Conference Presentation, Panel, Fireside Chat
Infrastructure as Destiny: The Compute-Capital-Cloud Trinity | RAISE Summit 2026
RAISE SummitStephanie Cohen, Don Barneston, Suresh Vasudevan, Greg Matson, Jeff Denworth, Jordan Nanos
- Demand for managed GPU and accelerator services has reached unprecedented levels with no cloud encountering off-take problems, a dynamic shift occurring over the last six months where demand exceeds even speculative investment plans.
- An endless supply of capital, including both investment-grade and non-investment-grade debt, remains available in the market with no signs of the current pace slowing down.
- Supply chain lead times for components such as those supplied by Credo have extended to years, prompting commitments to calendar year 2028 for customers deploying GPUs today.
- Capital expenditure for Cloudflare is planned to remain at 15% or less of revenue despite rapid growth, while planning horizons have extended to months rather than years due to the speed of global change.
- Customers are increasingly willing to sign multi-year supply agreements, including four-year contracts representing tens of billions of dollars annually, reflecting a new turn in demand shaping.
- Capacity for semiconductor fabs is expected to be underwritten until 2028, a situation never before seen, though funding models are anticipated to eventually align with profitable underlying business operations.
- Cloudflare predicts exponential growth in non-human traffic, which currently exceeds 50% of network volume, driven by an agentic internet where transactions between machines become opportunities for commerce.
- The industry is expected to mature on resiliency with defined levels, utilizing Service Level Agreements (SLAs) to ensure operating cadence and tune infrastructure to reduce error rates by three or four orders of magnitude.
- Vast's compression technology is projected to represent a significant portion of the world's largest storage deployments, potentially doubling or tripling existing hardware capacity to alleviate industry pressure.
- Infrastructure build-out for accelerated compute is anticipated to continue for several decades, with no expectation of the current capacity versus demand gap ending within the next five years, though the frantic scramble for capacity may diminish.
- A blurring of lines between vendors and cloud providers is expected, alongside consolidation among vendors and stabilization of the provider landscape over the next few years.
- Inference is predicted to become a mainstream cloud utility embedded in every application, with isolates where agents share runtime becoming the standard computing model.
- Major technology upgrades are expected every two years to meet customer requirements for bandwidth and scalability, while distributed systems may eventually evolve for operation in space or underwater.
- Socioeconomic questions regarding the impact of AI on employment and social unrest are expected to be resolved in a positive manner four to five years from now.
- The number of sovereign AI and neocloud providers is expected to grow, though not to the extent of hundreds of data center operators, as the industry focuses on operating assets profitably and efficiently under current capital constraints.