On September 22, 2026, TrendForce reported that global DRAM module market revenue reached $21.2 billion in 2025, up 59% year-over-year. That's a sharp acceleration from the previous year's $13.3 billion, which itself was only a 7% increase. But the growth didn't flow evenly to the companies that have long led the market. Kingston, the No. 1 player, grew just 48%, and its share fell from 66% to 62%. Why did the top player's slice shrink in a year when the market expanded by nearly 60%—and who ended up capturing the difference?
The Year $13.3 Billion Became $21.2 Billion
The DRAM module market added $7.9 billion in revenue in a single year. After a 28% decline in 2023 and a modest 7% recovery in 2024, this represents growth on an entirely different scale. At the exchange rate at the time of writing (157.41 yen to the dollar, September 22, 2026), $21.2 billion translates to roughly ¥3.34 trillion.
The companies whose numbers moved here aren't chip makers. Module manufacturers sit in the middle layer of the supply chain—buying dies and chips from chipmakers like Samsung, SK hynix, and Micron, then assembling them into DIMMs, SO-DIMMs, and RDIMMs for sale. Because they lack their own fabrication capacity, their revenue and profit hinge almost entirely on when and at what price they secured chips. Stockpile cheap and sell high, and profits swell; get the timing wrong, and rising procurement costs eat straight into margins.
In the first half of 2025, some module makers built up DDR4 chip inventory in Q2. There were two motivations: tariffs, and DRAM suppliers' planned end-of-life (EOL) schedules for DDR4. During the same period, supply also flowed to PC OEMs and consumer device customers who had front-loaded procurement in anticipation of tariffs.
In the second half, the demand landscape shifted. North American and Chinese cloud service providers (CSPs) ramped up server DRAM orders to meet computing demand from agentic AI applications. This surge squeezed the allocation DRAM suppliers had reserved for PC and consumer segments, pushing overall DRAM prices higher. With both shipment volumes and prices rising simultaneously, many module makers posted sharply higher revenue in the second half compared to the first.
In other words, 2025 rewarded companies that could unload cheaply stockpiled inventory during the high-price phase. Whether or not a company met that condition largely determined how much its ranking position grew.
Concentration Slipped Even as the Market Grew 59%
Kingston's 48% revenue growth in 2025 fell 11 percentage points short of the market's overall 59% growth. TrendForce attributes this to Kingston building up ample inventory throughout 2024 and 2025, backed by the industry's largest procurement scale, and actively supplying the channel market during the second-half price surge. Even so, it fell short of the market average.
As a result, in a year when the market grew 59%, the combined share of the top five companies fell from 81% to 77%, and Kingston's own share dropped from 66% to 62%—both declines of 4 percentage points. Since both share figures are published as rounded whole numbers, that 4-point gap includes some rounding error. Still, the direction is unmistakable: in a year the market ballooned, the top players' collective slice thinned.
One point worth clarifying: this doesn't mean Kingston shrank. A 48% revenue increase is a solid gain, and 62% still puts it far ahead of the No. 2 player. It's simply that the market's growth outpaced it.
So who drove the market average up? Lining up the year-over-year growth rates TrendForce published for nine companies: No. 3 Ramaxel posted 153%, the highest of the nine, followed by No. 7 Innodisk at 132%, No. 8 Apacer at 96%, No. 6 Patriot Memory at 89%, and No. 2 ADATA at 82%. Meanwhile, No. 4 Kimtigo grew only 20%, and No. 9 AGI just 26%. The nine companies split cleanly above and below the market average of 59%.
Comparing the explanations for the high-growth companies reveals two common threads. The first is inventory timing. ADATA's low-cost inventory built up during the previous cycle paid off during the upstream supply crunch, and partnerships with DRAM suppliers secured relatively stable chip allocations. Apacer, anticipating DDR4's EOL and worsening second-half supply-demand conditions, moved to secure additional supply starting in Q2.
The second common thread is customer mix. Innodisk focused on industrial markets, capturing both price gains and growth from edge AI projects. Patriot Memory's diversified presence across consumer, gaming, industrial, and edge computing segments paid off amid the supply shortage. By contrast, AGI—whose growth lagged—centers its own brand around the consumer retail market, relying on cross-border e-commerce as its primary channel for overseas expansion.
When chip allocation tightens, what determines revenue isn't how much a company wants to sell, but how much it can actually secure. The spread in 2025 growth rates reflects exactly that difference.
Only One Rank Actually Changed
Given how much share shifted, you might expect the rankings to have been reshuffled too—but the leaderboard delivers an anticlimax. The same eight companies that held the top eight spots in 2024 held them again in 2025, with not a single relative position swap. The only change was Ramaxel entering the rankings at No. 3, pushing the six companies from Kimtigo downward each dropped by exactly one rank. Kingston at No. 1 and ADATA at No. 2 remained unmoved even as their combined share shifted by 4 points.
