SC26’s Nov. 15–20 2026 program in Chicago puts scientific computing and AI workflows on the agenda before the Nov. 17–19 exhibits. For technology teams, the schedule creates separate windows for technical research and supplier discussions at McCormick Place.
Two calendars for computing teams
The organizer’s Nov. 5 attendee guide lists an exhibitor forum covering hardware, software and data infrastructure. The main exhibit floor is scheduled to open Tuesday through Thursday. A team attending only those days would follow a shorter calendar than colleagues participating in the full technical program.
Remote participation also requires planning around access. The Digital Experience offers livestreamed sessions and questions and answers according to registration category. Registered attendees are to receive sign-in instructions using their registration email and identification number. The guide lists Jan. 31, 2027, as the end of on-demand availability.
From physical models to infrastructure
Caltech’s Animashree Anandkumar is scheduled to deliver the keynote Tuesday, Nov. 17, from 8:30 to 10 a.m. CST. Her announced subject is AI that can simulate and understand the physical world. The abstract describes neural operators, which learn relationships between functions and work with inputs and outputs at different resolutions. Anandkumar’s abstract frames the subject as a research direction beyond language models.
The DOE booth demonstrations scheduled for Monday, Nov. 16, make the infrastructure question more concrete. Lawrence Berkeley National Laboratory’s planned presentation at 7 p.m. CST concerns machine-readable descriptions of research facilities, resources and access paths. An 8 p.m. demonstration will show scientific workloads using cloud resources through a common infrastructure interface. Another demonstration concerns preparing and assessing datasets for AI. These are announced demonstrations, not independently measured performance results.
Reliability before deployment
Argonne National Laboratory’s Nov. 9 announcement adds a reliability track to that agenda. A planned technical paper examines bottlenecks in large language model agents used for computing-system telemetry analysis. A separate workshop will consider faults, reproducibility and explainability in AI systems operating at scale.
Argonne’s booth participation also includes a presentation on its ALCF Inference Service, which the laboratory says makes models accessible across high-performance computing. That topic links the research discussion to how institutions make computing services available to users.
Argonne also lists a hands-on tutorial on building agentic workflows and discussions about the changing relationship between people and scientific computing. The planned sessions place workflow design, human collaboration and system behavior alongside model capability.
For enterprise readers, the useful distinction is between a research proposal, a demonstrated service and evidence for a deployment decision. The confirmed agenda identifies where those questions can be raised; it does not settle them in advance.
