IBM and Siemens Collaborate on Nanosheet Stacking Architecture Research, vivo Proposes Recording Agent Call Reasons

On October 15, 2026, two research updates landed within about an hour of each other: vivo's internet technology team published a proposal on Agent accountability, and IBM announced a partnership with Siemens to advance its Nanostack transistor architecture.

Earlier that day, the vivo team argued that an Agent should log not only which tools it calls but also why it calls them. A Tool call that executes successfully, they noted, does not mean the Agent has made a correct and authorized business action. The proposal introduces a reasoning object that serves as a traceable anchor between judgment and action, and assigns responsibility across Tool, Skill, Capability, and governance infrastructure layers, so the Agent can keep its dynamic coordination role without becoming the container for all accountability.

Later that day, IBM said it would work with Siemens to advance Nanostack, the stacked transistor architecture it unveiled earlier in 2026. Described as the industry's first known three-dimensional nanosheet stacking design, Nanostack breaks from the conventional approach of laying transistors flat on a two-dimensional plane and shrinking their spacing; instead, transistors are stacked vertically in a staggered arrangement, fitting more of them into the same area. In an earlier prototype demonstration, IBM packed nearly 100 billion transistors onto a fingernail-sized chip at the 0.7nm node, a density nearly twice that of the 2nm chip it announced in 2021. According to IBM's technical report, the sub-1nm chip delivers 50% higher performance than its 2nm predecessor, or 70% better energy efficiency, targeting generative AI, cloud infrastructure, and next-generation electronics.

The collaboration pairs IBM's semiconductor device innovation with Siemens' strengths in EDA, process simulation, lithography optimization, and advanced manufacturing software. The two companies will work on the design, simulation, and lithography capabilities needed for this and future generations of nanosheet stacking architectures. IBM noted that while the architecture has been shown to be feasible, it still needs further optimization before it is ready for volume manufacturing.