Samsung's HBM4 Yield Hits 80% - Now the Real Fight Is Production Volume

Reports confirm Samsung's HBM4 yield has climbed to roughly 80 percent, but the bigger number is capacity: HBM wafer input is projected to jump nearly 40 percent next year, from about 180,000 to 250,000 wafers a month.

Why an 80 percent yield actually matters

HBM (high-bandwidth memory) is built by stacking and connecting multiple layers of DRAM. A defect in a single layer can ruin the entire finished chip. That makes yield a far more direct driver of cost and supply than it is for ordinary DRAM.

According to industry sources cited by Seoul Economic Daily and TrendForce, Samsung's HBM4 yield has climbed from under 60 percent in February, when mass production began, to roughly 80 percent recently. The figure is an industry estimate, not an official disclosure from Samsung. There are also reports that SK Hynix's HBM4 yield has reached the 80 percent range.

An 80 percent yield does not mean "Samsung has won on technology." It means more chips can be sold from the same equipment and defect costs are falling, which is closer to the conditions needed for HBM4 to turn into real profit.

The number that matters more than 80 percent

  • HBM4 yield: about 80 percent. Based on industry sources, not an official disclosure.
  • Next year's monthly HBM wafer input: about 250,000 wafers. Up roughly 40 percent from about 180,000 wafers this year.
  • Outsourced glass-carrier cleaning: 2.5x. From 20,000 wafers a month to 50,000. This does not map one-to-one onto finished-product volume.

According to a September 20 report, Samsung's HBM production capacity, measured by average monthly wafer input, is expected to rise from about 180,000 wafers this year to about 250,000 next year. Outsourced cleaning volume for the glass carriers used in HBM4 and HBM4E is also reported to be rising from 20,000 to 50,000 wafers a month.

Glass-carrier numbers alone cannot be used to calculate finished-chip shipments, since reuse counts and process conditions vary. Still, the direction is clear. Samsung's HBM strategy is shifting from "can it pass customer qualification" to "how much can it supply, and how reliably."

Market share has already started to move

A report citing Counterpoint Research shows Samsung's share of HBM revenue rising from 21 percent in the first quarter of 2026 to 33 percent in the second quarter, while SK Hynix's share fell from 58 percent to 50 percent over the same period. That narrows the gap between them from 37 percentage points to 17 points.

UBS projects that, measured in shipped bits, SK Hynix will keep the number one spot this year at 48 percent, but that next year Samsung could reach 41 percent versus SK Hynix's 39 percent. These are forecasts. Whether the lead actually flips depends on product-by-product qualification and supply contracts with key customers such as Nvidia, AMD and Google.

QuestionWhat favors SamsungWhat still needs confirming
ProductivityHBM4 yield has improved rapidly to about 80%This is an industry estimate, not an official yield disclosure
CapacityMonthly wafer input projected to rise about 40% by 2027Actual equipment installs, ramp speed, product mix
Market shareRevenue share reportedly rose to 33% in Q2Sustained volume wins with major HBM4 customers
ProfitabilityRising mix of higher-margin HBM4How much yield gains translate into operating margin

Bernstein's data shows speed, not a confirmed reversal

An August Bernstein report used July's Korean multichip memory export data to estimate HBM revenue at Samsung and SK Hynix. The regression model put Samsung's third-quarter HBM revenue at about $12 billion and SK Hynix's at about $5.6 billion, implying roughly an 80 percent quarter-over-quarter increase for Samsung and a roughly 20 percent decline for SK Hynix.

This should not be read as confirmed results. The model uses regional export data as a proxy for company-level HBM revenue, and SK Hynix's shipments could recover later in the quarter. What matters is that multiple data sources are now pointing to the same thing: Samsung's HBM4 ramp is moving faster than the market expected.

Is this bad news for SK Hynix?

Looking at market share alone, competitive pressure is clearly increasing. But if the overall HBM market is growing fast and supply stays tight, Samsung ramping up production does not automatically mean lower revenue for SK Hynix. TrendForce expects HBM supply to keep falling short of demand growth even in 2027, which would let suppliers hold onto pricing power.

So the number SK Hynix investors should watch is not Samsung's yield in isolation, but whether SK Hynix keeps its share of high-spec product and contract pricing with core customers, including Nvidia. Technology leadership in this market is not a single yield figure. It has to be read alongside customer adoption, supply stability, packaging quality and price.

