Micron DRAM Revenue Growth Beat Its Rivals, but SK Hynix Still Leads HBM

Micron DRAM revenue growth reached 65.5% quarter over quarter in Q2 2026, exceeding both Samsung and SK Hynix during a record memory-market surge. The result did not make Micron the largest supplier. It did, however, expose an unexpected weakness in SK Hynix’s HBM-heavy product mix.

SK Hynix remained the leading supplier of high-bandwidth memory, or HBM, which stacks memory dies to feed data rapidly into AI accelerators. Yet that leadership limited its exposure to sharply rising conventional DRAM prices. Annual HBM agreements adjusted more slowly than shorter conventional-memory contracts.

Micron took the opposite side of that pricing divide. Its emphasis on higher-priced server DRAM helped it capture more of the quarter’s increase, narrowing its revenue gap with SK Hynix. Samsung remained the overall leader and recorded the strongest bit-shipment growth among the three companies.

This is not evidence that HBM has become a bad business. SK Hynix reported record companywide results and retained half of the HBM market by revenue. The quarter instead shows how contract timing and product mix can temporarily reverse competitive momentum, even during an AI-driven boom.

Micron DRAM Revenue Growth Reached 65.5%

Micron grew fastest because conventional DRAM prices rose much faster than physical supply, rewarding the product mix it had available to sell.

Global DRAM revenue reached approximately $154.73 billion between April and June, according to DRAM market findings published by TrendForce. That represented a 59.5% sequential increase.

The growth did not come from a comparable rise in memory output. TrendForce said suppliers’ inventories remained at historically low levels, while total bit shipments increased only modestly. Bit shipments measure the amount of memory capacity sold, independent of the revenue received for it.

The difference matters. When revenue rises much faster than shipped capacity, pricing and product selection become central to the result. Suppliers with inventory available for products receiving the largest contract increases gain disproportionate revenue.

Micron occupied that position in Q2. Its DRAM revenue rose to $36 billion, up 65.5% from the previous quarter. Its market share increased to 23.3% under TrendForce’s methodology.

SK Hynix generated $38.59 billion in DRAM revenue. Its 37.9% quarterly growth remained substantial, but its market share declined to 24.9%. The gap between the companies narrowed to $2.59 billion.

Samsung remained well ahead of both competitors. It generated $60.98 billion in DRAM revenue, representing 39.4% of the market. Its revenue increased 63.4%, slightly below Micron’s rate but from a much larger base.

These figures describe calendar Q2 supplier revenue, not necessarily each company’s reported fiscal-quarter revenue. That distinction is particularly important for Micron, whose fiscal calendar does not align directly with calendar quarters.

Micron’s 65.5% result also requires careful framing. The company was not the fastest-growing DRAM supplier across every manufacturer. Smaller suppliers grew more quickly from far lower bases.

Nanya Technology’s revenue increased 68.3% to $2.612 billion, while Winbond reached $998 million after 75.8% growth. Powerchip Semiconductor Manufacturing Corporation, or PSMC, recorded a 167.8% increase to $115 million.

Micron instead led the three companies that dominate advanced DRAM supply. Those companies serve much of the server and AI infrastructure market, making their relative positions more consequential for major data-center buyers.

Demand arrived from several directions. AI accelerators required more HBM, while server CPUs needed high-capacity registered dual in-line memory modules, known as RDIMMs. These modules provide the conventional system memory used by enterprise servers.

Training large language models contributed to demand, but inference also mattered. Inference is the process of running a trained model to generate answers, code, images, or automated actions. More deployed AI services require more servers, storage, and system memory.

TrendForce also linked RDIMM demand to agentic AI applications. These systems coordinate tools and complete multistep work, increasing the need for CPU-based control systems alongside GPU clusters.

Micron emphasized higher-priced server DRAM while operating under capacity constraints. That mix placed more of its available output in a segment experiencing strong contract increases.

The result was a growth-rate victory, not a complete change in the rankings. Samsung remained first, SK Hynix remained second, and Micron remained third. However, Micron moved close enough to make second place a practical competitive question.

Why SK Hynix’s HBM Lead Limited Its Upside

SK Hynix’s slower DRAM revenue growth reflected HBM pricing mechanics, not weak demand or a collapse in its technology position.