Ramaxel, meanwhile, posted the highest growth of the nine companies at 153%. TrendForce attributes this to strategic partnerships with PC OEMs and CSPs, along with expanded RDIMM collaborations with domestic Chinese DRAM suppliers and CSPs—contributing to scaling up and improving quality within China's domestic RDIMM supply chain. The partners are described only as "domestic DRAM suppliers," with no specific company names identified.
Worth noting: this doesn't mean Ramaxel was absent from the market in 2024. TrendForce's 2024 release named only the top eight companies in its body text, and Ramaxel wasn't among them. But the 2025 release covers the top nine, so whether Ramaxel was ranked ninth or lower the previous year, or whether TrendForce simply expanded the number of companies or scope it tracks, can't be determined from the published materials. What can be said with certainty is this: a company that fell outside the top-eight frame shown in the 2024 release was placed at No. 3 in the 2025 release.
Since individual companies' revenue figures aren't disclosed except for Kingston's, it's impossible to see from the outside how large a margin preserved the rankings. But given that growth rates ranged from 153% down to 20% without shuffling the order, the scale gap between top-tier and mid-tier players must be large enough that a single year's growth-rate differential couldn't close it. Concentration loosened, but the basic structure of the competitive landscape held. If you had to draw 2025's change on a single chart, the only line redrawn would be at No. 3.
The One Company TrendForce Left Unlabeled
Reading through the descriptions of all nine companies in order, one stands out for a different reason: its write-up is missing something. TrendForce's release names year-over-year growth rates for eight of the nine companies in its body text—but for No. 5 Team Group, no growth rate is mentioned at all.
What is written about the company is limited to a description of its business activity: supplying DDR5 products to the North American channel market during the second-half DRAM price surge. The number that accompanies every other company—Kingston's 48%, ADATA's 82%, Ramaxel's 153%—simply isn't there.
How to read this gap requires caution. The secondary outlet Evertiq reported the company's DRAM revenue as down 5% year-over-year, but this figure doesn't appear in TrendForce's press release body text, and Evertiq itself doesn't cite independent reporting beyond TrendForce as its source. The release does include accompanying charts, and it's possible the figure appears there rather than in the text. The absence of a number in the body text is not proof that the number is unfavorable.
There's another figure that's easy to conflate with this one. Team Group's company-wide revenue, according to financial data site stockanalysis.com, reached NT$20.43 billion for full-year 2025, up 2.46% year-over-year, with full-year net profit hitting a record high. But this is a company-wide metric that includes products beyond DRAM, such as SSDs—a different scope than the DRAM module revenue TrendForce tracks. Overall company growth and how the DRAM module business specifically performed need to be treated as separate matters.
What can be confirmed from a reader's standpoint is only this: according to TrendForce's tally, the company's rank slipped from No. 4 in 2024 to No. 5 in 2025, and that single-slot drop was caused by Ramaxel's entry. The actual growth figure will have to wait for either the company's own disclosure or next year's tally.
Where Will the 13–18% Price Hike Actually Land?
What happened in 2025 is still rippling into 2026 pricing. TrendForce forecasts that server DRAM contract prices in Q3 2026 will rise 13–18% quarter-over-quarter. This is a forecast, however, not a confirmed price increase.
Several U.S. CSPs have multi-year long-term agreements (LTAs) with suppliers. Because price increases for LTA-covered customers are limited, TrendForce expects that from Q3 2026 onward, price hikes will primarily fall on customers without LTAs and on supply that falls outside existing LTA volume caps.
This divide—between customers large enough to secure long-term contracts and those who aren't—closely mirrors the chip-allocation gap seen in the module market. The prices Japanese consumers see for memory modules at retail stores or e-commerce sites will depend on which side of that line the product they're buying was sourced from. No. 9 AGI, which relies on cross-border e-commerce as its primary channel for overseas expansion, lists Japan as one of its sales regions. The $21.2 billion, roughly ¥3.34 trillion market expansion flows from upstream chip allocation, through mid-tier inventory strategy, and ultimately reaches Japanese buyers as retail price.
There's also an escape valve on the supply side. Samsung, SK hynix, and Micron had planned to phase out DDR4 (EOL) from late 2025 through early 2026, but surging demand and prices pushed that timeline back to 2026. As long as older production lines keep running, industrial equipment makers and PC makers left out of the latest-generation allocation still have somewhere to source chips. Reports on the exact timing vary by outlet, so this should be treated as fluid information.
What's worth checking in next year's tally isn't who holds the top spot. It's whether the combined share of the top five companies falls further below 77%, or whether it holds steady there—and whether another company breaks into the rankings from outside, as Ramaxel did this year. If mid-tier players can keep securing cheap inventory and stable chip allocation, buyers gain more room to diversify their sourcing even as prices rise, moving the market one step further away from a structure where a single leading company's inventory decisions dictate supply for the entire market.