Besi Hybrid Bonding Delay Turns a Memory Boom Into Share Pressure

Besi shares are under renewed pressure as a hybrid bonding delay threatens one of the chip-equipment maker’s most important growth assumptions. The surprising obstacle is not weak demand for artificial intelligence hardware. It is a memory shortage so severe that manufacturers are favoring dependable capacity over a more ambitious packaging transition.

That choice challenges a familiar investment thesis around BE Semiconductor Industries, better known as Besi. Hybrid bonding was expected to move from advanced logic chips into high-bandwidth memory, or HBM, as AI accelerators demanded taller and denser memory stacks.

The technology remains relevant, and Besi continues to win orders across several applications. However, memory producers have found ways to extend existing bonding methods while factories operate near their practical limits. Investors are now questioning when HBM will generate the volume once expected from Besi’s flagship technology.

Besi Hybrid Bonding Delay Fears Hit the Share Price

The market is repricing the timing of Besi’s opportunity, not declaring hybrid bonding technically obsolete.

Besi stock fell 5.9% to €180.75 on October 7 after UBS cut its rating from Buy to Sell. The bank also reduced its price target from €370 to €159, according to a hybrid bonding outlook published that day.

The downgrade followed a separate reduction from Bank of America Global Research. That firm moved Besi from Buy to Neutral and cut its price target to €212. Barclays also lowered its target while retaining an Equalweight rating.

These changes matter because the analysts attacked the central assumption supporting Besi’s valuation. They did not simply lower estimates for a weak quarter or a routine semiconductor downturn.

UBS estimated that the market was pricing in about €936 million of hybrid bonding revenue for Besi in 2030. Its own estimate was €424 million, less than half that implied figure.

The bank also estimated that HBM would account for about 10% of hybrid bonding tool demand in 2028. It said consensus expectations appeared closer to 50%.

Those forecasts remain analyst estimates, not confirmed orders or company guidance. However, their scale shows why a change in adoption timing can produce a sharp stock reaction.

Besi sells equipment used during semiconductor assembly, when individual dies become complete packages or multi-die systems. Its hybrid bonding machines join prepared surfaces through direct copper connections, avoiding the solder bumps used in older stacking processes.

The method supports smaller connection pitches, higher interconnect density, and thinner assemblies. Those attributes make it attractive for advanced processors, chiplets, and vertically stacked memory.

The investment case assumed these technical advantages would eventually become manufacturing requirements. HBM seemed positioned to deliver the largest equipment volumes because every memory stack contains multiple dies that must be joined precisely.

That logic has not disappeared. The timetable has weakened.

Memory producers face intense demand from AI accelerator customers. When supply is scarce, manufacturers have a strong incentive to protect output, yields, and delivery schedules. Introducing a less mature production process can threaten all three.

The result creates an unusual reversal. The AI boom is increasing the need for HBM, yet that demand is reducing manufacturers’ willingness to change how HBM gets assembled.

Besi’s problem is therefore not a lack of customer interest. It is the gap between technical evaluation and high-volume production.

The company reported repeat HBM orders from one memory customer during the first quarter of 2026. It also shipped two evaluation systems to a second memory customer.

Evaluations establish that customers are working with the equipment. They do not establish when those customers will convert tests into full production lines.

Investors once rewarded Besi for the scale of that possible conversion. They are now demanding firmer evidence about its timing.

The Memory Shortage Rewards Output Over Packaging Ambition

Tight HBM supply gives manufacturers a reason to keep proven assembly methods running for another product generation.

HBM places several DRAM dies in a vertical stack beside an accelerator or processor. This structure provides the bandwidth needed to move large quantities of data through AI workloads.

Manufacturers traditionally connect those dies using microscopic solder bumps. Thermocompression bonding applies controlled heat and pressure, while mass-reflow processes melt many connections during a production step.

Hybrid bonding removes the solder layer. It aligns copper pads and surrounding insulating surfaces before joining them directly.

The method can support more connections in less space. It can also reduce stack height and improve electrical performance. Those advantages become increasingly valuable as HBM products add layers.

However, manufacturers do not select equipment through technical specifications alone. They evaluate throughput, contamination control, defect rates, integration costs, and production yield.

Yield measures the share of manufactured devices that meet required specifications. A modest reduction can become extremely expensive when each stack contains valuable memory dies.

