Integrated high-power package with matched TIM-1, TIM-2, and lid-attach thermal stack

The core claim here is a package-level system: an ordered, cooperating thermal stack that spans the full heat path from die junction to heat sink — zone-modulated TIM-1, low-bond-line aligned-filler TIM-2, an inorganic lid-attach layer, and a soft in-stack TIM-1.5 — captured as a single system claim. The strategic value is architectural rather than material: a buyer acquires rights to the entire cooperating arrangement in one instrument, rather than assembling rights layer by layer from separate material claims. The timing argument is concrete. High-TDP AI accelerators and HBM stacks are pushing package thermal budgets past what any single-layer point fix can address. Glass-core substrates and co-packaged optics add further stress at the bondline. These forces are converging on full-stack thermal management as a design requirement, not a nice-to-have — and the architecture claimed here is designed precisely for that convergence. The broader portfolio of high-power thermal-interface materials supplies the underlying constituent technologies. Each layer in this integrated package corresponds to a separately validated material family within that portfolio. The system claim functions as the portfolio's capture-everything backstop: a competitor who avoids any individual constituent material still reads on this claim if they assemble a qualifying multilayer stack.

This invention is not a new material; it is a package architecture defined by the ordered arrangement and cooperative function of its constituent layers. The stack assembles zone-modulated TIM-1 at the die-to-lid interface (managing peak-temperature spreading across a non-uniform die surface), a low-bond-line-thickness aligned-filler TIM-2 at the lid-to-cold-plate interface (providing pump-out resistance under liquid cooling pressure cycles), an inorganic lid-attach bondline that eliminates organic pump-out failure modes, and a soft compliant in-stack TIM-1.5 based on MgSiN2 with elastic modulus at or below 500 kPa. That last layer is mechanically critical: it absorbs differential thermal expansion between stacked die without transmitting damaging stress to fragile interconnects. An optional hexagonal boron nitride interposer sub-stack provides lateral heat spreading for packages where the die footprint is smaller than the heat-removal structure. Each layer is matched to its specific thermal and mechanical role, and the system claim captures the cooperative behavior that emerges from their combination — behavior no single-layer approach can replicate. The package-level reliability targets are drawn from industry-standard qualification protocols: JEDEC JESD22-A104/B116 thermal cycling, highly accelerated stress testing, and creep evaluation at the bondline. These are the qualification gates that determine whether a thermal stack survives the service life of a deployed accelerator package. Because this is a system architecture, conventional single-crystal phonon stability analysis does not apply; there is no periodic lattice to evaluate. Validation is instead conducted at the system level, with the constituent material families carrying their own individual computational evidence. A combined simulation of the hexagonal boron nitride interposer sub-stack with the zone-modulated TIM layer was performed as a prophetic example, modeling the coupled thermal and mechanical response of that sub-assembly. The constituent materials — zone-modulated TIM-1, aligned-filler TIM-2, MgSiN2 in-stack layer — have been evaluated within the broader high-power thermal-interface materials portfolio using multiple independent machine-learning interatomic potentials and density functional theory, with phonon consensus required before any material advances. The system claim inherits that evidentiary base.

The addressable market is estimated at $10 billion or more, spanning the full TIM-stack opportunity across AI accelerator packaging, high-bandwidth memory modules, co-packaged optics, and high-voltage power electronics. That estimate reflects the broadest scope in the high-power thermal-interface materials portfolio, because the system claim covers every thermal interface layer in a qualifying package simultaneously. The natural customers are full-package integrators: NVIDIA, AMD, Samsung's HBM business, hyperscaler co-packaged optics programs, and outsourced semiconductor assembly and test providers. These are the entities that actually assemble complete packages and therefore make the purchasing or licensing decision over the full stack. They also have the strongest economic motive for a single comprehensive grant — the alternative is assembling rights from multiple separate material claims across multiple transactions, with residual risk of gaps between layers. Royalty logic favors a per-package rate applied to the assembled article. A bundled rate capturing TIM-1, TIM-2, lid-attach, and in-stack TIM-1.5 in a single license simplifies negotiation, eliminates leakage from unlicensed layers, and converts the portfolio's constituent material rights into an enterprise-grade commercial instrument. At even a modest per-package royalty, aggregate value scales directly with the accelerator and HBM package volume that this thermal architecture is designed to serve.

single ordered article spanning TIM-1/TIM-2/lid-attach/in-stack for buyer-facing licensing

claimed by ordered configuration + cooperating-layer limitations; modulus<=500 kPa in-stack sub-claim adopts Samsung modulus limit as dependent only

The strongest acquisition candidates are the full-package integrators who actually assemble the complete ordered stack: NVIDIA, AMD, Samsung's HBM division, hyperscaler co-packaged optics programs, and major outsourced semiconductor assembly and test providers. Only these entities build the complete package article the system claim covers, so only they face direct infringement exposure — and only they capture the full licensing value of a single comprehensive grant spanning all TIM layers. For a strategic accelerator leader moving into glass-core or advanced 3D-stacked package architectures, outright acquisition or an exclusive license locks up the full-stack thermal IP position for the next platform generation and creates a licensing asset over competitors who build similar stacks. For an OSAT or HBM producer, a broad non-exclusive license covering all assembly flows eliminates cross-layer licensing risk at a single transaction cost. Because the system claim functions as the portfolio's commercial anchor — converting constituent material rights into an enterprise instrument — it should be priced and structured accordingly, with constituent family claims bundled rather than sold separately.

The primary risk is the absence of measured integrated results. The cooperating-stack reliability claim against JEDEC, HAST, and creep endpoints is supported today by a prophetic combined-stack simulation and module-level constituent recipes, not a tested package. The system claim's strength in a licensing or litigation context depends on that gap being closed. Until a full integrated test vehicle passes qualification, the specification's reliability assertions remain modeling-based. A secondary risk is inheritance from the constituent families. Each underlying material family carries its own modeling uncertainties — modeled thermal conductivity for zone-modulated TIM-1, modeled anti-pump-out performance for aligned-filler TIM-2, modeled effective thermal conductivity for cubic boron arsenide-loaded formulations. The system claim's credibility is bounded by the weakest link in those constituent proofs. Third, the design-around risk of a competitor omitting one cooperating layer to step outside the wrapper is real and requires robust claim construction to close. The roadmap to de-risk all three issues converges on the same action: fund and run the integrated package test vehicle. That single experiment is the gating milestone, and deal terms for any enterprise license should be structured to reflect pre- versus post-qualification value.