Exhibiting at SEMICON Japan 2026 (Dec 9–11)? We are now accepting advertising inquiries from exhibitors.Ads for SEMICON Japan 2026Learn more

Comparison · Packaging

Hybrid bonding vs flip chip: what is the difference?

Both methods connect a die to another die, a wafer or a substrate, and both bond the active face toward the partner. They differ in the joining mechanism, and that sets how small the pitch can go, how many connections fit per square millimetre and how much the process costs. Flip chip remains the workhorse; hybrid bonding is used where vertical bandwidth density matters most.

The short answer


The key difference is whether solder bumps are used. Flip chip connects a die face down through solder bumps, a mature process with pitches of tens of micrometres or more. Hybrid bonding has no bumps: flat copper pads and the dielectric around them are bonded directly, which allows single-digit micrometre pitch and far higher connection density, but demands front-end-class cleanliness and flatness.

Side-by-side comparison


AspectHybrid bondingFlip chip
Joining mechanismDielectric-to-dielectric bond, then anneal to fuse Cu padsSolder bumps reflowed or thermocompression-bonded to pads
BumpsNoneSolder bumps, or Cu pillars with solder caps for finer pitch
Typical pitchAround 10 µm and belowTens to over a hundred micrometres; micro-bumps reach a few tens of micrometres
Connection densityThousands to over a million per mm², depending on pitchTens to a few thousand per mm²
Gap between diesEssentially none; surfaces bonded directlyBump height gap, usually filled with underfill
Surface preparationCMP to very low roughness, precise Cu recess, particle-free surfacesStandard bumping and cleaning
ReworkNot possible after bondingLimited, but the process is well understood
Process maturityIn volume for specific 3D stacking productsMature high-volume process across many products
Typical uses3D logic stacking, cache on logic, image sensor wafer stackingProcessors and ASICs on substrates, micro-bump die stacking

How flip chip works


In flip chip assembly, bumps are formed on the die pads at the wafer stage. The die is flipped face down, aligned to matching pads on a substrate or another die, and joined by reflow or by thermocompression. Underfill is then dispensed into the gap to share mechanical stress between die and substrate.

Because the whole die area can carry connections, flip chip offers far more I/O and shorter paths than wire bonding. Moving from solder balls to copper pillars and micro-bumps has pushed pitch down to a few tens of micrometres, but solder spreading, bridging and bump height control make further shrinking increasingly difficult.

How hybrid bonding works


Both surfaces are planarised by chemical mechanical polishing so that the dielectric is extremely flat and the copper pads sit very slightly recessed. After cleaning and plasma activation, the surfaces are brought together and the dielectric layers bond at room temperature. A subsequent anneal expands the copper so that opposing pads meet and form a metal joint.

With no solder involved, pitch is limited mainly by alignment and pad design rather than by bump physics. Connection density grows with the inverse square of pitch, so going from tens of micrometres to a few micrometres multiplies the connection count by around a hundred. The catch is extreme sensitivity to particles and topography, which is why hybrid bonding uses front-end-class tools and cleanrooms.

Which to use when


Flip chip is the right choice for most processors, ASICs and network chips attached to package substrates, where its pitch is sufficient and cost and yield are well controlled. Hybrid bonding pays off where the number of vertical connections between two layers drives performance, such as stacking cache on a processor or joining logic dies in a 3D stack, as in TSMC SoIC.

HBM is a useful boundary case: current HBM stacks are joined with micro-bumps and TSVs, and hybrid bonding is being evaluated for future generations with more layers. Both methods will coexist, often in the same package.

FAQ


Why does pitch matter so much?

Connection density scales with the inverse square of pitch. Halving the pitch quadruples the number of connections per square millimetre, which directly increases die-to-die bandwidth and lets each link run slower and at lower energy.

What are wafer-to-wafer and die-to-wafer hybrid bonding?

Wafer-to-wafer bonds two full wafers and gives the best alignment and throughput but cannot sort out bad dies first. Die-to-wafer places tested individual dies onto a target wafer, which allows different die sizes and known good die selection at a lower throughput.

Is hybrid bonding the same as direct bond interconnect (DBI)?

DBI is a name used for the same basic approach of bonding copper pads and dielectric together. Hybrid bonding is the generic term used across the industry.

Does flip chip still need wire bonds or a substrate?

Flip chip replaces wire bonds for the die connection, but the die is still usually mounted on a package substrate, which connects to the circuit board through solder balls.

Terms compared


Japanese companies involved


More packaging comparisons


Sources


  1. TSMC: TSMC-SoIC
  2. Amkor Technology: Flip Chip Packaging
  3. Japan Institute of Electronics Packaging: Study of bump structures for fine-pitch flip chip assembly (Japanese) (PDF)

More on this topic


Semiconductor comparisons

Packaging terms

Last updated: November 2, 2026