Samsung Electronics has raised prices for advanced contract chipmaking services by up to 15% on new orders, according to a Reuters report published August 19 citing two sources familiar with the matter. The increases, which took effect in July for chips manufactured on Samsung’s 4-nanometer (SF4) and 5-nanometer processes, reflect a structural shift in the global semiconductor supply chain. Surging demand for artificial intelligence chips has absorbed the vast majority of leading-edge capacity at Taiwan Semiconductor Manufacturing Co., pushing customers toward Samsung and Intel as alternative foundries and giving Samsung pricing leverage it has not held since its foundry division began losing money in 2022.
Key Takeaways
- Samsung Electronics raised prices for advanced contract chipmaking by up to 15% on new orders starting in July 2026, with increases ranging from 5% to 15% depending on the process node
- Chinese customers are absorbing the steepest price increases, driven by U.S. restrictions on exports of advanced chipmaking equipment to China that have increased reliance on overseas foundries
- TSMC controls more than 70% of global foundry revenue and has pre-sold all 3-nanometer capacity through 2027 and all 2026 2-nanometer output to Apple, Nvidia, and AMD
- Samsung expects AI and high-performance computing applications to account for more than 30% of foundry revenue in 2026, up from 15% to 20% in late 2025
- Nvidia’s H200 chips are reaching China after Trump administration export approvals, with Chinese firms placing orders exceeding 2 million units at approximately $27,000 per chip against 700,000 in current inventory
TSMC’s Capacity Constraints Create an Opening for Samsung
The price increase did not originate with Samsung. It followed TSMC, which has effectively run out of leading-edge capacity to offer new customers. TSMC produced more than 70% of global foundry revenue in the first quarter of 2026, according to research firm Counterpoint. The company has pre-sold its entire 3-nanometer capacity through 2027. All of its 2-nanometer output for 2026 has been allocated to Apple, Nvidia, and AMD. When the dominant foundry cannot accept new orders at any price, customers migrate to the next available option.
Samsung, which accounted for approximately 7% of global foundry revenue in the same quarter, is the primary beneficiary of that migration. The company’s foundry business has operated at a loss since 2022, struggling to close the yield and technology gap with TSMC even as Samsung’s memory division reported record profits driven by high-bandwidth memory (HBM) chips used in AI systems. The price increases change the trajectory. Lee Min-hee, a Seoul-based analyst at BNK Investment & Securities, told Reuters that Samsung’s foundry business could become profitable as early as next year if prices continue to rise, a timeline that would be ahead of previous market expectations.
Samsung’s ability to raise prices reflects a market in which demand for advanced semiconductor manufacturing has outstripped the combined capacity of the world’s foundries. The company expects advanced processes to generate more than half of its foundry revenue in 2026, with AI and high-performance computing applications accounting for more than 30%, up from 15% to 20% in late 2025. That composition shift means Samsung’s foundry is no longer dependent primarily on mobile chip orders, which carry thinner margins and face more competitive pricing pressure.
Chinese Customers Pay the Highest Premiums
Chinese customers are accepting the steepest price increases among Samsung’s foundry clients, according to one of the Reuters sources. The dynamic is a direct consequence of U.S. export controls. Washington has restricted the sale of advanced chipmaking equipment to China, limiting the ability of Chinese semiconductor manufacturers to build out domestic leading-edge fabrication capacity. That restriction increases Chinese firms’ dependence on foreign foundries like Samsung and TSMC for the most advanced chips.
The demand from Chinese customers has been strong enough that Samsung has been unable to fill all orders. The company must also reserve capacity for U.S. clients and for its own internal chip production needs, creating a three-way allocation challenge. The result is a tiered pricing structure where customers in different regions face different cost structures for the same silicon, and where Chinese buyers, facing the fewest alternatives, absorb the highest premiums.
The geopolitical dimension extends beyond foundry services. Nvidia’s H200 AI chips are beginning to reach China after the Trump administration approved exports with a 25% fee. Chinese technology companies have placed orders exceeding 2 million H200 units for 2026 delivery, according to Reuters, but Nvidia currently holds only approximately 700,000 chips in inventory, including about 100,000 GH200 Grace Hopper superchip variants. Nvidia has asked TSMC to ramp up production of additional H200 chips starting in the second quarter of 2026. The H200 is priced at approximately $27,000 per chip for Chinese customers, and the eight-chip module runs about 1.5 million yuan, still cheaper than gray-market alternatives priced above 1.75 million yuan.
ByteDance and Chinese Tech Giants Compete for Limited Supply
ByteDance, the parent company of TikTok, is planning to spend 100 billion yuan on Nvidia chips in 2026, up from 85 billion yuan in 2025, if Chinese regulators approve H200 imports. Beijing has not yet formally greenlighted incoming shipments, creating regulatory uncertainty that sits alongside the supply constraints. The combination of regulatory approval delays, limited foundry capacity, and Nvidia’s inventory shortfall means that Chinese AI development timelines are partly governed by supply chain logistics rather than technical capability.
