Is Samsung falling further behind TSMC?

Is Samsung really falling further behind TSMC?

Yes. Samsung is falling further behind TSMC in semiconductor manufacturing, and the gap is now wider than at any point in their modern foundry rivalry.

According to TrendForce, TSMC captured 72.5% of the global foundry market in the second quarter of 2026, compared with Samsung's 5.9%. A year earlier, TSMC held approximately 70%, while Samsung remained above 7%. Samsung's foundry revenue increased just 1.8% from the previous quarter, against TSMC's 12.1%.

The latest results do not change that picture. TSMC reported $46.72 billion in third-quarter 2026 revenue, up 41.2% year over year in US dollars. Samsung's third-quarter results, released on 29 October 2026, cite a "notable revenue and earnings improvement" in foundry, but give no foundry revenue figure, so the two quarters cannot be compared directly.

We should be clear about what we're comparing. Samsung Electronics remains a major semiconductor manufacturer, particularly in memory chips. The gap examined here concerns Samsung Foundry, which manufactures chips designed by other companies, against TSMC, the world's largest dedicated chip foundry.

Samsung still has the technology, money and customers to compete, and its foundry is clearly busier than a year ago. What it currently lacks is evidence that this is translating into a larger share of the market. Comparable third-quarter market-share data is not yet available.

If you want more recent data on this point, please see our latest semiconductor industry report.

How much bigger is TSMC than Samsung Foundry today?

TSMC is now more than 12 times larger than Samsung Foundry by quarterly revenue. Their first- and second-place rankings make the competition look much closer than it really is.

TrendForce reported second-quarter 2026 foundry revenue of approximately $40.2 billion for TSMC and $3.26 billion for Samsung. That's a ratio of 12.3 to one. In a single quarter, TSMC generated roughly $36.9 billion more revenue, equivalent to over 11 quarters of Samsung's foundry revenue at its latest reported level.

The difference has grown quickly. In 2025, TrendForce estimated TSMC's annual foundry revenue at $122.54 billion and Samsung's at $12.63 billion, giving TSMC a 9.7-to-one advantage. Comparing an annual figure with a subsequent quarterly figure has limitations, but the widening ratio fits the broader market-share trend.

Samsung is also much closer to third-place SMIC than to TSMC. China's SMIC generated slightly more than $3 billion in the second quarter of 2026, against Samsung's $3.26 billion.

That means Samsung is fighting to preserve second place while TSMC operates at an entirely different scale. And the faster SMIC grows, the harder it becomes to describe Samsung as TSMC's only serious competitor simply because it ranks second.

Foundry metric	TSMC	Samsung
2025 revenue	$122.54B	$12.63B
2025 market share	69.9%	7.2%
Q2 2026 revenue	$40.2B	$3.26B
Q2 2026 market share	72.5%	5.9%
Q2 2026 revenue growth, QoQ	+12.1%	+1.8%
Q2 2026 revenue ratio	12.3× Samsung	1×
Q2 2026 revenue gap	$36.94B ahead	$36.94B behind
Q2 2026 market-share gap	66.6 points ahead	66.6 points behind

Has Samsung been losing foundry market share for years?

Samsung's foundry market share has been shrinking for several years, and the decline continued even as global demand for advanced chips surged.

Counterpoint Research's quarterly pure-foundry estimates show Samsung declining from 11% in early 2024 to 7% by the second quarter of 2026. TSMC moved from 63% to 73% over the same period.

We calculated that the gap widened from 52 to 66 percentage points in approximately two years. The trajectory is more revealing than any single quarter: Samsung lost ground during a period when AI accelerators, premium smartphone processors and custom cloud chips were driving record demand.

There was a brief improvement in Samsung's estimated share in early 2025, but it didn't last. By the second half of that year, Samsung was back near 7%, while TSMC continued climbing.

One methodological point is worth keeping in mind. Counterpoint's pure-foundry figures include Samsung's internal chip production, while TrendForce uses a different foundry-market methodology. The two series shouldn't be mixed, but both show Samsung losing share.

Quarter	TSMC share	Samsung share	Gap
Q1 2024	63%	11%	52 pts
Q2 2024	65%	10%	55 pts
Q3 2024	66%	9%	57 pts
Q4 2024	69%	8%	61 pts
Q1 2025	68%	9%	59 pts
Q2 2025	71%	8%	63 pts
Q3 2025	72%	7%	65 pts
Q4 2025	72%	7%	65 pts
Q1 2026	73%	7%	66 pts
Q2 2026	73%	7%	66 pts

Is TSMC taking nearly all the growth in the foundry market?

