Semiconductor Equipment Pricing Dynamics Shift Amid Surge in AI Demand

For a long time, the pricing power in the semiconductor supply chain has been structured as a distinct pyramid. At the top are giants like Apple, NVIDIA, Microsoft, Google, and Amazon, which control end-user demand, cloud computing orders, and system specifications. Below them are manufacturing leaders such as TSMC, Samsung, SK Hynix, and Micron, which dominate advanced manufacturing, advanced storage, and critical production capacity. In contrast, equipment suppliers, although positioned at the upstream of the manufacturing ecosystem and often possessing high technological barriers in certain areas, still frequently face pressures such as annual cost-reduction demands, repeated price negotiations, delayed acceptance timelines, and order cancellations within the procurement systems of major clients.

As a result, the semiconductor equipment industry has developed an unwritten rule: introducing new equipment (Design-in) often requires equipment suppliers to make significant price concessions; during subsequent repeat orders, wafer fabs typically follow supply chain management practices and demand continued price reductions from suppliers. Especially during downturns in the memory cycle, when wafer fabs reduce capital expenditures, it is not uncommon for equipment suppliers to accept price reductions of around 10% in order to secure orders, maintain market share, and sustain production line utilization.

But now, this long-standing buyer's market "rule" is beginning to loosen.

Recently, several tier-one equipment suppliers of SK Hynix have requested price increases of 3%-4% for their deliveries. According to Korean media reports, SK Hynix has asked the relevant suppliers to submit documentation justifying the price adjustments and is currently evaluating them. This is nearly unthinkable in the semiconductor equipment sector, where barriers have traditionally been high and buyers held absolute dominance.

Behind this anomaly is an imbalance in equipment supply and demand triggered by the explosive growth in AI computing power—when a wafer fab’s expansion speed directly determines whether it can secure large AI orders from major chipmakers, “acquiring equipment” has become the most urgent arms race.

A clear example is that TCB (Thermal Compression Bonding) equipment has recently been in high demand. As SK Hynix expands its HBM4 production, two Korean TCB equipment manufacturers—Hanmi Semiconductor and Hanwha Semitech—have recently received orders of similar scale for TCB bonders. In the complex structure of AI chips, TCB equipment plays a critical role, acting like a needle threading the fabric.

In the TCB equipment market, Hanmi Semiconductor and Hanwha Semitech from South Korea, along with ASMPT, are the three major players.

Among them, Hanmi Semiconductor is currently the market leader in HBM TC Bonder equipment. According to TechInsights, as of the first three quarters of 2025, Hanmi held a 71.2% market share by revenue in the HBM TC Bonder segment, outpacing SEMES, ASMPT, Yamaha Robotics, and Hanwha Semitech. Hanmi’s advantage lies in its early partnership with SK hynix and its coverage of both NCF and MR-MUF HBM production pathways.

According to The Elec on June 10, on June 8, Hanmi Semiconductor announced it received a 44.2 billion KRW order from SK hynix for TC Bonder equipment for HBM4 production. The equipment model is the TC Bonder 4.5 Griffin, with delivery scheduled by early September. Estimating each unit at approximately 3 billion KRW, the market believes this order corresponds to about 15 units of equipment.

However, the risks to Hanmi Semiconductor are also evident, as its customers are diversifying their supplier base; SK hynix has already introduced ASMPT and Hanwha, and Micron may also bring in additional alternative suppliers.

Hanwha Semitech is transitioning from a challenger to a preferred alternative for SK hynix. Recently, Hanwha Semitech secured an order from SK hynix, supplying not only the D2W hybrid bonding cluster system but also receiving an additional HBM4 TC Bonder order from SK hynix. Thus, Hanwha is pursuing two strategic paths to compete with Kimball and AMEC: one is capturing SK hynix’s HBM4 TC Bonder orders, and the other is expanding into hybrid bonding. According to The Elec, its SHB2 Nano hybrid bonding cluster system has already entered SK hynix’s production line for quality evaluation and optimization in April.

TrendForce stated that this order is seen as alleviating market concerns over cautious capital expenditure and delayed capacity ramp-up during the transition from HBM3E to HBM4. SK hynix has placed orders with multiple TCB equipment manufacturers, clearly adopting a multi-supplier strategy: Hanmi, Hanwha, and ASMPT are all entering its TCB supply chain. As early as 2025, The Elec reported that SK hynix planned to purchase up to 80 TCB bonders that year, exceeding its initial plan of 50; Hanmi also secured orders for approximately 50 TCB bonders from Micron.

Different from the markets dominated by Hanmei and Hanhua, ASMPT does not have a high market share in HBM, but it is very strong in C2S/C2W. Its publicly disclosed orders are primarily focused on AI chip C2S and logic chip C2W, and it claims to have installed over 500 TCB units globally, with an expected TCB TAM exceeding $1 billion by 2027 and a target market share of 35% to 40%. ASMPT is more of an advanced packaging platform player than a single HBM equipment provider.

