Samsung Art Store launches exclusive Moomin collection with 60 artworks

Samsung is quietly building something that could uproot how mobile memory works. Samsung Galaxy S28 or S29, which could launch with Exynos 2800 or 2900, may be equipped with the next-gen mobile HBM technology.

Industry sources (via ETNews) claim that Samsung is developing “Multi Stacked FOWLP,” a packaging approach designed to bring server-grade HBM bandwidth to smartphones and tablets.

As mobile AI continues to evolve, seamless execution demands high specs. Faster DRAM and NAND flash memory chips land right next to the application processor.

The future of mobile AI is going to be even more spec-hungry. So instead of going aggressive with DRAM, the Korean tech giant seems to be exploring HBM-level bandwidth for future smartphones.

Mobile DRAM vs HBM

The memory inside your phone right now uses copper wire bonding, which caps out at roughly 128 to 256 I/O connections, which leaks signal, runs hot, and drinks power.

Samsung’s existing VCS technology stepped things forward by stacking DRAM dies in a staircase configuration and filling them with copper pillars instead of wires.

The new approach pushes past them by stretching those copper pillars to aspect ratios between 15:1 and 20:1, up from the current 3:1 to 5:1 range, which is a dramatic change.

Copper pillars thinner than 10 micrometers start to bend and break. That said, Samsung is wrapping the design in FOWLP (fan-out wafer-level packaging), which molds the chip and extends the wiring outward.

The mold holds the pillars upright and the result is more I/O connections in the same footprint, projected bandwidth gains of 15% to 30%, and memory stacking capacity jumping more than 1.5 times over current solutions.

An industry insider stated, “With demand for server and data center HBM still exploding, it is difficult for Samsung to devote all of its resources to mobile HBM development. The key challenge will be balancing technological maturity with the timing of mass production.”

Samsung may have the technology roadmap right. The question is whether the market will let them get there on schedule. Server demand could delay mobile HBM by a year, maybe more.

Related article: