The relentless surge in artificial intelligence workloads has pushed TSMC’s most advanced fabrication plants to their limits, forcing many chip designers to seek alternative manufacturing partners. Samsung has emerged as the primary beneficiary, and according to insiders who spoke with Reuters, the South Korean giant is now leveraging this newfound demand to raise its contract manufacturing prices by as much as 15%.
Starting in July, Samsung began charging customers in the United States and China between 10% and 15% more for its 4nm process node, known internally as SF4, compared to the previous month. Taiwanese firms are also feeling the pinch, with price increases ranging from 5% to 10% on the same node. But the adjustments aren’t confined to the relatively recent 4nm technology. The 5nm node (SF5) has seen a similar 10-15% uptick, while the older, more mature 8nm node is now almost 10% more expensive than it was just a few weeks ago.
This strategic pricing move signals a significant shift in the semiconductor landscape. For years, Samsung trailed TSMC in both technology leadership and yield rates, often having to undercut its rival to win orders. Now, with AI accelerators, custom silicon, and high-performance computing chips flooding the market, the dynamics have changed. Samsung’s foundry division is no longer just a backup option; it’s a critical supplier with the leverage to command premium rates.
Why the Sudden Price Surge?
The core driver behind this price hike is the unprecedented demand for AI-specific chips. Companies like NVIDIA, AMD, and a host of cloud providers are scrambling to secure manufacturing capacity for GPUs and custom ASICs. TSMC’s 3nm and 5nm fabs are running at near-total utilization, leaving little room for new orders. As a result, many fabless chip designers are turning to Samsung as their second source, even if it means paying more.
But there’s more to the story than just overflow demand. Samsung has also been working diligently to improve its yield rates on advanced nodes. Higher yields mean fewer defective dies per wafer, which translates to lower effective costs for Samsung. However, the company is choosing to pocket those efficiency gains as profit rather than pass them on to customers. This is a classic supply-and-demand response, but it also reflects a deliberate strategy to reposition its foundry business as a premium service.
The Role of AI Accelerators and Custom Silicon
AI accelerators are notoriously large and complex chips, often exceeding 800 square millimeters in die size. This means they consume a huge number of wafers, and even a small number of orders can tie up a fab’s capacity for months. With TSMC’s leading-edge nodes fully booked, Samsung’s 4nm and 5nm lines have become the go-to alternative for companies like Qualcomm, Apple (for some components), and a growing list of Chinese AI startups.
Chinese firms, in particular, are aggressively seeking domestic and non-TSMC manufacturing options due to export controls and geopolitical tensions. This has created a perfect storm: US-based companies want to diversify away from Taiwan, while Chinese companies are often forced to do so. Samsung, with its fabs in South Korea and the US, is uniquely positioned to serve both markets, and it’s capitalizing on that advantage.
Breaking Down the Price Increases by Node
Let’s take a closer look at the specific price adjustments across Samsung’s key process nodes, as reported by industry insiders.
4nm (SF4): The Premium Tier
The 4nm node is Samsung’s most advanced in terms of high-volume manufacturing for mobile and AI applications. The 10-15% price increase for US and Chinese customers is the most aggressive move. For Taiwanese clients, the 5-10% hike is slightly more moderate, possibly reflecting long-term partnership agreements or volume commitments. This node is used for flagship smartphone processors and mid-range AI inference chips, making it a critical revenue generator for Samsung.
5nm (SF5): The Workhorse
The 5nm node is still widely used for a variety of products, including networking chips, automotive controllers, and some AI accelerators. The 10-15% price increase here is notable because this node is more mature, and one might expect prices to drop over time as yields improve. Instead, Samsung is using the capacity crunch to push prices up, signaling that even older nodes are in high demand due to the overall semiconductor shortage in specialized segments.
8nm (SF8): The Legacy Cash Cow
Even the 8nm node, which has been in production for several years, is seeing a nearly 10% price increase. This is particularly interesting because 8nm is considered a legacy node for many applications, but it’s still used in a wide range of products, from Wi-Fi chips to power management ICs. The price hike on this node suggests that the demand for semiconductor capacity is so broad that it’s spilling over into older technologies.