Memory makers currently have little room for experimentation that interrupts shipments. AI accelerator customers are competing for available HBM output, and suppliers are expanding production as quickly as their facilities permit.

UBS argued that tight supply could last through 2028. Under those conditions, accelerator customers prioritize available capacity over the maximum performance enabled by a new bonding process.

That priority changes the economics of adoption. A manufacturer can keep using a qualified process, sell nearly everything it produces, and defer the risks attached to a major production change.

A standards decision has also given existing methods additional life. JEDEC, the industry organization that defines important memory specifications, raised the permitted HBM package height from 720 to 775 microns.

The additional vertical space means manufacturers can fit taller stacks within the standard without immediately removing solder bumps. Sixteen-layer HBM4 products therefore have another path to production.

A September analysis of the HBM height change found that the decision lets manufacturers continue using microbumps for products once expected to require direct copper bonding.

The change does not make hybrid bonding less capable. It delays the moment when older methods encounter an unavoidable physical limit.

This distinction explains why the memory crunch is bad news for Besi’s adoption schedule. Scarcity raises the value of every working production line, including lines based on older packaging methods.

Manufacturers can revisit hybrid bonding when taller stacks, finer connection pitches, or thermal requirements outweigh the transition risk. Until then, supply pressure encourages caution.

The relationship between scarcity and innovation can look counterintuitive. Strong demand usually encourages investment in better technology.

Here, demand is unquestionably encouraging capital spending. Yet much of that spending is directed toward expanding proven capacity rather than changing the core assembly process.

Memory makers are still evaluating hybrid bonding equipment and building advanced packaging facilities. They simply do not need to place the technology into every near-term HBM product.

That distinction affects tool suppliers differently from memory producers. A memory company benefits immediately from selling more output at favorable market conditions.

Besi benefits when manufacturers install its equipment. A decision to extend existing processes can postpone that revenue even if final HBM demand remains exceptionally strong.

The shortage therefore separates the AI memory boom from the hybrid bonding equipment cycle. Those two trends were often treated as if they would advance together.

They no longer appear synchronized.

The Flagship Technology Is Meeting Manufacturing Reality

Hybrid bonding offers a convincing technical destination, but Besi’s valuation depends on how quickly factories travel there.

Besi entered 2026 with tangible momentum. In its first-quarter results, the company said hybrid bonding unit orders more than doubled from the fourth quarter of 2025.

Those orders exceeded the previous quarterly record in both units and value. Besi attributed most of the increase to a larger-than-expected capacity build by one customer.

The company also reported repeat orders from a memory customer and evaluation shipments to another. Overall hybrid bonding adoption reached 20 customers, according to Besi.

That customer count demonstrates broad technical engagement. It does not reveal the size, production status, or revenue potential of every installation.

A research tool, an evaluation machine, and a volume-production line serve different commercial purposes. Counting them together can make the adoption curve appear smoother than it really is.

Besi said new applications were emerging across logic, memory, co-packaged optics, and consumer products. It connected those developments with AI-related introductions expected between 2027 and 2030.

The statement supports a diversified opportunity. It also places a meaningful part of the anticipated payoff several years into the future.

This timing gap sits at the center of the current selloff. Investors had treated memory adoption as a near-term accelerator for a technology already gaining ground in logic.

Logic chips provide evidence that hybrid bonding works outside the laboratory. AMD has used direct bonding in products containing vertically stacked cache, while TSMC has deployed its SoIC packaging process.

Intel also uses direct bonding within its Foveros family. These examples show that the process can reach commercial production when product architecture and economics justify it.

Memory presents a different manufacturing problem. HBM stacks contain many repeated dies, so throughput and cumulative yield become especially important.

Every additional interface adds another opportunity for alignment or bonding defects. A process that performs well on a lower-volume logic product does not automatically meet the economics of mass-produced memory.

The older assembly route also keeps improving. Suppliers are refining thermocompression equipment, mass-reflow processes, molding materials, and stack designs.

Hybrid bonding is not competing against a frozen technology. It is competing against familiar processes that continue to gain capability.

The increased package-height allowance strengthens that competition. It lets manufacturers use physical space to postpone a more complex transition.

Besi has its own response to this reality. The company is developing TC Next, a thermocompression platform aimed at next-generation AI devices.

