PAPERS
Published research by David Borish on enterprise AI strategy, on-premises deployment economics, and AI capability growth. These frameworks are used by Fortune 500 organizations and the Open-Prem Strategy Accelerator workshop is delivered in partnership with IBM for enterprise clients.
THE OPEN-PREM INFLECTION POINT
The Open-Prem Inflection Point identifies the economic crossover point where on-premises AI deployment becomes more cost-effective than cloud computing for enterprise organizations. Created by David Borish, this framework analyzes open-source model performance, infrastructure costs, and data security requirements to help enterprise leaders make informed decisions about their AI infrastructure.
Released: April, 2025
THE OPEN-PREM INFLECTION POINT V2
Updated analysis reflecting the rapid acceleration of open-source AI models including DeepSeek V3, GLM-4, Llama 4, and Qwen3, alongside hardware developments from NVIDIA Blackwell and AMD MI350. This update recalculates the inflection point using current benchmark data and incorporates the emergence of agentic AI architectures as a new factor in the on-premises deployment case.
​
Released: October, 2025
THE OPEN-PREM INFLECTION POINT
V2 UPDATE
This V2 Update documents the moment open-source models achieved genuine frontier parity with proprietary systems. DeepSeek V3.2 matches GPT-5 performance and achieved gold medals at the International Mathematical Olympiad and International Olympiad in Informatics. GLM-4.6 is the only frontier-scale model (355B parameters) available under MIT license. Cost savings expanded to 80-90% at API level and 95%+ with self-hosting at scale.
​
Released: December, 2025
THE EXPONENTIAL REPLACEMENT CURVE
The Exponential Replacement Curve measures the rate at which AI capabilities are growing relative to human task performance. Using METR benchmarks, the framework establishes that AI capability doubles approximately every 5.5 months across three converging exponentials: model intelligence, cost reduction, and robotic capability. The resulting 27-64x annual growth rate provides enterprise leaders with a concrete timeline for workforce and operational planning.
​
Released: April, 2025
THE EXPONENTIAL REPLACEMENT CURVE V2
The Exponential Replacement Curve V2 is updated with current data showing AI systems completing 7-hour autonomous tasks, robotic labor costs reaching $5-10 per hour, and refined doubling-time calculations. This version incorporates the correlation between the framework and findings independently published by Wharton professor Ethan Mollick using METR benchmark analysis.
​
Released: June, 2025