In December 2025, ASMPT received orders for 19 and then 15 C2S TCB units, respectively, from a leading OSAT partner serving the AI chip business of a top-tier wafer foundry. ASMPT states that it is the sole supplier and POR for this customer’s C2S TCB solution.

On June 8, 2026, ASMPT announced another repeat order from a leading global IDM for eight C2W TCB systems to support advanced client and data center CPU production. ASMPT highlighted that the adoption of chiplet architecture in client and data center processors is driving demand for C2W TCB.

Overall, this wave of TCB orders is essentially the convergence of three trends: HBM stacking, AI chip C2S, and logic Chiplet C2W.

The market once believed that as line width and pitch continued to shrink, more advanced hybrid bonding would replace TCB. However, it now appears that this replacement timeline has been extended.

First, at the HBM4 stage, TCB remains the more realistic path for mass production.

HBM4 requires higher stacking, greater bandwidth, and improved thermal dissipation, but hybrid bonding demands stricter requirements for surface flatness, particle control, cleanliness, and yield ramp-up. Therefore, memory and logic fabs are continuing to use TCB bonding while simultaneously preparing for hybrid bonding production lines.

Although SK hynix purchased a hybrid bonding inline system developed jointly by Applied Materials and BESI in April this year (Applied Materials acquired a 9% stake in BESI in 2025, and the two companies are collaborating on die-based hybrid bonding systems), according to The Elec, this approximately KRW 20 billion equipment order is primarily intended for R&D preparation for next-generation HBM, rather than an immediate full-scale replacement of TCB in mass production. This inline system integrates Applied Materials’ chemical mechanical polishing (CMP) and plasma processing equipment with BESI’s hybrid die bonding machine and is expected to be installed soon on a development production line. The same system is already in mass production at TSMC.

Kinex, Applied Materials’ own system, also emphasizes that hybrid bonding requires integration of modules such as wet cleaning, plasma activation, in-situ metrology, and queue time control, indicating that it is not merely a die bonder but a more complex system that bridges front-end and back-end processes.

Kinex system (image source: Applied Materials)

The semiconductor foundries' bet on hybrid bonding is also driving BESI's rapid growth. In the first quarter of 2026, BESI's orders increased by 104.5% year-over-year to €269.7 million, with Reuters attributing the growth primarily to demand for hybrid bonding, and noting that a second customer in the memory market has entered qualification for HBM-related applications.

Secondly, the relaxed standards have also extended the life of the TCB.

According to a April report by TrendForce, JEDEC is reportedly discussing relaxing the height specification for the next-generation HBM from 775 micrometers to approximately 900 micrometers, which could slow the adoption of hybrid bonding. Once the stack height limit is relaxed, manufacturers can continue using mature TCB processes to support more layers without immediately bearing the yield risks associated with hybrid bonding.

Finally, TCB devices are also being upgraded, not standing still.

For example, ASMPT recently launched its AOR TCB technology, which focuses on flux-free bonding, active oxide removal, reduced residue contamination, and improved bonding uniformity—aimed at addressing the challenges of next-generation HBM in terms of stack height, precision, and yield.

Therefore, the more reasonable industry assessment at this stage is that TCB and hybrid bonding will coexist during the HBM4/HBM4E phase; the share of hybrid bonding is likely to increase significantly only with HBM5 and higher-layer generations.

Overall, TCB is not a minor trend but a structural shift in backend equipment. According to Yole’s related report, the backend equipment market is transitioning from traditional packaging support processes to a strategic market for advanced packaging equipment; among these, TCB and hybrid bonding are the two fastest-growing segments. Yole forecasts that the TCB market will reach $936 million by 2030, with a CAGR of approximately 11.6% from 2025 to 2030; the hybrid bonding equipment market is expected to reach $397 million by 2030, with a CAGR of approximately 21.1%.

Counterpoint data also shows that AI GPUs and custom AI ASICs are driving growth in advanced manufacturing and advanced packaging; it forecasts that the industry’s advanced packaging capacity could expand by approximately 80% year-over-year by 2026, and states that advanced packaging has become a “gating factor” for AI deployment.

The AI expansion boom is not only causing wafer fabs to compete for equipment, but also straining equipment manufacturers' own supply chains due to shortages of key components such as FPGAs, CPUs, and driver ICs.

According to Elec on May 29, South Korean semiconductor test equipment manufacturers are facing the "most severe" parts shortage in history, with an ironic saying emerging in the industry: "Without semiconductors, you can't build semiconductor test equipment." The report states that the lead time for FPGAs used in test equipment operation has stretched from the previous 8–10 weeks to as long as 52 weeks; Driver ICs, which were previously available instantly through distributors, now require at least a 10-week wait; x86 CPUs and GPUs are also in short supply, with some product prices rising from around 1 million won to 3 million won—an increase of up to three times.

Because AI data centers have absorbed high-end chip production capacity, allocation priorities, and inventory buffers, test equipment manufacturers have become "downstream of the downstream," squeezed in the allocation of critical components. For example, Sourceability recently noted that FPGA lead times have extended to over 52 weeks, primarily due to demand from data centers—hyperscale cloud providers and AI infrastructure companies, with their larger orders and stronger bargaining power, have secured higher-priority supply allocations, pushing other industries relying on similar components to the back of the line. The same applies to CPUs and GPUs; although test equipment manufacturers play a vital technical role, their procurement volumes struggle to match those of cloud providers and AI server manufacturers.