The bigger picture in Korea's export data

Provisional figures from Korea's Customs Service show semiconductor exports of $34.12 billion between September 1 and 20, up 259.4 percent from a year earlier and accounting for 47.8 percent of total exports. Total exports over the same period rose 78.3 percent to $71.4 billion.

These figures are not HBM-specific. They mix in commodity DRAM pricing, server demand, shipment timing and base effects. Still, they show that the HBM race between Samsung and SK Hynix is not purely zero-sum right now, because the broader memory market itself is running very hot.

Samsung's HBM4 yield gains are starting to show up as real market share against SK Hynix, and the next test is who can supply the most volume to Nvidia, AMD and Google.

Insight Times Editorial Desk

More by topic

Starting with this story's topics, then what readers are following now

Investor psychology, decision-making and books on investing

HBM

This story's topic77 storiesSee all →

Memory stacked next to the GPU.

The Real ASML Number Isn't the 3% Pop, It's the 30% Capacity Hike

ASML's rally isn't just an AI trade. Customers are already locking in 2027 tool slots, and ASML is raising Low-NA EUV capacity by roughly 30% to keep up.

Investing

A 3x P/E is not a discount. It is a verdict on how long memory profits last

Nomura's buy case on Samsung Electronics and SK Hynix does not stop at higher memory prices. It argues that if AI has turned memory from a cyclical component into strategic infrastructure, the formula the market uses to value both companies has to change too.

Investing

2027 Isn't the End: Citi's Case for a Memory Supercycle Running to 2031

Citi says the memory shortage timeline is stretching again. If AI shifts from "train once" models to continual learning, HBM, server DRAM and enterprise SSDs could all fall short at the same time.

Tech

AI ran out of memory before it ran out of GPUs

The HBM bottleneck is spreading into DRAM prices, server costs and the price of the next laptop or phone.

Tech

Has Memory Peaked? Prices Are Cooling, but the Shortage Looks Set to Run to 2028

Price growth has slowed and consumer demand is already breaking. Server and HBM supply remains short, and Micron says more than 75% of its 2027 volume is already committed to customers.

Tech

AI Chip Peak Check: Demand Still Rising, Risk Building in 2027-2029 Supply and Funding

DRAM and NAND contract prices are still climbing and HBM suppliers hold the edge. Power shortages, AI financing costs and simultaneous memory expansion are the risks for later years.

Investing

AI Chips Haven't Peaked. The Real Risk Sits in 2028 Supply

HBM and server DRAM shortages are now spreading into enterprise SSDs. The question isn't today's strength, it's how much new capacity these prices are pulling in and when it lands.

Investing

NVDA

This story's topic96 storiesSee all →

Who Hits $1 Trillion Next: Reading AMD and ASML by Different Rules

One company is chasing Nvidia's GPU empire. The other sells the machine that draws the chips for every advanced chipmaker, including Nvidia. Both get floated as 2028 trillion-dollar candidates, but the numbers investors need to watch are entirely different.

Investing

Nvidia and Palantir Put AI Inside the Supply Chain - The Real Shift Is That It's Learning Judgment

A single Vera Rubin rack holds 1.3 million parts. The real breakthrough in Nvidia and Palantir's new supply chain AI isn't automating that complexity, it's capturing the judgment calls that used to live only in planners' heads.

Tech

Musk Sees the Next Phase of AI Infrastructure: Who Can Switch GPUs On, Fast

If neoclouds struggle under financial pressure, AI demand may not shrink. It may simply favor the few operators with the capital and execution to serve it, a shift that could matter for SpaceX and Tesla.

Tech

China's AI Labs Went Open Source Because They Had to, Not Because They Wanted To

A GPU shortfall pushed Chinese AI labs toward efficiency and open weights, and that constraint is now reshaping the global competition over who controls AI's basic infrastructure.

Tech

The Cheaper Tokens Get, the Bigger Your AI Bill Gets

As AI shifts from training to inference, the token is becoming an industrial commodity. Falling prices are not cutting demand, because agents and automation consume tokens at a scale no human chat session ever did.

