SC25 Insights: Trends Impacting Technology Investments in 2026
What engineering and R&D technology leaders need to know from SC25 to stay competitive
Last week’s SC25 conference in St. Louis brought together 16,500 attendees and a record 560 exhibitors representing the leading edge of computational engineering and scientific discovery. While the supercomputing community includes academics, government researchers, and cloud providers, the most compelling insights from this year’s conference speak directly to product development organizations racing to bring innovations to market faster than ever before.
If your company develops physical products–whether you’re in automotive, aerospace, life sciences, manufacturing, or energy–the trends and announcements from SC25 offer a clear roadmap for the technology investments that will separate industry leaders from followers in 2026 and beyond. This isn’t about theoretical research or future possibilities. The conversations at SC25 centered on production-ready capabilities that engineering and R&D organizations are deploying right now to compress development cycles, reduce costs, and unlock innovations that competitors can’t match.
Here are the actionable trends from SC25 to factor into your 2026 planning.
Where Innovation-Driven Companies Are Investing
Hyperion Research’s market analysis revealed that the HPC+AI sector grew more than 20% in the first half of 2025 and 23% in 2024, triple the historic growth rate. Total sector spending reached $60 billion, projected to nearly double to $120 billion by 2029.
This isn’t speculative investment. This growth is driven by engineering and R&D organizations that have deployed AI-accelerated workflows, measured the results, and determined the ROI justifies scaling these capabilities across their entire product development process. The growth is fueled by new technologies and use cases expanding rapidly, governments making major AI investments, and AI infrastructure being measured in gigawatts rather than megawatts.
Your industry peers and competitors are likely already investing in these capabilities or planning to 2026. The question isn’t whether to invest in computational acceleration and AI-driven simulation, but where to start and how quickly you can deploy capabilities that deliver measurable productivity gains. Organizations at SC25 discussed reducing product development cycles from months to weeks, cutting simulation time from days to hours, and reclaiming significant engineering time previously lost to manual workflows.
As you evaluate these capabilities, focus on three main factors: quantifiable engineering productivity gains, operational simplicity in deployment and management, and clear economics that demonstrate positive ROI within your planning horizon.
AI Physics Methods Move From Research to Production
AI for specialized engineering applications is moving rapidly from research to production. There was continued discussion throughout the conference about the role of AI methods like surrogate models and reduced order modeling, and how they can be used in tandem with traditional modeling and simulation methods. A general consensus emerged that AI physics methods are highly effective at exploring large design spaces early in the R&D process as a complement to traditional methods, which can then be used for validation.
Rescale’s presentation at the Microsoft theater highlighted real-world applications of these approaches, showcasing an aerodynamics optimization use case for General Motors Motorsports that demonstrates how AI physics methods integrate with traditional simulation workflows to accelerate product development timelines.
The competitive advantage will come from organizations that can rapidly train proprietary models on their unique engineering data and methods. Generic foundation models provide a starting point, but models trained on your specific thermal characteristics, aerodynamic trade-offs, or structural behaviors will deliver insights competitors can’t replicate. Start identifying which simulation domains would benefit most from AI acceleration: thermal analyses you run thousands of times per program, structural simulations with long solve times, or fluid dynamics studies exploring large parameter spaces.
NVIDIA’s SC25 announcements showcased the rise of AI methods for engineering and scientific exploration, highlighting Apollo, a family of open AI physics models covering semiconductor design, fluid dynamics, structural mechanics, and weather modeling. Companies like Applied Materials, Cadence, LAM Research, Siemens, and Synopsys are already using Apollo to accelerate design and simulation workflows. Apollo will be available in 2026 on Rescale along with additional toolkits from NVIDIA as well as other third-party model architectures, providing access to diverse AI simulation prediction methods.
AI and Agents for Engineers Promise to Supercharge Productivity
Agentic AI and autonomous workflows emerged as a primary theme throughout SC25, with widespread discussion of systems that can independently orchestrate complex engineering tasks, test hypotheses, and optimize workflows with human-in-the-loop oversight rather than constant manual intervention.
However, among practitioners, the general consensus is that agentic automation requires solid foundations of data and process understanding. Organizations can’t deploy intelligent agents to automate workflows if the underlying data is fragmented, inconsistent, or inaccessible. The most successful implementations start with capturing workflows at the point of execution as engineers work, facilitating user-led process mapping and data standardization without requiring heavy upfront governance frameworks.
The vision is compelling: simulation agents that automatically launch analyses when parameters change, orchestrate compute workloads across architectures while balancing cost and performance, generate automated reports, and sync validated results back to PLM systems, all without manual intervention. But getting there requires first solving the data intelligence challenge: automatically capturing metadata as simulations run, establishing traceable digital threads, and enabling AI-powered search across engineering knowledge.