TSMC captured roughly 79% of the combined quarterly revenue growth among the world's ten largest foundries in the second quarter of 2026. Samsung accounted for only about 1%.

These percentages come from our calculations using TrendForce's reported figures. The top ten foundries generated approximately $53.49 billion in combined revenue, up from about $47.97 billion, an increase of roughly $5.5 billion. That covers the top ten only, not the whole market.

TSMC increased quarterly revenue from roughly $35.9 billion to $40.2 billion, adding about $4.3 billion. Samsung's revenue increased by approximately $58 million.

The contrast becomes even clearer when we look at demand. NVIDIA's AI processors, AMD's advanced chips and custom accelerators from major cloud companies have helped drive TSMC's growth. Samsung has benefited from additional advanced-node orders and demand for HBM base dies, but on a much smaller scale.

That explains how Samsung can report improving foundry sales while its market share keeps falling. Growing 1.8% is little comfort when your biggest competitor grows 12.1% in the same quarter. Samsung's stronger third quarter may narrow that difference, but without a revenue figure it cannot yet be measured.

For Samsung to narrow the gap, it needs to capture a much larger portion of the industry's new spending. The latest reported numbers show TSMC doing almost the opposite: taking most of that growth for itself.

If you want more recent data on this point, please see our latest semiconductor industry report.

Has TSMC already beaten Samsung in the 2nm chip race?

TSMC is currently ahead of Samsung in commercial 2nm production, even though Samsung adopted some of the underlying transistor technology earlier.

Samsung introduced gate-all-around transistors at 3nm in 2022, becoming the first major manufacturer to bring that architecture into production. TSMC continued using FinFET technology at 3nm before introducing nanosheet gate-all-around transistors with its N2 process.

Both companies started their initial 2nm production ramps in 2025. However, TSMC has already shown how quickly the new process is becoming a meaningful business.

According to its July 2026 earnings announcement, 2nm represented 3% of TSMC's quarterly wafer revenue. In its next quarterly report, that figure reached 8%. That's an increase of five percentage points in just three months.

Given TSMC's revenue scale, even a single-digit percentage represents billions of dollars in production. Samsung hasn't disclosed an equivalent breakdown for its own 2nm manufacturing.

Samsung's case is getting stronger. In its October 2026 results, it reported "continued order wins centered on 2nm HPC customers, including major hyperscalers," without naming them or giving order values. A foundry executive told the Herald that the number of 2nm projects won in 2026 should more than double from a year earlier. That is a project count, not revenue. Samsung also guides for new second-generation 2nm mobile products in the fourth quarter.

The commercial numbers are enough to show which company is ahead. Samsung now has a real 2nm customer pipeline, but TSMC has already demonstrated a much larger revenue-generating 2nm operation.

Are Samsung's 2nm chip yields still worse than TSMC's?

Samsung's 2nm yields remain a serious concern, although the exact performance gap with TSMC is impossible to establish from public financial disclosures.

A semiconductor yield measures how many chips manufactured on a wafer meet the required specifications. When yields are low, manufacturers spend money producing chips they cannot sell, increasing the effective cost of each working processor.

The public estimates disagree. In August 2026, Tom's Hardware discussed industry estimates placing Samsung's 2nm yields around 55%. In September, DIGITIMES reported SF2 yields of nearly 60%, while TrendForce News, citing Maeil Business Newspaper, put Samsung's 2nm yield above 70%. None of these is an audited company figure, and they may describe different chips or test conditions.

Still, the economics are worth understanding. Imagine two factories each producing 100,000 potential chips. At a 55% yield, the first produces 55,000 usable chips. At an 80% yield, the second produces 80,000.

That's roughly 45% more working chips from the same nominal output. The example is hypothetical, not an estimate of TSMC's actual yields, but it shows why even relatively small manufacturing differences can influence which foundry a customer chooses.

Meanwhile, TSMC's growing 2nm revenue share provides concrete evidence that customers are already buying chips manufactured using its newest process in significant volumes.

The commercial evidence is stronger than the leaked yield estimates. Samsung's yields are probably improving, but until it discloses production data or shows substantially greater 2nm revenue, the safer judgment is that TSMC remains ahead.

Why do Apple, NVIDIA and AMD keep choosing TSMC over Samsung?

Apple, NVIDIA and AMD help keep TSMC ahead because they repeatedly use its factories for some of the world's most valuable processors. These long-running relationships give TSMC a steady stream of advanced-chip orders that Samsung hasn't been able to match.