Investing

Box Shows AI Its Manuals, and the AI Starts Redesigning the Workflow

Box Architect, built with NVIDIA, moves AI from reading documents to designing how work gets done. It has not yet reached beta, so the real-world payoff remains to be seen.

Tech

Marvell's $115 Billion Lumentum Deal: Is AI's Next Bottleneck Light, Not Chips?

Marvell's purchase of Lumentum points to a contest beyond GPU speed, in data transfer and power efficiency. Whether the price is right is not yet confirmed.

Tech

Semiconductors

This story's topic59 storiesSee all →

The whole business of designing, fabricating and testing chips.

The Next AI Chip Race Is About How Chips Connect

Applied Materials will lift its stake in Dutch equipment maker BESI from about 9% to 20%. The bet is less about control than about a new way to link chips.

Tech

White House Swaps AI for "SI." The Real Investment Signal Is Infrastructure, Not the Name

An executive order renames artificial intelligence "Super Intelligence" and six frontier labs sign a voluntary safety accord. What investors should watch is the data center, power, nuclear and chip policy now gathered under one national-competitiveness frame.

Tech

Before Robots Sell, Chips Sell First: Following the Money in the Humanoid Race

Billions of dollars are moving before humanoid robots reach mass sales. Compute suppliers may see revenue first, but that does not guarantee stock returns.

Tech

Tech Selloff: Is the Bull Market Over? Wall Street's Four-Year Chart Says Not Yet

The uptrend has survived five double-digit pullbacks in four years. This time, AI companies' debt and cash flow are the new variables.

Investing

Samsung Posts 107 Trillion Won Profit, TSMC Sales Jump 50%. Why Did the Shares Fall?

The AI chip supercycle has not ended. But the market's question has changed: not how fast earnings grow, but how long today's enormous margins can last.

Tech

AI Data Centers' Next Bottleneck: Getting Power From the Grid to the GPU

Securing a gigawatt of power does not mean a gigawatt reaches GPU compute. The next fight in AI infrastructure is happening inside the data center, over how little electricity gets wasted before it reaches the chip.

Tech

Memory Chip Supercycle: Not at Peak Yet, But 5 of 6 Warning Signs Are Flashing

Memory prices are still climbing, but the pace is slowing and PC and smartphone demand is starting to crack under the weight of higher chip costs. This looks less like a price peak and more like a profit plateau building toward the next supply cycle.

Investing

GOOGL

This story's topic25 storiesSee all →

Google Catches Up, and Rivals Run Ahead Again

The real test for Gemini 4 Argon is not AI performance. It is whether Google can win customers and keep them.

Tech

Same cloud, three different ways to earn the money back

Microsoft, Alphabet and Amazon are all spending enormous sums on AI infrastructure. The route each one takes to recover that capital is not the same, and that is what decides the winner.

Investing

Why Meta jumped 6.5%: Muse is the first price tag on AI compute

Muse is not a big revenue line yet. But in a capex race running toward $145 billion a year, it is the first product-shaped answer to the question every AI investor is asking: how does this money come back?

Tech

Big Tech's Cash Is Draining. Its AI Business Is Earning More

Nomura's J-curve points to a timing gap between falling free cash flow and the payback on AI spending. For investors, the two are not the same thing.

Tech

Dark navy cover. Horizontal bar chart of Reddit Q3 2026: revenue +53%, ad revenue +54%, global DAU +12%, US DAU -2%, stock about -29% on Oct 30. Title: Google AI Is Taking Reddit's Visitors.

Google's AI Search Is Cutting Reddit's Traffic. Are the Web's Economics Changing?

Reddit's revenue grew fast, but search visits fell. The shock points to a shifting relationship between companies that produce content and platforms that answer with AI.

Tech

Google Cloud Grew 103%. Alphabet Fell 10% Anyway. Here Is What Wall Street Fears

AI demand is showing up in the numbers, but search ads missed and AI spending is not yet returning enough cash. Alphabet's results ask not how fast it is growing, but how profitably.

Investing

Why Oracle Is Buying More Nvidia GPUs and Adding Google TPUs Too

Oracle will expand Nvidia's next-generation AI systems to multi-gigawatt scale while becoming the first big outside cloud to offer Google's TPUs. The move shows competition shifting from single-chip speed to the running cost of whole systems.

Tech