The H200, based on Nvidia’s Hopper architecture and manufactured on TSMC’s 4-nanometer process, offers approximately six times the performance of the now-banned H20, a weaker chip that Nvidia had specifically designed for the Chinese market before Beijing blocked it. Chinese firms view the H200 as a meaningful upgrade that justifies both the higher per-unit cost and the diplomatic complexity of securing import approval. The first deliveries from Nvidia’s existing inventory are scheduled to reach Chinese clients ahead of the Lunar New Year, with additional supply following once TSMC completes its production ramp.
The scale of Chinese demand raises concerns about whether Nvidia can balance its obligations to Chinese customers with supply commitments elsewhere. Nvidia has stated that licensed H200 sales to authorized Chinese customers “will have no impact on our ability to supply customers in the United States,” but the math of 2 million units ordered against 700,000 in stock suggests that global AI chip supply will remain tight through at least the second half of 2026.
Samsung’s Foundry Business Approaches a Profitability Inflection
Samsung’s foundry division has been a persistent drag on the company’s semiconductor segment. While the memory business has posted record results, fueled by HBM3E chips that Samsung and SK Hynix both priced 20% higher for 2026 deliveries, the foundry has lost money for four consecutive years. The yield gap with TSMC, particularly on the most advanced nodes, has limited Samsung’s ability to win high-volume orders from the largest fabless chipmakers.
The AI demand wave is changing that equation. Samsung raised prices on its SF4 process in July, with increases varying from 5% to 15% depending on the customer and process node. The company is simultaneously expanding its Pyeongtaek fabrication complex in South Korea to add capacity. If the pricing holds and order volumes continue to rise, the foundry division could reach profitability by 2027, according to BNK Investment & Securities. That timeline would represent a turnaround that Samsung’s foundry leadership has pursued for years without success in a market where TSMC’s manufacturing advantages consistently attracted the most valuable orders.
Intel, the third major advanced foundry, is also positioned to absorb overflow demand from TSMC. The three-way dynamic between TSMC, Samsung, and Intel is reshaping foundry economics globally. For the first time in the current AI cycle, all three leading-edge manufacturers are operating near or at capacity, and all three are raising prices. The era of foundries competing primarily on cost to win orders is giving way to a market where access to any leading-edge capacity at any price is the primary constraint.
The Supply Chain Tightens Across Every Layer
The pricing pressure is not confined to foundry services. Samsung and SK Hynix raised HBM3E memory prices by 20% for 2026 orders, reflecting the same demand-supply imbalance in the memory layer of the AI chip stack. TSMC has raised its own prices by 8% to 10% for advanced processes below 5 nanometers, increases that flow through to the cost of every chip Apple, Nvidia, Qualcomm, and AMD manufacture. Nvidia’s H200 chips, priced at $27,000 each for Chinese customers, reflect the combined effect of constrained foundry capacity, limited inventory, and a 25% export fee imposed by the U.S. government.
The cumulative result is an AI hardware supply chain where every major component, from the foundry wafer to the finished accelerator, is priced higher than it was 12 months ago. For hyperscale cloud providers, AI startups, and sovereign computing programs, the cost of building AI infrastructure is rising faster than the cost of the models themselves. For Samsung, the moment represents an opportunity to convert its long-struggling foundry business into a profitable operation by charging prices the market will now bear. Whether that window stays open depends on how quickly TSMC, Samsung, and Intel can add capacity, and whether AI chip demand continues to grow at a rate that outpaces their ability to build it.
FAQs
How much has Samsung raised chipmaking prices?
Samsung raised prices for advanced contract chipmaking services by up to 15% on new orders starting in July 2026. The increases apply to chips manufactured on Samsung’s 4-nanometer and 5-nanometer processes, with the size of the increase varying from 5% to 15% depending on the process node and the customer’s region.
Why is Samsung able to raise foundry prices now?
TSMC, which controls more than 70% of global foundry revenue, has pre-sold its leading-edge capacity through 2027. Customers unable to secure TSMC manufacturing slots are turning to Samsung and Intel as alternatives, giving Samsung pricing leverage it has not held since its foundry business began losing money in 2022.
Why are Chinese customers paying the highest premiums?
U.S. restrictions on exports of advanced chipmaking equipment to China have limited domestic fabrication capacity, increasing Chinese firms’ reliance on overseas foundries. With fewer alternatives available, Chinese customers face the steepest price increases and are accepting them to maintain access to advanced chip manufacturing.
How does Nvidia’s H200 chip demand affect the supply chain?
Chinese technology companies have placed orders exceeding 2 million H200 chips at approximately $27,000 per unit, but Nvidia currently holds only about 700,000 in inventory. Nvidia has asked TSMC to ramp up production, and ByteDance alone plans to spend 100 billion yuan on Nvidia chips in 2026 if Chinese regulators approve H200 imports.
