than offset lower volumes.

For years, many PC users knew Micron through its Crucial brand. In December 2025, Micron announced its exit from Crucial consumer products, with shipments continuing through February 2026 and warranty support remaining available. It linked the decision to serving larger customers in faster-growing markets.

That does not mean Micron stopped supplying chips to phone and PC manufacturers. Its continuing Mobile and Client business is substantial.

Automotive and Embedded makes money from devices beyond computers

This unit serves automotive, industrial and consumer applications. Its products support systems such as vehicle electronics and industrial equipment.

Here, operating conditions matter. A memory component inside a car must handle temperatures and reliability requirements very different from those of a desktop computer. Micron offers specialised products and extended product lifecycles for these applications.

Our view is that investors should value automotive for its design requirements and growing memory content, while allowing for car production cycles and pricing pressure. The market remains a smaller contributor than data centres or phones and PCs.

Which Micron Business Segment Generates the Most Revenue and Profit?

Core Data Center had the fastest percentage growth in Q3. Cloud Memory added the most revenue in dollars.

Cloud Memory generated approximately $10.38 billion of additional revenue versus the year-earlier quarter, followed by $9.99 billion from Core Data Center. Together, these two units produced 61.0% of Micron’s revenue and 63.4% of its year-on-year revenue increase, calculated from the reported figures.

However, percentage growth, profit margin and total profit answer different questions.

Source: Micron Q3 FY2026 Form 10-Q. These are management-reported segment results, not segment net income or cash flow. Margins are company-rounded. Segment operating income totalled $33.68 billion, compared with consolidated GAAP operating income of $33.32 billion after unallocated expenses, predominantly stock-based compensation.

Cloud Memory led in operating profit because of its scale. Mobile and Client had the highest operating margin, and generated more operating income than Core Data Center.

That is a useful challenge to a simple AI narrative. Ordinary memory can become extremely profitable when supply tightens. Investors should not assume that the most advanced product always produces the highest margin in every quarter.

Micron does not disclose standalone HBM profit, or capital expenditure and cash flow by business unit. We can identify the largest reported profit contributor. We cannot honestly rank the units by disclosed free cash flow.

How DRAM and NAND Prices Drive Micron’s Revenue and Profits

Micron’s revenue has a simple starting point:

Revenue = quantity of memory shipped × average selling price.

The company discusses quantity through bit shipments. A bit is a basic unit of information. Tracking bits helps separate the amount of memory shipped from the number of packages shipped, since packages can have different capacities.

In Q3, DRAM prices increased in the low-60s percentage range from the previous quarter, while bit shipments increased in the low-single-digit range. NAND prices increased in the mid-80s percentage range, with bit shipments up in the mid-single-digit range.

Much of the revenue jump therefore came from higher prices for output, not a similar jump in factory output.

Higher prices can add revenue without requiring proportionately more manufacturing activity.

Consider an illustrative memory manufacturer with $100 of revenue, $60 of manufacturing costs and $20 of other operating expenses. Assume operating expenses stay fixed and manufacturing costs rise in proportion to volume.

Illustration only. These are assumed figures, not Micron’s reported results or forecasts.

The same revenue increase produces very different profits depending on its cause. Price-driven growth can be spectacular, but also vulnerable to reversal.

For investors, a revenue beat becomes more informative when management explains how much came from pricing, shipments and product mix.

The Hidden Cost of HBM Is the Memory It Replaces

HBM can improve Micron’s business through higher performance and stronger customer relationships. It also consumes scarce manufacturing capacity.

A wafer is the circular silicon base on which chips are manufactured. Micron has explained that HBM3E needs roughly three times as much silicon as DDR5 to produce the same number of bits. That is a generation-specific comparison, not a fixed rule for every future HBM product.

The opportunity cost is the conventional memory Micron could have produced with that silicon.

Here is a simplified example using equal quantities of memory as the unit. Assume the same wafer capacity makes three units of conventional memory or one unit of HBM.

Illustration only. Profit amounts are assumptions. Actual outcomes depend on chip design, manufacturing yields, packaging costs and negotiated prices.

HBM can earn more per unit yet less from the same wafer capacity. The premium must compensate for the memory output displaced and the extra complexity.

Our preferred question is therefore how much gross profit Micron earns from its scarce wafer capacity. Product margin alone does not answer that. The company does not disclose enough detail to calculate this precisely by product, but the framework helps investors interpret pricing and manufacturing updates.

There is also a second effect. Allocating more capacity to HBM can constrain conventional DRAM supply, supporting prices elsewhere. Micron’s disclosures identify higher HBM resource requirements and limited cleanroom space among the constraints on supply growth.

AI can therefore improve Micron’s earnings through both advanced memory and the broader supply balance. How long that benefit lasts depends partly on competing manufacturers’ investment.

How Long-Term Customer Contracts Could Change Micron’s Business

Memory manufacturers have historically faced a difficult mismatch: expensive factories last for years, while product prices can change much faster.

Micron is trying to reduce that mismatch through strategic customer agreements, or SCAs.

In its June 2026 earnings discussion, management said it had completed 16 agreements. Most generally run for five years, while automotive agreements generally run for three. They include binding volume commitments, and many contain fixed prices or price floors and ceilings.

A take-or-pay arrangement requires the customer to take agreed volumes or pay according to the contract. A price floor limits how far the agreed price can fall. A ceiling limits how far it can rise.

Management disclosed approximately $100 billion of minimum contracted revenue across 14 agreements over their remaining terms. This included agreements signed after the Q3 quarter-end. The RPO recorded at the quarter-end itself was over $5 billion. RPO means revenue from remaining contractual performance obligations, not cash already earned.

Our view is that the potential improvement in downturn profits matters more than the impressive contract total.

A factory supported by committed demand and minimum pricing can be easier to finance and plan. Customers gain supply assurance, while Micron gains protection against part of the industry’s volatility.

The trade-off deserves equal attention. Price ceilings may surrender some upside during shortages. Higher production costs can also reduce profits despite a contractual revenue floor.

These agreements could improve Micron’s earnings durability. They do not justify assuming that every future quarter will match today’s exceptional margins.

Micron Capex and Free Cash Flow: Where Does Its Cash Go?

None of Micron’s four business units reported an operating loss in Q3. Its largest visible cash investment was manufacturing capacity, while research and development was its largest operating expense category. Q3 R&D expense was $1.32 billion.

The cash-flow statement shows why those investments deserve their own analysis.

Source: Micron Q3 FY2026 earnings release and its reconciliation. Adjusted free cash flow is operating cash flow minus Micron’s net capital expenditure measure.

The resulting adjusted free cash flow was 44.2% of revenue, calculated from unrounded figures. The quarter funded substantial factory spending and still left a large surplus.

The cash requirement is still growing. In June, Micron forecast approximately $27 billion of net capital expenditure for FY2026. That was guidance, rather than a reported full-year result. Its expansion includes wafer manufacturing in the US and Taiwan, alongside advanced packaging in Singapore.

However, factories require continual reinvestment. Depreciation spreads an asset’s cost across its useful life in the profit statement. Capital spending measures cash used to acquire assets. Both matter, and they should not be treated as interchangeable.

New factories can also consume cash before producing meaningful output. Calling that spending a loss would be misleading, but assuming every new factory will earn an attractive return would be equally careless.

Three years show why one quarter is not enough

Sources: Micron fiscal 2023, 2024 and 2025 year-end earnings releases. Adjusted free cash flow follows the company’s reconciled definition for each year.

Across those three years, Micron generated approximately $78.03 billion of revenue but a cumulative $1.35 billion adjusted free cash outflow, calculated using unrounded figures.

The current earnings surge is a substantial improvement. The historical record also explains why investors should assess cash generation across a full memory cycle, including weak years.

Customer deposits are another source of funding

Management projected $22 billion of deposits and related financial commitments under agreements signed by the June results announcement, including approximately $18 billion of cash deposits. These were projected commitments, not a statement that all the cash had arrived.

Micron said the deposits would appear in financing cash flows and would not increase its free cash flow. It also said the cash would be returned to customers over time, towards the latter half of the agreements.

For investors, that distinction is valuable. Customer funding can help build factories, but it carries obligations to those customers. Revenue, operating cash generation and financing must remain separate in the analysis.

How Investors Can Analyse Micron Stock and Its Business Cycle

Micron competes with major memory manufacturers including Samsung, SK hynix and other DRAM and NAND suppliers. Scale, manufacturing efficiency and product execution matter because competitors can add supply or compete on price.

Our investment framework starts with three questions.

• First, what is driving profit growth? Higher shipments and successful product adoption tell a different story from a sharp industry-wide price increase. Both can create value, but the latter needs a more demanding downside test.
• Second, how much profit is protected? Contract coverage helps only when commitments, pricing terms and delivery economics support attractive returns. A minimum selling price should be compared with future manufacturing costs, not today’s costs alone.
• Third, how much cash remains after reinvestment? Strong reported earnings are more valuable when Micron can fund its technology and capacity needs without repeatedly consuming its cash reserves during weaker periods.

This also changes how investors should interpret the price-to-earnings ratio, or P/E. P/E compares the share price with annual earnings per share.

For illustration, a $100 stock earning $10 per share trades at 10 times earnings. If earnings fall to $5, the same share price represents 20 times earnings. The apparently inexpensive valuation depended on earnings staying elevated.

For Micron, we would estimate earnings and free cash flow under several pricing environments before deciding that a low P/E signals attractive value. We would also give genuine weight to evidence that its contracts are raising the profit floor.

We would give little weight to a valuation that simply multiplies an exceptional quarter by four. Equally, dismissing Micron as an unchanged commodity business overlooks the potential value of specialised products and binding customer commitments.

Author’s View on How Micron Makes Money

Micron’s business is becoming more valuable because advanced computing needs better memory, and customers increasingly care about dependable access to that memory.

The opportunity extends beyond HBM. Server memory, enterprise storage and conventional chips can all contribute. The latest segment results demonstrate why investors need to examine the whole company.

We find the business transformation persuasive. We would demand more evidence before treating its current profit margins as permanent.

The durable Micron thesis depends on earning more from scarce manufacturing capacity, protecting those earnings through customer commitments, and retaining enough cash after building the next generation of factories. Those are the results investors should track as the AI story develops.