The stockout logic for driver ICs differs from that of FPGAs, CPUs, and GPUs. Their shortages stem from the fact that they are niche, high-performance analog/mixed-signal devices facing increased demand from test equipment, resulting in very low supply elasticity. ADI’s website lists Automatic Test Equipment as a dedicated product line, indicating that these chips are specialized, critical components within the test equipment supply chain.

The shortage of these critical components has already impacted equipment delivery. Elec reported that a semiconductor inspection equipment manufacturer recently signed a supply contract worth over 10 billion KRW with Samsung Electronics, but has been forced to delay delivery by three months due to component shortages. The report also stated that equipment manufacturers have begun discussing order quantities and delivery timelines with customers months in advance, prior to the formal issuance of purchase orders, in order to secure components ahead of time.

Thus, in the AI era, a paradoxical chain has emerged: shortage of AI chips → wafer foundries expanding capacity → increased demand for test equipment → test equipment requiring FPGA/CPU/Driver ICs → these chips are prioritized and snapped up by AI data centers → delays in delivery of test equipment.

If the shortage of TCB and test equipment is an isolated surge at individual nodes, a broader perspective reveals that the entire semiconductor equipment industry has entered a sweeping, comprehensive upward cycle driven by the genuine strength of AI.

SEMI expects global semiconductor manufacturing equipment sales to rise from $133 billion in 2025 to $145 billion in 2026, reaching a record high of $156 billion in 2027. SEMI specifically notes that this growth is primarily driven by AI-related investments, particularly in advanced logic, memory, and advanced packaging.

SEMI also expects global 300mm wafer fab equipment spending to increase by 18% to $133 billion in 2026, followed by a further 14% increase to $151 billion in 2027, stating that AI is resetting the scale of semiconductor manufacturing investment.

This round of equipment opportunities primarily stems from three expansion initiatives:

First, leading logic foundries—TSMC, Intel, and Samsung—are expanding production for AI accelerators; TSMC forecasts that the global semiconductor market will exceed $1.5 trillion by 2030, with AI and HPC accounting for 55% of that total. Meanwhile, TSMC plans to build nine phases of wafer fabrication facilities and advanced packaging infrastructure by 2026, with 2nm and A16 production capacity expected to grow at a 70% CAGR between 2026 and 2028.

Second, in the storage sector, HBM has reignited the DRAM capacity expansion cycle. In June, SK Hynix Chairman Choi Tae-won stated in Taipei that SK Hynix plans to double its overall wafer capacity over the next five years and believes global storage supply constraints may persist until 2030. According to Counterpoint data, SK Hynix captured 58% of the global HBM market share in the first quarter of 2026. In the same period, SK Hynix reported significantly increased profits and indicated that customer demand for HBM supply over the next three years already far exceeds its current production capacity. The company also announced plans to substantially increase investments, with key focus areas including the expansion of M15X, construction of the Yongin cluster, and critical equipment.

In March this year, SK Hynix disclosed that it would purchase approximately KRW 11.95 trillion worth of EUV equipment from ASML, with the transaction to be completed by the end of 2027, for use in mass-producing new products; analysts say these tools will be deployed at the Yongin facility and the Cheongju M15X plant, supporting HBM and advanced DRAM production.

Micron stated in its earnings materials that it has increased its fiscal 2026 capital expenditure plan from $18 billion to approximately $20 billion, primarily to support HBM supply capacity and 1-gamma DRAM production, and is accelerating equipment orders and installation timelines.

Third, advanced packaging: CoWoS, C2S, and C2W are becoming bottlenecks in AI chip delivery; in the AI era, advanced packaging equipment is emerging as one of the most elastic components of this cycle. TSMC has disclosed that the compound annual growth rate of CoWoS capacity is expected to exceed 80% from 2022 to 2027, and demand for AI accelerator wafers is projected to grow 11-fold from 2022 to 2026.

Therefore, in the semiconductor equipment sector, AI computing demand is reigniting a major cycle encompassing front-end, back-end, testing, and fab equipment.

Today, leading semiconductor equipment manufacturers are no longer selling just cold machinery, precision lenses, and complex algorithms—they are selling, at their core, the most scarce resource for wafer fabs and tech giants: the ability to deliver capacity in the AI era.

In this reshuffling battle for pricing power, not all equipment manufacturers can share the spoils equally. The true winners are the absolute market leaders firmly positioned at critical process nodes—including advanced logic fabrication, HBM stacking, advanced packaging (such as CoWoS), and high-end chip testing. Armed with irreplaceable technological barriers and control over production capacity, they are reshaping the entire semiconductor industry’s landscape of value distribution like never before.

This article is from the WeChat public account "Semiconductor Industry Watch" (ID: icbank), authored by Du Qin (DQ).