How AI Will Reshape Computer Systems by 2035: A Jeffersonian Dinner in San Francisco About Our 10,000x Future

Key Takeaways

• AI inference projected to grow 10,000× by 2035.
• 50× algorithmic gains expected from transformer advances and data efficiency.
• 50× hardware improvements driven by specialization and AI‑automated design.
• Solar and battery costs falling, enabling greener data center power.
• Human‑interpretable abstractions needed for AI oversight and guidance.

Pulse Analysis

The CRA‑Industry Salon in San Francisco served as a rare think‑tank where academia and industry converged to forecast the next thirty years of computing. Attendees argued that the AI revolution will dwarf previous paradigm shifts—microprocessors, the internet, smartphones—by delivering an order of magnitude more intelligence per watt. By 2035, they anticipate a 10,000× increase in AI inference, a figure derived from multiplicative progress: 50× improvements in algorithms, another 50× in specialized hardware, and a 4× expansion of data‑center capacity. This aggressive trajectory forces a re‑examination of core performance metrics, pushing designers to prioritize intelligence‑per‑energy and latency‑bounded token throughput.

Technical drivers underpinning the forecast are rooted in both software and silicon breakthroughs. Transformer‑based models continue to unlock efficiency gains, while research into data‑efficient learning promises another 50× leap, echoing human learning’s low data requirements. Simultaneously, AI‑assisted hardware design promises rapid, cost‑effective specialization, potentially automating large portions of the design flow. The panel stressed that abstractions will remain essential, but future interfaces may evolve beyond human comprehension, requiring new oversight mechanisms that retain human‑interpretable layers for safety and governance.

Beyond the chips, the conversation turned to energy and societal impact. Falling solar and battery prices make renewable‑powered data centers viable, reducing the carbon footprint of massive compute growth. Yet the rapid AI expansion raises policy and workforce challenges, from misinformation to job displacement. Participants called for proactive education, lifelong learning pathways, and informed regulation to harness AI’s benefits while mitigating risks. The insights from this three‑hour dialogue are poised to shape CRA’s roadmap and guide industry leaders as they navigate the coming AI‑driven computing era.