Apple relies on TSMC for the advanced processors powering iPhones, Macs and other devices. NVIDIA uses TSMC manufacturing and packaging technologies for its AI accelerator platforms. AMD depends on TSMC for major CPU and GPU families, including EPYC server processors and Instinct accelerators.

And those customers keep coming back.

A successful processor generation can lead to another, with design work beginning years before commercial launch. The foundry gains experience with demanding products, improves its manufacturing processes and develops a closer working relationship with the chip designer.

Over time, these relationships make switching more difficult. A chip built for TSMC's manufacturing process cannot simply be transferred unchanged to Samsung's equivalent node. Redesigning and qualifying a processor takes engineering work, money and time.

Samsung has manufactured important chips for Qualcomm, Google and its own Exynos processors. But several major customer decisions have favored TSMC, particularly at advanced nodes. Samsung's new hyperscaler orders are unnamed, so they cannot be counted as wins from these three companies.

There is some risk for TSMC in depending heavily on large customers. A slowdown in AI spending or weaker demand from Apple could affect its growth. For now, however, its customer base gives it access to far more advanced-chip volume than Samsung.

Did Samsung lose Qualcomm because its chips were worse?

Samsung lost major Qualcomm processor orders partly because of reported manufacturing problems, and the consequences extended well beyond one disappointing smartphone chip.

Samsung manufactured Qualcomm's original Snapdragon 8 Gen 1 using its 4nm process. Qualcomm subsequently moved production of the Snapdragon 8+ Gen 1 to TSMC, and later flagship Snapdragon generations reinforced TSMC's role.

The Elec reported in February 2022 that poor advanced-node yields were an important factor behind Qualcomm's manufacturing decisions. Its reported yield estimates were not independently audited, but the switch in manufacturing partner was publicly visible.

Qualcomm is an especially valuable customer because its Snapdragon processors power flagship smartphones from multiple manufacturers. Winning a Qualcomm generation brings volume, manufacturing experience and the chance to secure follow-on products.

Samsung's difficulties therefore had consequences that lasted longer than the original Snapdragon 8 Gen 1 production cycle.

A foundry can recover from a weak chip generation, but customers designing expensive flagship processors generally need reliable manufacturing schedules and predictable output. Repeatedly moving those programs between factories would be costly.

Qualcomm's transition was one of the clearest examples of Samsung losing an important external customer relationship to TSMC. The return is not settled either. In September 2026, The Bell reported that a possible Qualcomm 2nm deal had stalled over price, with Qualcomm pushing for lower prices and Samsung resisting. No new award has been confirmed.

If you want more recent data on this point, please see our latest semiconductor industry report.

Can Samsung's $16.5 billion Tesla deal help it catch TSMC?

Samsung's $16.5 billion Tesla agreement is one of its most important recent foundry wins, but it would take several deals of similar scale to make a serious dent in TSMC's lead.

Samsung disclosed the agreement in July 2025, reporting a contract worth approximately KRW 22.8 trillion, or about $16.5 billion, running through the end of 2033.

Elon Musk subsequently confirmed that Samsung's Taylor, Texas factory would manufacture Tesla's next-generation AI6 processor. On its October 2026 earnings call, Samsung said it had commenced clean-room operations there, and its outlook places the start of operations at the first Taylor fab in 2027. Commercial Tesla shipments have not started.

Tesla gives Samsung the opportunity to prove its capabilities with a demanding outside customer. Producing processors for Tesla's AI ambitions could also help Samsung attract other companies developing custom chips.

Yet the contract's headline value needs perspective.

Spread evenly over approximately eight years, $16.5 billion would represent around $2.1 billion per year. This is only an illustrative average, since actual production and payments could be concentrated in particular years.

TSMC generated $46.72 billion in revenue in the third quarter of 2026 alone. Tesla's entire announced contract is worth approximately 35% of that single TSMC quarter.

The deal matters more for the type of customer Samsung has won than for the immediate revenue it guarantees. If Samsung manufactures Tesla's AI6 chips successfully, secures follow-on generations and attracts more AI chip designers, the contract could mark the beginning of a broader recovery.

It's a big contract. The revenue gap is bigger.

Is Apple starting to move chip production from TSMC to Samsung?

Apple is giving Samsung new chip-manufacturing business, but there is currently no evidence that it is moving its main iPhone or Mac processors away from TSMC.

In August 2025, Apple announced a partnership involving Samsung's Austin, Texas factory and a new chip-manufacturing technology intended for products including iPhones.

Reuters and the Financial Times linked the initiative to advanced image sensors, an area where Sony has historically been a major Apple supplier.