HBM occupied a larger share of SK Hynix’s shipments than it did at Samsung or Micron. That exposure had supported SK Hynix during the earlier phase of the AI infrastructure buildout. In Q2, it reduced the company’s ability to capture conventional DRAM price increases.

HBM is generally sold through long-term agreements negotiated with accelerator companies and cloud customers. These agreements can provide volume certainty, supply visibility, and boundaries for future pricing.

Those advantages come with a tradeoff. A contract negotiated before a rapid market increase can prevent a supplier from immediately receiving the new market price. Conventional DRAM contracts often reset more frequently, giving those products faster exposure to changing conditions.

Counterpoint Research found that conventional DRAM prices rose sequentially during Q2, while average HBM prices declined year over year. Existing HBM3E pricing and delays in broader HBM4 adoption contributed to the difference.

Research Director MS Hwang said SK Hynix faced two related pressures. It received more of its shipments and revenue from HBM, while its earlier long-term agreements contained prices below later market levels.

That explanation makes SK Hynix’s HBM lead look like a quarterly liability. However, it would be misleading to treat the lead itself as a strategic failure.

The company still controlled 50% of HBM revenue in Q2, according to Counterpoint’s memory market tracker. Samsung held 33%, while Micron held 18%.

SK Hynix’s share had declined from 58% in the previous quarter. Samsung increased its share from 21%, while Micron slipped from 21%. These movements reinforce the conclusion that Micron’s outperformance came primarily from conventional server memory.

The distinction between price and volume is also essential. A company can maintain strong shipments, customer relationships, and technical capability while reporting slower revenue growth. That happens when contracted prices adjust more slowly than other parts of the market.

SK Hynix’s companywide figures hardly resemble a business suffering from weak demand. Its Q2 financial results listed revenue of 79.3187 trillion won, up 51% sequentially and 257% year over year.

The company also reported operating profit of 60.5426 trillion won and a 76% operating margin. SK Hynix characterized these figures as preliminary and subject to the independent auditing process.

Those companywide results include more than the DRAM revenue measured by TrendForce. They cover HBM, conventional DRAM, NAND flash, enterprise solid-state drives, and other products. This explains why they should not be compared directly with TrendForce’s supplier figures.

SK Hynix said it had finalized long-term agreements with around 10 customers. The company presented those contracts as a source of stability during a period when customer demand exceeds available supply.

That stability carries real value. AI companies need confidence that memory will arrive alongside expensive accelerators and networking equipment. An unexpected HBM shortage can delay an entire computing system, not simply one component.

The same contracts can suppress near-term revenue growth when outside prices rise quickly. SK Hynix accepted some pricing flexibility in exchange for longer demand visibility and stronger customer commitments.

Therefore, the main Micron vs SK Hynix tension is not conventional memory against an obsolete technology. It is short-cycle price capture against contracted HBM leadership.

That balance can reverse again. If conventional DRAM price growth moderates while HBM4 commands higher values, SK Hynix’s mix will regain an advantage. Q2 captured one point in that pricing cycle, not its final outcome.

Samsung Pressured Micron and SK Hynix From Both Sides

Samsung’s recovery matters because it combined conventional DRAM scale with an early HBM4 ramp, limiting the room available to both challengers.

A simple Micron vs SK Hynix comparison omits the company that generated the most revenue. Samsung retained 39.4% of the market under TrendForce’s calculations and delivered the strongest bit-shipment growth among the three leaders.

That shipment growth helped Samsung convert the rising average selling price into a 63.4% quarterly revenue increase. Micron grew slightly faster, but Samsung added substantially more absolute revenue.

Samsung also gained ground in HBM. Its revenue share rose from 21% in Q1 to 33% in Q2, while SK Hynix’s share fell by eight percentage points.

The shift followed Samsung’s early HBM4 production push. HBM4 doubles the memory interface from HBM3E’s 1,024 bits to 2,048 bits, increasing the amount of data that can move between memory and a processor.

Samsung announced that it had begun HBM4 mass production and commercial shipments on February 12. The company reported a consistent transfer speed of 11.7 gigabits per second per pin, with support reaching 13 gigabits per second.

Those performance figures remain company claims. Actual results depend on accelerator design, packaging, thermal behavior, system configuration, and customer qualification.

Still, commercial shipments gave Samsung a way to participate in two favorable segments. It could sell more conventional DRAM into a sharply rising market while building HBM4 revenue.

