Applied Materials introduces two chipmaking systems

Applied Materials (AMAT) introduced two chipmaking systems designed to create the smallest features in the world’s most advanced logic chips. The company said, “By controlling materials deposition with atomic-level precision, the technologies enable chipmakers to build faster and more power-efficient transistors at the scale required to sustain the pace of today’s global AI infrastructure buildout. Driven by surging demand for AI compute, the semiconductor industry is pushing the limits of scaling to squeeze more energy-efficient performance from each of the hundreds of billions of transistors in a processor chip. To address this challenge, the world’s leading logic chipmakers are introducing new Gate-All-Around transistors at 2nm and beyond. The GAA transition enables much higher performance at the same power, but achieving these gains comes with dramatically higher process complexity. Building the complicated 3D structures inside a GAA transistor takes more than 500 process steps, many of which require entirely new ways of depositing materials with precision, repeatability and control – all within tolerances approaching the size of individual atoms. Applied today unveiled two chipmaking systems that leverage material innovations to create some of the most complex features associated with GAA transistors. The new technologies enable deposition of metals and insulating dielectrics – essential materials that dramatically impact the performance and power efficiency of advanced chips… Next-generation AI GPUs now in development are expected to pack more than 300 billion transistors into a space the size of a postage stamp. Without proper isolation, electrons can easily diffuse into neighboring transistors, leading to parasitic capacitance, an unintended electrical drag between transistors that slows signals, wastes power and negatively impacts a chip’s performance-per-watt… Each GAA transistor is a switch controlled by a gate stack composed of multiple layers of metal that determine the threshold voltage needed to turn the transistor on and off. To meet the unique needs of different AI workloads, from the data center to the edge, chipmakers provide designers with a range of transistor options, with some tuned to switch faster for peak performance and others tuned to switch using the lowest amount of power. Meeting these trade-offs comes down to metal gate stack optimization based on high-precision metal deposition.”