Conference Presentation, Panel
2035 Decoded: Navigating the Decade Ahead | Dust, Glean & More | RAISE Summit 2026
- Enterprise AI adoption is projected as a 10 to 15-year journey, with the current build-out representing only 1% of total required infrastructure.
- Capital expenditure (CapEx) is forecast to escalate from an estimated $75 billion in 2025 to between $250 billion and over $300 billion by 2027.
- Agentic AI systems face significant hurdles regarding models, hardware, and the nature of living systems, while Large Language Models (LLMs) are deemed unsuitable for formal reasoning, necessitating Energy-based Models (EBMs) for fields like financial analysis and robotics.
- Self-improving systems are expected to create insurmountable competitive gaps for leading labs, with organizational compounding projected to accelerate growth at rates difficult to overestimate.
- The transformation of white-collar work is predicted to be significantly underestimated long-term, though frontier model capabilities may be overestimated in the short term, fundamentally altering the profession of programming.
- Automation is expected to replace human underwriters and claims adjusters in insurance, while repetitive word-based tasks will drive efficiency renaissances in language-based industries.
- Future hiring is projected to increase in high-leverage areas rather than reduce headcount, with job functions evolving incrementally to allow engineers to assume broader roles such as designers or product managers.
- Organizational structures are anticipated to become flatter with wider spans of control, though hierarchies will persist, while the average employee's leverage is expected to rise orders of magnitude, enabling small teams to generate massive output.
- Enterprises will face a choice between single-vendor context graphs and maintaining replaceability via standard interfaces, alongside a critical need for memory infrastructure at personal, team, departmental, and agent levels to retain compounding learnings.
- Formal verification and safety measures are expected to become the top priority within 10 years to prevent costly hallucinations in critical infrastructure like banking and autonomous vehicles.
- Risk tolerance in non-deterministic systems is predicted to allow specific players to advance rapidly, potentially generating the most significant value dispersion since the Internet, with a P&L difference between those who transition to the agentic era and those who do not.
- Market dynamics may see a "hollowing out of the middle," with concentration at the top due to data and talent access, and new opportunities at the bottom due to reduced friction, while value destruction is expected to occur first in technological transitions.
- Of the 391 non-semiconductor S&P 500 companies, approximately 250 are projected to lose value while 125 are expected to emerge as winners possessing a moat of defensible, deep context data.
- Success in the agentic, generative, and transformative (AG&T) landscape will depend on urgency, the right personnel, and context utilization, with winners needing to own specifications that constrain AI behavior and maintain human judgment.
- The market is expected to constantly shift to new innovations within a two-year cycle, meaning winning and losing will be determined by mindset and agility rather than specific technologies, with bad companies likely struggling as current market heat subsides.
- Productivity gains will be driven by executing tasks faster and cheaper, improving quality, and accomplishing previously impossible feats, with public markets continuing to reassess the value of private AI companies over time.