That combination pressured SK Hynix at the high end and reduced Micron’s opportunity to gain share at Samsung’s expense. Counterpoint described Samsung as benefiting from conventional DRAM demand, price increases, and a rising HBM position.

Micron was not absent from HBM4. Its fiscal Q3 update said HBM4 was shipping in volume for a lead customer’s platform. Micron also sent qualification samples to multiple additional customers.

Qualification is the customer testing required before a component receives approval for broader production use. It can include performance, reliability, thermal, and compatibility checks.

Micron said its HBM4 uses 1-beta DRAM technology. The company is developing HBM4E with its newer 1-gamma process and expects volume production during calendar 2027.

SK Hynix also began mass shipments of HBM4 during Q2. It said the product met customer operating-speed requirements while providing competitive power efficiency and manufacturing economics.

Each supplier therefore entered the second half with HBM4 activity. The meaningful differences concern qualified customers, delivered volume, yield, packaging availability, and realized revenue.

Those details are not fully visible in market-share percentages. A supplier can ship early units without reaching the scale needed to move its total results. Another can possess substantial contracted demand but recognize revenue under previously agreed terms.

Samsung’s position makes the second-place contest more difficult to interpret. Micron needs enough output to challenge SK Hynix, but it must also prevent Samsung from capturing incremental demand across both memory categories.

Meanwhile, SK Hynix must defend its HBM relationships without allowing fixed agreements to trail market conditions for too long. It also needs sufficient conventional DRAM exposure to benefit when CPU memory prices rise.

The competitive map has consequently become more fluid. Samsung operates from scale, SK Hynix owns the largest HBM position, and Micron has the strongest recent growth rate among the three.

No single metric settles that contest. Revenue share shows commercial scale, while HBM share indicates positioning around accelerators. Shipment growth, contract timing, product qualifications, and available capacity determine how those shares move.

What Micron’s Growth Rate Does Not Prove

One exceptional quarter does not prove that Micron has secured second place, because capacity and future contract resets will determine whether the gain persists.

TrendForce placed Micron 1.6 percentage points behind SK Hynix in Q2. Counterpoint used a different methodology and estimated shares of approximately 25% and 26%, respectively.

Differences between research firms can result from product classification, shipment timing, revenue allocation, estimates, and rounding. The direction is consistent across both sources, even though the exact shares differ.

Micron closed much of the gap. It did not complete the overtake.

Counterpoint Vice President Neil Shah said Micron was positioned to surpass SK Hynix and claim second place. However, he attached that possibility to available capacity and the structure of long-term agreements.

Capacity is the central constraint because suppliers cannot quickly add advanced memory output. New cleanrooms require extensive construction, equipment installation, process tuning, and customer qualification.

Process migration can increase the number of memory bits produced from each wafer. It does not create unlimited supply, and early production on a new node can face yield challenges.

Yield is the share of manufactured chips that meet the required specification. A more advanced process can eventually produce more usable chips per wafer, but weak initial yields can delay that benefit.

TrendForce expects the leading suppliers to increase output primarily through advanced-process migrations during 2026 and 2027. It expects only modest growth in wafer starts through optimization, acquired cleanrooms, and space reallocation.

Micron has taken a direct step toward expanding that footprint. It completed its purchase of PSMC’s P5 site in Tongluo, Taiwan, after announcing a total cash consideration of $1.8 billion.

The site adds approximately 300,000 square feet of existing cleanroom space. Micron expects meaningful product shipments from the facility beginning in its fiscal 2028, according to the Tongluo acquisition announcement.

That schedule limits the facility’s near-term effect. It supports Micron’s longer competitive position, but it cannot explain the Q2 increase or immediately solve every capacity constraint.

Nearer-term growth depends on existing fabs, process improvements, product allocation, and yields. Micron must decide how much capacity to direct toward server DRAM, HBM, mobile memory, and other products.

The company also faces a moving pricing environment. TrendForce expects conventional DRAM contract-price growth to moderate to between 13% and 18% sequentially in Q3.

That remains strong growth, but it is well below the Q2 industry revenue increase. PC and smartphone customers have less ability to absorb additional increases, while some server demand is shifting toward lower-capacity RDIMMs.

Micron’s mix produced exceptional leverage during a quarter of sharp conventional-memory increases. A slower pricing environment reduces that advantage unless shipments or product values rise enough to compensate.

SK Hynix’s HBM exposure can move in the opposite direction. HBM4 adoption gives suppliers an opportunity to negotiate new pricing around a more complex product.