During the first quarter, Besi reported two new TC Next orders and said adoption had expanded to six customers. This gives the company exposure to the bridge technology that memory producers can use before hybrid bonding becomes necessary.

That diversification reduces the idea that Besi has only one path into AI packaging. It does not eliminate the valuation problem attached to delayed hybrid bonding revenue.

Traditional assembly systems generally do not carry the same growth expectations investors assigned to direct bonding. The concern is therefore about product mix as much as total equipment demand.

Besi also works with Applied Materials on Kinex, an integrated die-to-wafer hybrid bonding system. The platform combines surface preparation, bonding, and process control within a coordinated production environment.

Integration can reduce contamination risks and shorten movement between separate manufacturing steps. Those benefits address some barriers to high-volume adoption.

Yet an improved tool cannot force customers to adopt before their products require it. Equipment readiness and customer urgency remain different variables.

Competitors are also pursuing the same opportunity. ASMPT has developed hybrid bonding systems, including tools aimed at memory applications. Other partnerships combine bonding specialists with surface-preparation suppliers.

Besi retains substantial expertise in precise die placement and advanced packaging. Still, a delayed market gives rivals additional time to qualify equipment and deepen customer relationships.

Bank of America highlighted another competitive uncertainty by pointing to possible involvement from ASML. The extent and form of any future move remain unclear.

ASML is best known for lithography rather than die bonding. Its process-control capabilities and deep semiconductor relationships would still make any strategic entry closely watched.

This competitive backdrop increases the cost of waiting. Besi needs HBM adoption to arrive before a wider field captures too much of the equipment opportunity.

At the same time, moving too early would not necessarily help customers. Memory producers will adopt the process when it improves their complete manufacturing economics, not because equipment suppliers need a faster revenue ramp.

That balance defines the Besi hybrid bonding delay. The technology can remain the likely destination while the journey becomes slower and more competitive.

Why the Bearish Case Is Not Yet a Technical Verdict

The strongest skeptical argument concerns timing and valuation, while the strongest counterargument is hybrid bonding’s expanding use beyond memory.

The bearish case starts with a straightforward claim. Besi’s share price reflected a rapid HBM transition that manufacturers no longer need to make.

If HBM contributes only a small portion of tool demand through 2028, projected revenue moves further into the future. Investors must then discount those distant sales more heavily.

The UBS estimates illustrate that gap. The bank’s €424 million hybrid bonding revenue forecast for 2030 stood far below the roughly €936 million it believed the market had priced into Besi shares.

Forecasts can change, and neither number represents a disclosed customer commitment. Nevertheless, such a difference makes the stock sensitive to every production update.

The downside case grows stronger if memory customers repeatedly extend microbump-based assembly. It also strengthens if HBM4E arrives without broad hybrid bonding adoption.

Competition adds another risk. A later market provides more time for ASMPT, Hanwha Semitech, EV Group, Applied Materials, and other suppliers to improve their positions.

Besi could still participate in the eventual transition while capturing less revenue than earlier forecasts assumed. A technology can succeed without delivering the economics expected by every supplier.

There is also a customer-concentration question. Besi said one customer drove most of its first-quarter hybrid bonding order increase.

The company did not identify that customer in the release. A large capacity build proves demand, but it can also create uneven quarterly comparisons.

Investors should therefore distinguish broad evaluation activity from repeat production orders. The second signal carries more weight because it suggests customers need additional capacity for actual products.

The bullish response begins with the physical limits of existing connections. Solder bumps occupy space, add vertical height, and restrict how closely electrical connections can be placed.

Those constraints become harder to manage as designers stack more dies. They also matter when chips must move more data while controlling power and heat.

Hybrid bonding replaces the bump with direct copper connections. That architecture offers a scaling path after conventional bonding reaches practical limits.

Research firm Counterpoint described a broader memory adoption path involving SK hynix, Applied Materials, and Besi. Its analysis emphasized continuing co-development rather than abandonment.

Besi’s opportunity also extends beyond HBM. Logic chips, chiplets, image sensors, co-packaged optics, and consumer processors can all benefit from dense vertical connections.

Co-packaged optics places optical communication components near computing silicon. The design aims to reduce the power and distance involved in moving data between processors and networks.

That market could create another source of hybrid bonding demand. However, it is also developing gradually and should not be treated as an immediate replacement for delayed HBM revenue.

