Macroplane › Investment Theses › Hybrid-bonder duopoly with ASMPT

Hybrid-bonder duopoly with ASMPT

AI supply-chain thesis — mapping bottlenecks, focus companies, and supply-chain exposure for investors.

**Bottleneck theme:** Outliers **Focus:** $BESIY — BE Semiconductor Industries N.V. BE Semiconductor is one of two players (with ASMPT) in the hybrid-bonding equipment duopoly, the rate-limiting tool category for the most demanding 3D logic-on-logic (TSMC SoIC, AMD MI3xx GCD-on-IO chiplets) and ultra-high-stack HBM packaging steps. Where TCB (Kulicke & Soffa) is the volume answer for HBM3E/HBM4, hybrid bonding is the high-end answer where stack count and bandwidth-per-mm² both push pitch below 10µm. The Q1 26 order print of €269.7M was a record, signaling the long-anticipated qualification-to-volume inflection at TSMC and the leading HBM customers. The investment case is structural: as 3D-IC and ultra-high-stack HBM scale, hybrid bonding moves from "research pilot" to volume production tool, and BESI captures premium-priced equipment revenue that previously didn't exist. Applied Materials owns ~9% of BESI for a strategic reason — the AMAT/BESI integrated tool is the production answer for the most demanding nodes. The bear case is a slower hybrid-bonding ramp (TCB stays the volume choice longer than expected) and an ASMPT share gain. Pair with $KLIC (TCB), $AMAT (front-end + BESI partnership), $TOWCF (compression molding).

Focus companies in this thesis (1)

  • BE Semiconductor Industries N.V. (BESIY)

Supply-chain categories covered

  • Semiconductor Wafer Fabrication — Production of semiconductor wafers through foundry processes, foundational for chips used in AI infrastructure.
  • Die Prep — Wafer thinning, dicing, and preparation services for advanced packaging including hybrid bonding wafers.
  • Wire Bonding Equipment — Equipment for traditional wire bonding used in semiconductor packaging.
  • Advanced Packaging — 2.5D/3D packaging, CoWoS, chiplets, fan-out wafer-level packaging
  • HBM — High Bandwidth Memory — 3D-stacked DRAM (HBM2E/HBM3/HBM3E/HBM4) connected via through-silicon vias, delivering 1+ TB/s of bandwidth per stack. Co-packaged with GPUs, TPUs, and custom AI accelerators for datacenter AI training/inference and HPC workloads.
  • AI GPUs — Compute accelerators and GPUs powering AI training, inference, and large language models.
  • Hyperscalers — Major cloud operators (AWS, Azure, GCP, Meta, Oracle, Alibaba, Tencent, Baidu, Naver) and tier-2 / neocloud providers (DigitalOcean, OVHcloud, Rackspace, Kingsoft) tracked as a demand signal across multiple theses (photonics, HBM, AI accelerators, power, cooling). Excludes SaaS apps, telcos, REITs, and IT services firms.

Thesis milestones & bottleneck markers

  • HBM4 sampling — MU — First HBM4 samples requiring hybrid bonding
  • TSMC CoWoS-L production ramp — TSM
  • BESIY hybrid bonder revenue >€500M — BESIY

Browse all AI supply-chain theses · Macro trends · Industries · Product categories