Every chiplet design rests on a die-to-die (D2D) interconnect: the short, dense, high-speed link that lets two separately manufactured dies behave like one chip. The UCIe standard defines the physical and protocol layer for that link, which means the link itself is increasingly commodity. So where is the defensible IP? The June 2026 record points somewhere telling — not to the foundries, but to interconnect specialists fencing the functions that sit above and beside the standardized link.
Eliyan Corp.'s US12650936B1, "Chiplet-based multi-chip module with die-to-die (D2D) interface that employs bandwidth balancing circuitry" (issued 2026-06-09), is a B1 grant — note the kind code: a patent issued without any pre-grant publication, which often signals a company that wanted the application kept dark until grant. Its classifications are pure G06F 13 (bus/interface control: 13/1684, 13/1668, 13/1678), not packaging classes. That is the tell — Eliyan is not claiming a structure, it is claiming the circuitry that balances bandwidth across the D2D interface. The limitation that matters is "bandwidth balancing," and it reads on the controller logic, not the wires.
ProteanTecs Ltd.'s US12644924B2, "Die-to-die connectivity monitoring" (issued 2026-06-02), fences a different adjacent function: telemetry. Classified in G01R 31/3016 / 31/3173 (testing of integrated circuits) and H10W 70/65 (interconnect), it claims monitoring the health of the D2D links in the field. For chiplet products — where a single bad link can fail an expensive multi-die module — in-field connectivity monitoring is a genuine moat, and it is orthogonal to whoever owns the link itself.
Map this and the picture inverts the usual assumption. People expect the D2D thicket to be owned by TSMC, Intel, or the UCIe consortium members. But a standardized physical link is, by design, hard to monopolize — the standard exists precisely so multiple vendors can implement it. The patentable value migrates to the functions the standard does not specify: how you arbitrate and balance bandwidth across the link (Eliyan), and how you monitor its integrity over the product's life (ProteanTecs).
The white space follows from that logic. UCIe defines the lane; it does not define adaptive power management of the lane, security/authentication across the lane, or workload-aware routing among many lanes. Those are the next cells a CI team should expect to see filled, and an entrant who reads only the foundry portfolios will miss where the contested IP actually is.
For a portfolio strategist, the lesson generalizes: when a physical interface gets standardized, do not map the thicket on the interface — map it on the functions stacked on top. The June 2026 D2D grants are a clean case study, with two specialists already holding the bandwidth-balancing and monitoring corners.