That makes the deal interesting for Samsung, but it also limits what we can conclude.

Apple's A-series processors for iPhones and M-series processors for Macs remain closely tied to TSMC. In May 2026, Bloomberg reported that Apple had explored using Intel and Samsung for main device chips in the US, but said neither effort had produced orders. No such order has been confirmed since.

Apple also has good reasons to work with multiple semiconductor manufacturers. It can diversify suppliers, develop new components and expand American production without changing its primary processor partner.

When we combine the Apple and Tesla agreements, Samsung's strategy becomes clearer. The company is attracting major US customers through specific manufacturing projects.

Those projects could grow into broader relationships, especially if Samsung delivers well. But calling Apple's sensor-related manufacturing work a major defeat for TSMC would considerably exaggerate its significance.

Is TSMC's advanced chip packaging becoming another problem for Samsung?

TSMC's lead in advanced packaging has become a major obstacle for Samsung, particularly in the AI chip market where the processor and its memory must work together at extremely high speeds.

Modern AI accelerators often combine large logic chips with several high-bandwidth memory stacks. Connecting these components efficiently requires sophisticated packaging technologies.

TSMC's CoWoS platform has become central to manufacturing high-performance AI processors, including products designed by NVIDIA and other major semiconductor companies.

Its 2025 annual report described multiple CoWoS technologies already in volume production, alongside further development of larger packages and advanced three-dimensional integration.

Demand has encouraged significant industry investment. A June 2025 DIGITIMES analysis projected global CoWoS and related packaging capacity to reach approximately 1.31 million wafer-equivalent units in 2026, with TSMC remaining the principal supplier. That was a capacity forecast rather than confirmed production.

Samsung is also developing advanced packaging and has a potentially valuable advantage: it can manufacture logic chips, HBM memory and packaging technologies within the same group.

However, offering all three capabilities doesn't necessarily mean customers will buy the complete package. Each component must meet performance, yield and delivery requirements.

TSMC already has large AI customers using its manufacturing and packaging technologies together. Samsung needs to prove that its integrated approach can compete with those established production relationships.

Could Samsung's HBM business help it beat TSMC in AI chips?

Samsung's HBM business could help it win more AI chip manufacturing orders, but its success in memory hasn't yet translated into a foundry comeback.

Samsung is one of the world's major DRAM and NAND manufacturers and is developing new generations of high-bandwidth memory. TSMC doesn't manufacture these memory chips internally.

This difference creates an opportunity for Samsung to offer chip designers a combination of advanced logic production, HBM and packaging.

The memory side is now very strong. On its October 2026 earnings call, Samsung said HBM4 sales more than tripled from the previous quarter and reached 60% of its HBM sales. Reuters reported that Samsung says its HBM supply for next year is already sold out.

HBM base dies are particularly relevant because they contain logic circuitry that helps control the memory stack. Samsung can manufacture these components in its own foundries, giving its logic business another source of demand.

But Samsung's semiconductor division includes more than the foundry business. Strong memory results tell us that Samsung's overall semiconductor operations are doing well, not that its contract chip manufacturing has suddenly caught TSMC.

There is a catch: memory suppliers can sell HBM for processors manufactured by TSMC. Samsung can benefit financially from an AI accelerator without manufacturing its main logic chip.

For Samsung Foundry, the opportunity is to persuade customers that combining these capabilities under one manufacturer offers better performance, pricing or delivery. We haven't yet seen that strategy produce a major shift in foundry market share.

Can Samsung afford to keep spending as much as TSMC on new chip factories?

Samsung can afford enormous semiconductor investments, but TSMC currently has a much stronger financial engine dedicated to foundry manufacturing.

In the third quarter of 2026, TSMC reported revenue of $46.72 billion, a gross margin of 66.4% and an operating margin of 58.7%.

Those margins are exceptional for a company running expensive semiconductor factories. They also give TSMC substantial resources to fund new processes, packaging facilities and manufacturing capacity.

TSMC raised its full-year revenue growth outlook to approximately 44% in US dollars and maintained a capital expenditure plan of $60 billion to $64 billion.

Samsung Electronics is far from financially weak. On its October 2026 call, it guided for about KRW 84.4 trillion of group capital spending this year, including KRW 76.8 trillion for semiconductors. It does not say how much of that goes to foundry rather than memory.

The difficulty is that Samsung Foundry has not demonstrated the same financial performance as TSMC's manufacturing business. Samsung doesn't publish a standalone foundry income statement, and analysts quoted by Reuters in October 2026 said the business remained at a loss.

A foundry with strong customer demand can spread the huge cost of new equipment across more wafers. Higher utilization can improve profitability and make the next investment easier to fund.

Samsung has to build expensive manufacturing capacity while working to attract enough customers to fill it. Its Taylor, Texas investment, for example, provides potential capacity for Tesla and future customers, but the commercial returns depend on successful production.

Samsung has the money to remain in the race. TSMC currently has the better economics for staying ahead.

If you want more recent data on this point, please see our latest semiconductor industry report.

Is Samsung Foundry finally recovering?

Samsung Foundry is recovering in activity, but not yet in profit or market share. The improvement is real, and still too small to reverse its competitive decline.

Samsung's third-quarter 2026 results, published on 29 October, reported a notable improvement in foundry revenue and earnings, alongside continued 2nm order wins from HPC customers, including major hyperscalers. A foundry executive told the Herald that high-performance computing should exceed 50% of foundry sales next year, a forecast rather than an achieved mix.

For the fourth quarter, Samsung guides for double-digit foundry revenue growth, helped by new second-generation 2nm mobile products and a 4nm NPU production ramp.

There are several reasons to take these developments seriously.

Tesla's AI6 agreement gives Samsung a major outside customer for its Taylor factory, where clean-room operations have begun. Apple's specialized manufacturing project offers another long-term relationship. Demand for HBM base dies also provides more work for Samsung's existing foundry lines.

Together, these developments suggest Samsung is becoming better positioned to fill manufacturing capacity and generate additional revenue.

Still, Samsung gave no foundry revenue figure, and Reuters reported that analysts consider the foundry still loss-making. The last measured quarter, Q2 2026, showed only 1.8% sequential growth at Samsung against 12.1% at TSMC.

The fourth-quarter growth target is guidance, not a result, and TSMC itself guides for 12% to 16% sequential growth in the same quarter.

A recovery is now credible, but the market-share data still favors TSMC. Samsung needs several quarters of growth faster than TSMC's, and a return to profit, before this counts as a real change in the rivalry.

How fast would Samsung have to grow to catch TSMC?

Samsung would need extraordinarily fast foundry revenue growth to close the gap with TSMC, especially if TSMC continues expanding at anything close to its recent pace.

We can calculate the challenge using TrendForce's second-quarter 2026 market shares: 72.5% for TSMC and 5.9% for Samsung.

That gives TSMC a revenue advantage of roughly 12.3 to one.

Suppose TSMC grows its foundry revenue by 20% annually over the next three years. If Samsung grows 30% annually during the same period, Samsung's revenue relative to TSMC would increase by approximately 27%.

That's real progress, but TSMC would still be nearly ten times larger.

To halve the revenue ratio within three years under the same 20% TSMC growth assumption, Samsung would need annual revenue growth of approximately 51%.

These are hypothetical scenarios rather than forecasts. They show the size of the challenge created by years of unequal growth.

Samsung's most realistic near-term objective is therefore to stop losing market share. The fourth quarter is a first test: Samsung guides for double-digit foundry growth, while TSMC guides for revenue of $52.3 billion to $54.3 billion, 12% to 16% above its third quarter. Faster 2nm production, Taylor's 2027 start and more external customers could help.

Even getting back to a double-digit foundry market share would require a major improvement from today's position.

Is Samsung falling further behind TSMC, or can it still catch up?

Yes, Samsung is still behind TSMC by a wider margin than in the past, and the latest comparable data shows no narrowing. TSMC has a much larger foundry business, its market-share lead has widened for years, and it is capturing most of the new spending on advanced chip manufacturing.

Counterpoint's figures show the gap between the two companies growing from 52 percentage points in early 2024 to 66 points by the second quarter of 2026. TrendForce's revenue data shows that TSMC is now more than 12 times larger.

The technological picture also favors TSMC. Its 2nm process has already become a meaningful source of wafer revenue, while its advanced packaging business benefits from strong demand for AI processors.

Samsung's position is improving, though. Its October 2026 results show better foundry revenue and earnings, 2nm orders from hyperscalers and a Taylor factory moving toward a 2027 start. Its HBM business gives it a way to combine memory, logic manufacturing and packaging.

These wins have not yet produced measurable share gains, and the foundry is still reported to be losing money.

Our judgment is that Samsung can remain a major foundry competitor and potentially recover some lost ground. But catching TSMC would require years of faster growth, successful advanced-node manufacturing and repeated large customer wins.

For now, TSMC remains far ahead. Samsung has stopped looking stuck, but it has not yet shown it can grow faster than TSMC.

If you want more recent data on this point, please see our latest semiconductor industry report.