First in Three Years! Samsung Wafer Foundry Returns to Profitability in June, 4nm Process Yield Rate Improved to Around 80%
Author | Zhang Yaqi
Editor | Hard AI
Samsung Electronics' foundry business, after years of losses, is finally showing signs of a turnaround.
According to sources in the Korean semiconductor industry,in June of this year Samsung Electronics’ foundry division achieved monthly profitability for the first time since 2023.This change was driven by a steady increase in HBM (High Bandwidth Memory) base die shipments as well as a significant improvement in 4nm process yield.
This positive momentum has made Samsung internally optimistic about achieving positive quarterly earnings in Q3. Meanwhile, the company is swiftly strengthening its presence in the AI chip foundry sector – following the Tesla AI6 chip orders, the production of Groq chips and potential collaborations with Meta and Anthropic have also surfaced. Market expectations for Samsung’s foundry business recovery are on the rise.
01
Turning a monthly profit, with the possibility of quarterly profitability in Q3
Samsung Electronics foundry division turned a monthly profit in June 2024, its first since 2023. The unit had long been under pressure—suffering from poor yields in advanced processes, major client losses, and low capacity utilization, leading to expanding losses. Samsung does not disclose the foundry segment's financials separately,but market estimates put combined operating losses for foundry and System LSI at about 2.5 trillion won in 2023, climbing to 5.3 trillion won in 2024, and further to about 6 trillion won in 2025.
As for Q2 as a whole, April and May were still loss-making months, so a quarterly turnaround is not yet confirmed. However, given that June’s profit came not from a one-off settlement but from sustained improvements in capacity utilization and process yield, Samsung internally believes the possibility of a quarterly profit in Q3 is high.
02
HBM4 boosts 4nm utilization; yield improvements reduce unit cost
The main drivers behind this single month's turnaround werethe rising shipments of HBM Base Die and simultaneous improvements in 4nm process yield.
The HBM Base Die is the logic die located at the bottom of DRAM stacks, responsible for interfacing with the GPU, and is produced on Samsung's foundry lines. Increased HBM production synchronously drives Base Die wafer starts, thereby improving utilization in advanced process lines—particularly since HBM4 Base Die uses the 4nm process, creating continuous supplementary demand for that line. In February this year, Samsung became the first in the world to mass-produce and ship HBM4; in May, they also delivered HBM4E 12-stack samples to global customers.
The foundry business typically has high fixed costs such as depreciation, labor, and maintenance, so repeated shipments help improve equipment utilization and lower unit costs. At the same time, industry insiders estimate Samsung has now improved its 4nm yield to about 80%. Higher yields mean more chips are shipped for the same wafer starts and lower costs for scrap and rework. In April and May, the initial ramp-up phase was still affected by costs and process stabilization pressures, but in June the simultaneous effect of Base Die volume ramp and yield improvement brought single-month profitability.
03
Major clients return, as a more diversified AI chip supply chain brings new opportunities
The return to monthly profitability coincided with an improved order structure. Last year, Samsung Electronics foundry won the Tesla AI6 chip order, and has recently also taken on the AI inference chip business based on the Groq architecture as announced by Nvidia. In addition, Meta and Anthropic are also mentioned as potential foundry partners for Samsung's own AI chips, further boosting expectations for a recovery in Samsung's foundry business.
With the rapid growth in AI training and inference demand, TSMC’s advanced process and advanced packaging capacity is tightly constrained. Large technology companies seeking to reduce their dependence on Nvidia are ramping up self-developed AI chips and simultaneously looking to break TSMC's single-source dependency. In this context, thanks to its advanced 2nm process layout, HBM Base Die capacity, and a US manufacturing base, Samsung is gradually attracting attention as an alternative supply chain option.
Despite the positive signs, Samsung Foundry still lags significantly behind TSMC. According to market research firm TrendForce,in the first quarter of 2025, TSMC will have a global foundry market share of 72.3%, while Samsung is second at 6.5%.Compared to the same period last year, TSMC's share rose by 4.7 percentage points from 67.6%, while Samsung's dropped by 1.2 percentage points from 7.7% to 6.5%. The gap between the two has widened from 59.9 to 65.8 percentage points.
Hard·AI
Disclaimer: The content of this article solely reflects the author's opinion and does not represent the platform in any capacity. This article is not intended to serve as a reference for making investment decisions.
You may also like

Wall Street giants collectively raise rate hike forecasts, US Treasury futures continue to fall, gold drops over 1% to break 4000
After the US October CPI exceeded expectations, Goldman Sachs and Nomura added a forecast for an interest rate hike in March next year, while Deutsche Bank included a forecast for a rate hike in January next year. Currently, the market has fully priced in a 25 basis point rate hike in December, with a 13% probability for a 50 basis point hike. On Wednesday, the US Treasury spot market was closed, but Treasury futures continued to fall, with corresponding yields rising by about 2 to 4 basis points, and spot gold came under pressure, dropping by 1.1%.

No Need to Worry About the Cooling Down of AI Data Center Construction? The US Import Boom Is Unlikely to Subside in the Short Term, Real Impact May Come in 2029
It takes an average of 4.8 years from finalizing investment to starting construction of a data center project, and an additional 2.6 years from construction start to becoming operational. Servers usually cannot be installed until after the building structure and racks are completed. For trade partners such as Taiwan (China), South Korea, and Mexico who rely on US data center supply chain demand, the inertia in construction timelines provides at least a one- to two-year buffer for future exports.

Federal Reserve Major Debate: When Should the Tenure of “Number Three” Williams End?
The main point of contention is whether Williams should step down in June 2028, upon completing a full 10-year term, or in June 2027, when he reaches the mandatory retirement age of 65. Analysts point out that the Federal Reserve Board holds ultimate discretion; if he leaves early, Fed Chair Waller would be able to assert greater control over the appointment to this critical position. However, forcing Williams to shorten his term could easily be interpreted as political manipulation. Williams himself is expected to serve until June 2028.

