Applied Materials and Intel deepen AI chipmaking R&D pact

Most AGI conversations fixate on models, but capacity is ultimately bounded by the physics and economics of chipmaking. Applied Materials and Intel teaming up at the EPIC Center to co-develop new transistors, interconnects and 3D packaging is a reminder that sustained model scaling requires breakthroughs in process technology, not just bigger GPU clusters. By co-locating Intel engineers inside Applied’s pilot fab environment, the two firms are trying to compress the cycle from lab concept to volume manufacturing for AI centric nodes.

From a competitive standpoint, this is also about catching up to TSMC and Samsung as AI demand stretches today’s roadmaps. Intel Foundry wants to be a credible supplier for AI accelerators and custom silicon; Applied wants its tools baked into those flows as early as possible. If the collaboration succeeds, it could bring denser interconnects and better thermal management for stacked AI dies to market faster, enabling higher effective compute per rack and better energy efficiency. Those are exactly the bottlenecks that appear when models push into multi-trillion parameter and agent-heavy territory. Even without headline numbers, this kind of deep process partnership tilts the medium term compute outlook slightly toward more, cheaper AI capacity.