Backside power delivery and double-sided wiring solve one problem and create another. Once you route signals on the front and power on the back, you have two active surfaces — and two surfaces means new paths for electrostatic discharge, substrate noise, and edge cracking that a single-sided die never faced. The patents addressing those new failure modes are where the interesting mid-2026 claims sit, and IBM's grant is a clean specimen for construction.
International Business Machines Corporation's US12653030B2, "Double-sided integrated circuit with electrostatic guard ring" (issued 2026-06-09), looks at first like a broad "double-sided IC" claim. It is not. Read the classification list: H01L 23/585 is specifically a guard-ring structure; 23/562 and 23/564 are crack/stress protection; 23/481 is a through-via; 24/05 and 24/08 are bonding pads/structures. The center of gravity is the guard ring. The "double-sided" language sets the context; the electrostatic guard ring is the limitation the claim is built around.
Construe it accordingly. A guard ring is a conductive ring around active circuitry that intercepts stray charge and isolates it. In a double-sided IC, the question the independent claim must answer is: a guard ring positioned where, relative to the two wiring levels and the through-vias that connect them? The protected scope is almost certainly a guard-ring structure with a defined relationship to the backside power network and the inter-level vias — not any guard ring, and not any double-sided chip. Strip the guard-ring limitation and you have an unpatentable generality; it is the recited element that earns the grant.
The design-around space, therefore, is in the guard-ring geometry and connectivity. A competitor building a double-sided IC does not infringe merely by having two wired faces — that is the era, not the invention. They would touch this claim only if their electrostatic protection ring has the specific structure and placement the independent claim recites. Move the ring's connection to the backside network, or protect against discharge by a different structure (e.g., distributed clamps rather than a ring), and the claim is avoided. That is the lever.
Why this matters beyond one patent: it is a template for reading the whole backside-power patent wave. The marketing category — "double-sided," "backside power" — is never the claim. The claim is some specific structure that the double-sided architecture newly requires: a backside contact (covered in our claim-construction read of the IBM and TSMC contact patents), a power bridge, or here, an electrostatic guard ring. Construe the specific structure; ignore the category.
The takeaway for a freedom-to-operate review on a double-sided design: do not be reassured that "everyone is going double-sided." The architecture is open; the protection structures it forces — guard rings, contacts, bridges — are individually claimed. IBM's US12653030B2 fences one of them, and clearing it means construing the guard-ring limitation against your specific ESD-protection layout, element by element.