TrendForce has said annual HBM pricing prevents contract values from reflecting quarterly changes immediately. It also expects HBM wafer input among the three leading suppliers to reach 22% of DRAM wafer input by late 2026 and 30% by late 2027.

A larger HBM4 mix would consume production resources that might otherwise serve conventional DRAM. That can support broader memory pricing, but it raises the importance of manufacturing yields and packaging capacity.

Counterpoint’s figures introduce another warning against a simple victory narrative. Micron’s HBM share declined to 18% in Q2 while Samsung gained sharply.

Micron therefore needs to defend two positions at once. It must retain its conventional server-memory momentum while securing enough HBM4 qualifications and shipments to remain relevant around next-generation accelerators.

Smaller suppliers also complicate the market. Nanya, Winbond, PSMC, and CXMT are serving demand left behind as the largest manufacturers prioritize advanced server products.

Counterpoint estimated that CXMT’s DRAM revenue increased 716% year over year. Its expansion could place additional pressure on mature and mainstream memory segments, particularly if its production capacity and customer reach continue growing.

Micron DRAM growth explained through product mix is convincing for Q2. Extending that explanation into a permanent ranking change would go beyond the available evidence.

Three Signals Will Decide the Next DRAM Ranking

The next phase depends on Q3 conventional-memory pricing, HBM4 revenue conversion, and each supplier’s ability to create usable capacity.

The first signal is Micron’s Q3 calendar-quarter DRAM revenue and market share. The company must show that its Q2 growth was more than a one-quarter response to unusually favorable contract resets.

A sustained share increase would strengthen the case that Micron’s server-memory allocation is producing a structural gain. A retreat toward its earlier position would suggest that pricing timing created most of the outperformance.

The important measure is not revenue growth alone. Analysts should compare revenue, bit shipments, average selling prices, and product mix. Higher revenue generated mainly by another price increase would say less about manufacturing progress.

The second signal is HBM4’s contribution at all three companies. Announced production or sample shipments matter less than qualified volume and recognized revenue.

SK Hynix must show that its HBM lead translates into better pricing as next-generation contracts take effect. A rebound would weaken the claim that HBM leadership remains a liability beyond Q2.

Samsung must prove that its early commercial shipments can support its rapid share gain. Its position becomes more defensible if HBM4 volume expands without sacrificing conventional DRAM output.

Micron needs additional qualifications beyond its lead customer. More qualified platforms would reduce customer concentration and support its claim to compete across the AI accelerator market.

The third signal is usable capacity. This includes mature yields on advanced DRAM nodes, available packaging, cleanroom conversions, and the timing of new facilities.

SK Hynix plans to expand HBM4 production during the second half of 2026. It is also accelerating the M15X schedule and preparing for the opening of its first Yongin cleanroom in early 2027.

Micron’s Tongluo acquisition adds a longer-term expansion path, but meaningful shipments are not expected until fiscal 2028. Its immediate challenge is extracting more output and value from existing capacity.

Samsung’s scale gives it another route. It can allocate output across conventional DRAM and HBM while using its internal foundry and packaging operations. That breadth also creates execution complexity across several advanced processes.

Buyers should watch these signals because memory supply affects entire AI systems. HBM availability can limit accelerator deliveries, while server DRAM shortages can delay the CPU infrastructure supporting data preparation, inference, and agent control.

Higher memory costs also influence which AI deployments remain economical. Large cloud providers may secure long-term supply, while smaller operators face tighter allocation and less negotiating leverage.

For developers and enterprise customers, the competitive ranking matters less than the resulting availability. More balanced competition can broaden qualified supply and reduce dependence on one HBM provider.

However, every supplier is directing more resources toward high-value infrastructure products. Consumer devices and conventional memory buyers can still face higher costs when that shift removes capacity from other segments.

Micron posted the fastest Q2 growth among the Big Three, while SK Hynix retained its HBM lead and Samsung remained the largest DRAM supplier. All three statements are simultaneously true.

The next ranking will reveal which advantage matters most after Q2’s pricing surge fades. Watch whether Micron holds its share, whether SK Hynix converts HBM4 leadership into faster growth, and whether Samsung sustains gains across both markets.

Those results will answer the central question behind Micron DRAM revenue growth: did product mix create a temporary opening, or has Micron built a credible path to second place?