PD01 – Resilient Science: Sustaining Computation in a Market Optimized for AI
Event Type
Panel Discussion
TimeTuesday, June 3011:30 – 12:15 CEST
LocationPlenary Room (Bldg. 6 – 001)
DescriptionWhat happens when the world’s most powerful computers stop being built for science?
Supercomputing, once the quiet workhorse of climate models, quantum chemistry, and
astrophysics, now fuels a global race for dominance in artificial intelligence and data driven
industries. The result: a market optimized for inference speed and low precision arithmetic,
not for 64 bit calculations to understand the physical world. As compute demand explodes, so
too does competition for scarce resources—chips, memory, and even trusted supply chains.
This shift poses an uncomfortable question for researchers: how much of our digital
infrastructure do we truly own or control? Proprietary architectures and vendor locked
ecosystems are eroding the openness that once defined scientific computing. Transparency
and reproducibility—cornerstones of credible science—risk being replaced by dependency
and opacity.
This panel invites thought leaders from academia, industry, and policy to explore whether
science can remain sovereign in an increasingly commercial, constrained, and contested
computational landscape. What pathways exist for developing sovereign, open, and
interoperable computing environments that preserve transparency, reproducibility, and long
term access? Can we rebuild trust through open standards, interoperable platforms, and
transparent governance? Or will scientific computing become a tenant in someone else’s
silicon empire?
Supercomputing, once the quiet workhorse of climate models, quantum chemistry, and
astrophysics, now fuels a global race for dominance in artificial intelligence and data driven
industries. The result: a market optimized for inference speed and low precision arithmetic,
not for 64 bit calculations to understand the physical world. As compute demand explodes, so
too does competition for scarce resources—chips, memory, and even trusted supply chains.
This shift poses an uncomfortable question for researchers: how much of our digital
infrastructure do we truly own or control? Proprietary architectures and vendor locked
ecosystems are eroding the openness that once defined scientific computing. Transparency
and reproducibility—cornerstones of credible science—risk being replaced by dependency
and opacity.
This panel invites thought leaders from academia, industry, and policy to explore whether
science can remain sovereign in an increasingly commercial, constrained, and contested
computational landscape. What pathways exist for developing sovereign, open, and
interoperable computing environments that preserve transparency, reproducibility, and long
term access? Can we rebuild trust through open standards, interoperable platforms, and
transparent governance? Or will scientific computing become a tenant in someone else’s
silicon empire?