Impact on Chip Designers and Consumers
For fabless companies, these price increases will inevitably lead to higher production costs. In the short term, many will absorb the extra expense to maintain market share. However, over time, these costs will be passed down the supply chain. Smartphone manufacturers, PC makers, and cloud service providers will likely see their component costs rise, which could translate into higher prices for end consumers.
But there’s a silver lining for Samsung. The increased revenue from these price hikes will allow the company to invest more heavily in R&D for next-generation nodes, such as 2nm and 1.4nm. This could help Samsung close the technology gap with TSMC, which has traditionally held a 1-2 year lead in process technology.
What This Means for the Foundry Market
The foundry market is now a seller’s market, and Samsung is clearly taking advantage of it. This is a stark contrast to just a few years ago when Samsung was aggressively discounting its wafers to win orders from TSMC. The shift in pricing power is a testament to the strategic importance of advanced packaging and process technology in the AI era.
However, it’s not all smooth sailing for Samsung. The company has faced criticism in the past for yield issues on its 3nm gate-all-around (GAA) technology. If Samsung wants to maintain these price increases in the long term, it must deliver reliable, high-yield processes that meet customer expectations. Otherwise, clients may flock back to TSMC once its capacity expands.
Geopolitical Factors at Play
The price increases also come against a backdrop of heightened geopolitical tensions. The US has imposed strict export controls on advanced semiconductor technology to China, which has forced Chinese companies to seek alternatives. Samsung, with its US-based fab in Austin, Texas, and its South Korean facilities, is seen as a relatively safe supplier for both US and Chinese clients, though it must navigate complex regulations.
For Chinese customers, paying 15% more for Samsung’s 4nm node is still a better option than not having access to advanced chips at all. This dynamic gives Samsung significant pricing power in the Chinese market, even as it faces potential political backlash from Washington for supplying technology that could be used in AI systems.
Long-Term Contracts vs. Spot Pricing
It’s worth noting that these price increases likely apply to new orders or spot purchases, not necessarily to long-term contracts that were signed months ago. Many major clients, such as Qualcomm and NVIDIA, have multi-year agreements that lock in prices. However, as these contracts expire, they will be renegotiated at the higher rates, which means the full impact of these price hikes will be felt over the next 12-18 months.
For smaller fabless companies without long-term agreements, the price increase is immediate and severe. This could lead to consolidation in the semiconductor industry, as smaller players struggle to absorb the higher costs, potentially being acquired by larger firms.
What’s Next for Samsung’s Foundry Business?
Looking ahead, Samsung is expected to continue its aggressive pricing strategy as long as the AI boom persists. The company has already announced plans to expand its advanced packaging capabilities, which are critical for AI chips that require high-bandwidth memory (HBM) integration. This expansion will require significant capital investment, and the current price hikes will help fund those efforts.
Additionally, Samsung is ramping up its 2nm GAA process, which is expected to enter mass production in 2025. The company has already secured several early customers, including a major AI chip designer. If Samsung can successfully ramp 2nm with good yields, it could further solidify its position as a leading foundry and justify even higher prices in the future.
Advice for Chip Buyers
For companies that rely on Samsung for their chip manufacturing, now is the time to renegotiate contracts and lock in prices where possible. Diversifying across multiple foundries, including Intel Foundry Services, which is also ramping up its advanced nodes, could provide leverage in future negotiations. However, with TSMC’s capacity fully booked and Intel still catching up, Samsung currently holds the upper hand.
In the short term, the price increases are likely here to stay. The AI boom shows no signs of slowing down, and the demand for advanced chips will only grow as more applications—from autonomous vehicles to generative AI—come online. Samsung’s decision to raise prices by up to 15% is a clear signal that the semiconductor industry is entering a new era of pricing power for manufacturers.
For consumers, this might mean slightly higher prices for next-generation smartphones, laptops, and AI-powered devices. But for the industry as a whole, it’s a sign of health and investment, as companies like Samsung reinvest their profits into the next wave of innovation.
As the situation evolves, industry watchers will be closely monitoring whether TSMC responds with its own price adjustments, and whether Samsung can maintain its yield improvements to justify the premium. One thing is certain: the days of cheap, abundant chip manufacturing capacity are over, at least for the foreseeable future.
