Only 20 chips out of 100 met quality standards
At a foundry in Hwaseong, South Korea, Samsung has crossed a threshold that chipmakers have long pursued — mass production of 3-nanometer semiconductors built on an entirely new transistor architecture. The shift from fin-shaped to gate-all-around transistors marks not merely an incremental refinement but a structural reimagining of how electricity flows through silicon. Though the first chips will find their way into Bitcoin mining rigs rather than smartphones, this moment belongs to the longer human story of bending physical limits ever further inward, toward the imperceptibly small.
- Samsung has officially begun mass production of 3nm chips, leapfrogging conventional transistor design with a gate-all-around architecture that promises 45% less power draw and 23% faster performance over current 5nm chips.
- A critical bottleneck looms beneath the announcement: yield rates below 20% mean most chips rolling off the line fail quality checks, making them unsuitable for the unforgiving standards of smartphone manufacturers.
- Rather than stall, Samsung is routing first-generation production toward Bitcoin mining processors — a less demanding application that lets the company refine the process while still generating commercial output.
- The mobile market remains the true destination, with Samsung targeting 2023 for smartphone-grade 3nm chips and a second-generation process that pushes gains even further — 50% power reduction, 30% performance boost, 35% smaller footprint.
- Industry observers are watching closely, as this transition to gate-all-around transistors sets the architectural direction for the entire semiconductor industry in the years ahead.
In late June, Samsung announced that its foundry in Hwaseong, South Korea had begun mass production of 3-nanometer chips — a milestone built on a fundamental change in how transistors are designed. Rather than the fin field-effect transistors that have defined the industry for years, Samsung's new chips use a gate-all-around approach, with its proprietary Multi-Bridge-Channel FET technology employing nanosheets to give manufacturers greater flexibility in tuning power and performance.
The first-generation results are striking on paper: 45% lower power consumption, 23% better performance, and a 16% smaller physical footprint compared to existing 5nm chips. A second generation, slated for 2023, promises to push those figures further still — 50% power reduction, 30% performance gain, and a 35% smaller die.
But the road from announcement to your next smartphone is not a straight one. Yield rates — the share of chips that pass quality inspection — remain below 20%, a figure too low for the exacting demands of mobile device makers. Samsung's pragmatic response is to direct initial production toward a Chinese firm making Bitcoin mining processors, where tolerances are more forgiving and the technology can be proven at scale before it graduates to flagship phones.
Dr. Siyoung Choi, who heads Samsung's foundry division, placed the achievement within a lineage of manufacturing firsts — high-K metal gate, FinFET, extreme ultraviolet lithography — framing gate-all-around transistors as the next step in a deliberate progression. The message is one of patience as much as ambition: prove the process, stabilize the yields, then bring the gains to the billions of people who carry a chip-dependent device in their pocket.
Samsung's foundry division announced in late June that it had begun mass production of 3-nanometer semiconductor chips at its manufacturing facility in Hwaseong, South Korea. The chips represent a significant leap in transistor design, moving away from the fin field-effect transistor architecture that has dominated the industry toward a new gate-all-around approach. This shift in how transistors are constructed promises meaningful gains in both power efficiency and computational speed.
The company's proprietary technology, called Multi-Bridge-Channel FET, uses nanosheets with wider channels to give customers flexibility in how they balance power consumption against performance. For the first generation of these 3nm chips, Samsung claims a 45 percent reduction in power consumption compared to its existing 5nm solutions. At the same time, the chips can deliver a 23 percent performance improvement while drawing the same amount of power as older designs. The physical size of the chips themselves shrinks by 16 percent. The second generation, expected to enter production in 2023, pushes these numbers further: 50 percent less power draw, 30 percent better performance, and a 35 percent smaller footprint.
Yet there is a practical constraint shaping how these chips will reach the market. The first batch of Samsung's 3nm production will not power smartphones or tablets. Instead, the company will supply them to a Chinese semiconductor firm specializing in processors for Bitcoin mining operations. Samsung's own announcement described the chips as intended for "high performance, low power computing applications," a characterization that aligns with the mining sector's needs. The reason for this detour around mobile devices is straightforward: yield rates remain problematic. Earlier reporting suggested that fewer than 20 percent of manufactured chips met quality standards—a rate too low for the demanding specifications of smartphone makers. Samsung has likely improved those numbers since then, but the company is not yet ready to commit its 3nm production to mobile processors.
Dr. Siyoung Choi, who leads Samsung's foundry division, framed the achievement as part of a longer trajectory of innovation. The company has previously pioneered high-K metal gate technology, FinFET architecture, and extreme ultraviolet lithography in the foundry space. The move to gate-all-around transistors at 3 nanometers represents the next step in that progression. Choi emphasized Samsung's intention to continue pushing the boundaries of manufacturing technology while working to bring the new process to full maturity.
What this means in practical terms is that the smartphone chips powering flagship devices in 2023 and beyond will likely benefit from these advances, but not immediately. The path from initial production to reliable, high-volume manufacturing for consumer devices typically requires months of refinement and yield optimization. Samsung is taking the measured approach of proving the technology in less demanding applications first, then scaling up to the mobile market once the process is stable enough to meet the exacting standards of device makers and consumers who expect their phones to work reliably for years.
Citas Notables
Samsung has grown rapidly as we continue to demonstrate leadership in applying next-generation technologies to manufacturing. We seek to continue this leadership with the world's first 3nm process with the MBCFET.— Dr. Siyoung Choi, head of Samsung's foundry division
La Conversación del Hearth Otra perspectiva de la historia
Why would Samsung send its first 3nm chips to Bitcoin miners instead of Apple or Qualcomm?
The yield rate is too low. Only about one in five chips coming off the line actually meet the quality standards needed for smartphones. Bitcoin miners are more forgiving—they just need raw computing power. Samsung uses this window to refine the process.
So this is essentially a proving ground?
Exactly. They're manufacturing at scale, learning where the defects happen, improving the process, and building confidence. By the time they're ready for mobile in 2023, they'll have months of real-world production data.
The performance gains sound dramatic. Forty-five percent less power, 23 percent faster. Are those real?
They're real, but they're measured in specific ways. The 45 percent figure assumes you're comparing apples to apples—same workload, same voltage. In a phone, the actual benefit depends on how the chip is used. But the direction is unmistakable: this is a genuine step forward.
What about the second generation? Fifty percent power reduction sounds almost too good to be true.
It's ambitious, but not impossible. They're promising it for early 2023. The gap between first and second generation is often where the real maturity happens. By then, yield rates should be much higher, and they'll have learned what works and what doesn't.
When will consumers actually see these chips in phones?
Probably late 2023 at the earliest, more likely 2024. Samsung needs to prove the yields are solid, then Qualcomm or Apple needs to design chips using the process, then those need to be validated and integrated into devices. These things take time.