The path to agentic automation isn’t an overnight transformation. It starts with building the data foundations that make automation possible: unified data fabrics that capture simulation metadata automatically, AI assistants that provide instant access to cross-disciplinary analysis through natural language queries, and platforms that can templatize common workflows for reuse across teams.
Public Sector Signals Strategic Computing Investments
SC25 was well-attended by government agencies and public sector organizations, alongside the private sector contractors who support national R&D initiatives. Their attendance and the timing of additional recent announcements further validates the public sector’s commitment to investing in computational capabilities as essential infrastructure for innovation.
The White House announced the Genesis Mission executive order, representing a historic national effort to invest in high performance computing resources and build an integrated AI platform that harnesses federal scientific datasets to train foundation models and create AI agents for accelerated discovery. Amazon also announced an investment of up to $50 billion to expand AI and supercomputing capabilities for U.S. government customers, adding nearly 1.3 gigawatts of capacity across secure AWS regions and enabling agencies to achieve in hours what once took weeks.
Notably, the public sector continues to invest heavily in cloud-based solutions for the flexibility and agility they provide. This should serve as a strong signal for private sector organizations that have been slower to adopt cloud infrastructure. Government agencies, traditionally known for strict security requirements and on-premises, increasingly embracing cloud at this scale validates the maturity, security, and strategic value of cloud deployment models for mission-critical R&D workloads.
The public sector is signaling where critical capabilities are heading. The emphasis on integrated platforms that unify compute, data, and AI (not patchwork of point solutions), validates the same challenges facing private sector R&D. For the established contractors and startups serving government projects or competing in domains like energy and defense, these announcements highlight the technology capabilities that will be expected in future programs.
Cloud Considerations for Engineering Organizations
Cloud adoption for engineering workloads continues accelerating, with a critical insight emerging from SC25: the most sophisticated organizations aren’t choosing between cloud and on-premises. They’re strategically leveraging both.
Customer feedback highlighted key cloud advantages: lower costs when used strategically, better on-premises integration, sovereignty controls, tighter HPC+AI integration, and access to specialized hardware. The rise of “neoclouds” like CoreWeave, Nebius, and Lambda reflects the expanding ecosystem.
The strategic question isn’t “cloud or on-premises” but rather “which workloads gain the most from cloud capabilities?” Cloud excels at handling peak demand periods requiring 10x normal capacity, access to latest GPU architectures without capital investment, collaborative projects involving multiple sites, and rapid scaling for new programs. On-premises remains optimal for steady-state baseline workloads, applications with specific security requirements, and situations where hardware investments are already amortized.
The critical success factor is maintaining unified engineering capabilities regardless of where compute happens. Engineers should have consistent access to results, unified search capabilities, and seamless collaboration whether simulations run on-premises or in cloud. Cloud also provides continuous technology refresh through access to infrastructure advancements without the long lead times and overhead required to deploy and maintain them on-premises.
Recommendations for Engineering Digital Transformation Initiatives
Based on SC25 insights, here are steps for engineering and R&D leaders to consider when planning 2026 technology investments:
1. Develop a workload-specific cloud strategy. Which engineering activities are capacity-constrained or require specialized hardware? Consider project backlogs, time-sensitive and high-priority programs, workloads that would benefit from newer hardware only available in cloud, and peak capacity requirements. Start with these cloud candidates rather than attempting wholesale migration.
2. Identify AI and agentic opportunities with a phased approach. Start with quick wins: AI-powered search to find previous simulation work, automated job summaries, and natural language queries. Build toward advanced capabilities: AI physics models for high-volume domains, workflow automation for repetitive tasks, and eventually autonomous agents orchestrating complex multi-step processes.
3. Assess data foundations. Can engineers easily find and reuse previous simulation work? Do you capture metadata automatically or depend on manual processes? Is institutional knowledge at risk when engineers work in silos or leave the company?
4. Evaluate integration strategy. Are you planning to connect disparate systems that will require complex integrations? Would integrated platforms reduce administrative burden while improving capabilities?
5. Plan for adoption. Look for solutions that fit naturally into existing workflows rather than requiring extensive training. Consider how to deliver new and specialized capabilities to broader user communities within your organization, enabling citizen users to leverage advanced simulation and AI methods.
Not Sure Where to Get Started? Build a Foundation for Next-Gen Use Cases
The convergence of trends visible at SC25 (unified platforms, AI-driven automation, intelligent data management, hybrid cloud strategies) represents where product development is heading in 2026 and beyond. One critical capability threading through these themes is data intelligence: the ability to automatically capture and act on simulation data at scale with the goal of building a data pipeline for AI training.
We’re hosting an in-depth product demo on December 10th showcasing Rescale Data Intelligence capabilities purpose-built for computational engineering and scientific R&D. Learn more about the new product and reserve your spot for the live demo.
Companies tracking these trends and taking action now will lead their industries in 2026 and beyond!