Logic adoption provides the clearest existing validation. Manufacturers have already used direct bonding in commercial processor designs, proving the technique can meet demanding production requirements.

The unanswered question is volume. HBM attracted investors because a single memory stack requires many repeated bonding steps, potentially multiplying equipment demand.

A diversified collection of smaller applications can support growth without matching that projected scale. Besi needs enough of those markets to convert from evaluation into repeat orders.

The company’s first-quarter performance offered reasons for confidence. It guided for second-quarter revenue growth of 30% to 40% from the first quarter, based on backlog and customer feedback.

That guidance covered Besi’s wider business, not solely hybrid bonding. It therefore cannot settle the HBM adoption debate.

The skeptical reading should remain equally disciplined. A delayed transition does not prove that memory manufacturers have rejected direct bonding.

It shows that current capacity incentives and revised package rules have weakened the near-term requirement. Product roadmaps can change again as stack heights and connection densities increase.

Investors should also avoid turning analyst downgrades into technical conclusions. Analysts evaluate the relationship among expectations, timing, valuation, and risk.

A Sell rating can coexist with a favorable long-term technology outlook. UBS described Besi as a high-quality semiconductor equipment company while challenging the revenue implied by its valuation.

That distinction is crucial. The share-price decline reflects reduced patience with the adoption schedule, not evidence that hybrid bonding has stopped working.

Three Signals Will Decide Whether Besi’s Wait Gets Longer

Orders, customer production plans, and competitive positioning will show whether the latest reset is temporary or structural.

The first signal arrives with Besi’s third-quarter results, scheduled for October 22, 2026. Investors need more than a total order number.

They should look for repeat hybrid bonding orders, the applications behind those purchases, and any distinction between evaluation and volume-production systems.

Memory orders deserve special attention. Another evaluation shipment would show continued technical interest, while a repeat capacity order would provide stronger evidence of commercialization.

Management’s language will matter as much as the headline figure. A shift in expected memory adoption from HBM4E toward HBM5 would imply a longer revenue delay.

The second signal is the packaging strategy disclosed by SK hynix, Samsung, and Micron. These companies control the production decisions that determine when hybrid bonding enters HBM at scale.

Specific evidence would include confirmed use in a shipping stack, qualification milestones, or capital spending tied to production lines. Demonstration samples alone cannot establish mass adoption.

The package-height change has given manufacturers room to extend established methods. The key question is whether upcoming HBM4E designs consume that extra room without direct bonding.

If major suppliers continue using advanced mass-reflow or thermocompression processes, the bearish timing argument strengthens. If one supplier commits a high-volume product to hybrid bonding, competitive pressure could accelerate adoption elsewhere.

The third signal is the equipment field around Besi. Investors should track production wins from competing bonding suppliers and any clearer strategic move from ASML.

Competition matters before the market reaches full scale. Tool qualifications often begin long before factories place large orders.

A customer that validates several vendors gains negotiating leverage and reduces dependence on a single supplier. Besi’s early position remains valuable, but it does not guarantee eventual market share.

Conversely, repeat orders for the Besi and Applied Materials platform would show that their integrated approach is becoming embedded in customer processes. That outcome would weaken concerns about a delayed market inviting successful challengers.

These signals should be read together. Rising orders without a named production application may still represent extended evaluation.

A product announcement without equipment details may reveal little about Besi. A technical win also matters less if the customer can meet only limited production yields.

The broader AI infrastructure cycle remains supportive. Accelerators need HBM, manufacturers need more packaging capacity, and denser chip designs continue testing existing interconnects.

However, a favorable industry direction does not settle the timing of supplier revenue. Besi’s share price demonstrates how quickly markets punish that distinction.

For semiconductor buyers, developers, and enterprise technology teams, the issue reaches beyond one European equipment stock. Packaging constraints affect accelerator availability, system costs, and the pace of AI infrastructure deployment.

The Besi hybrid bonding delay shows that component scarcity can slow a technically superior manufacturing change. When existing capacity becomes exceptionally valuable, factories optimize for dependable output first.

That tension will persist until hybrid bonding becomes necessary rather than merely attractive. Watch Besi’s repeat orders, the memory makers’ shipping roadmaps, and confirmed production wins from competing equipment vendors.

The next question is not whether direct copper bonding has technical advantages. It is whether the next HBM generation finally makes those advantages worth the manufacturing risk.