HBM

# HBM Micro-Bump Pitch Scaling and Interconnect Density

Each die in an HBM stack connects to the die below it through thousands of tiny solder micro-bumps, and the pitch — the center-to-center spacing between adjacent bumps — directly caps how many TSV connections can physically fit across a given die area, making bump pitch one of the fundamental limits on how wide an HBM interface can be. A micro-bump is a small solder ball that mechanically and electrically joins a TSV on one die to a matching pad on the die below it. Because every TSV needs its own bump, and every bump needs enough surrounding space to avoid shorting into its neighbors during the bonding process, the pitch between bumps sets a hard geometric ceiling on how many TSV connections can be packed into the die's bonding area — which in turn caps the total interface width the stack can offer the host.

Shrinking bump pitch directly increases how many parallel data and power/ground connections fit in the same die area, which is exactly why successive HBM generations have paired wider interfaces with progressively tighter bump pitch rather than growing die area to compensate. Each new HBM generation has widened its per-channel interface and added more independent channels, and doing that within roughly similar die sizes has required shrinking micro-bump pitch generation over generation — more bumps in the same area means proportionally more TSV connections, which is the direct enabler of a wider aggregate interface and therefore higher total bandwidth per stack.

Shrinking bump pitch doesn't come free — tighter spacing raises the stakes on bonding alignment precision, since a misaligned bump has proportionally less margin before it bridges into an adjacent connection, and it tightens the tolerance on exactly the kind of coplanarity and warpage issues that multi-die stacking already has to manage. As pitch tightens, the bonding process has less room for error before two adjacent bumps that were supposed to stay electrically separate actually touch, turning what used to be a comfortable alignment tolerance into a much more demanding one. This is the direct manufacturing-difficulty cost that comes bundled with every generation's pitch reduction, and it's part of why bump-pitch scaling and bonding-process capability have had to advance together rather than pitch simply shrinking on its own schedule.

Spare-TSV redundancy exists to repair individual defective vias after the fact; micro-bump pitch is instead about how many vias can physically exist in the first place — tighter pitch gives a stack more raw interconnect density to work with, but also makes every one of those connections a little more fragile, which is exactly the kind of yield pressure spare-via redundancy schemes are built to absorb.