Interview, Fireside Chat
Fast by Design: The Infrastructure Powering the Agentic Era | Rodrigo Liang | RAISE Summit 2026
- Speed for executing large models without accuracy loss is projected to become a critical priority, particularly as latency threatens to limit the effectiveness of stacked agents.
- The enterprise market is anticipated to shift toward specific intelligence for defined workflows, utilizing smaller language models and open-weight alternatives rather than relying solely on general intelligence.
- Future deployments are expected to favor heterogeneous and hybrid architectures, including on-prem air-gapped appliances, Neo Clouds, Sovereign Clouds, and cloud workloads to address privacy, security, and agent certifiability.
- Market dynamics are predicted to avoid a winner-take-all outcome, with enterprises seeking diverse options that combine frontier-level intelligence for complex tasks and open models for simpler use cases.
- Organizations plan to distribute systems through networks of over 90 enterprise software companies leveraging thousands of sales personnel, emphasizing flexible models and choice behind firewalls.
- Open source is expected to prevail over the long term due to cost and flexibility advantages, though this trajectory remains subject to short-term geopolitical constraints and issues.
- Model architectures are projected to continue expanding in size and becoming more distributed as they self-learn, while the company aims to deliver real-time performance for big models without directly competing against NVIDIA.
- Capitalization efforts will rely on strategic partnerships with major supporters like Intel to secure large financial backstops, while solving complex enterprise sales requirements is intended to enable service for any AI lab.