1、Accelerating aFrugal AIEcosystemWhite Paper|November 2025 CONTENTS Executive Summary-61 Introduction-72 Why Is the AI Compute Stack Expensive and Often Not Profitable-8 2.1 Massive Compute and Infrastructure Expenditure 2.2 Hardware Limitations and Vendor Lock-in 2.3 Software Complexity and Maintena

2、nce 2.4 Token-Based Pricing and Output Costs 2.5 Model Size and Compute Requirements 2.6 Cloud Compute Pricing Schemes 2.7 Bandwidth and Storage Overheads 2.8 Unsustainable Burn Rate and Lack of Optimisation3 Key AI Startup Failure Drivers-124 The Frugal AI Technical Stack:Architecture for Cost-Effi

3、cient,Scalable AI-13 4.1 The Frugal AI Stack 4.2 Quantitative Comparison between Traditional and Frugal AI Stacks 4.2.1 Average Cost Savings 4.2.2 Energy Footprint 4.2.3 Scalability Metrics5 Security Implications Across the Frugal AI Stack-22 6 The Socio-Economic Stack for Frugal AI-25 7 Sustainabil

4、ity-298 Key Startups and Platforms in the Frugal AI Stack/Examples-319 The Frugal AI Ecosystem and Observatory-3510 Further Research-3711 Conclusion-38 References-40|5 The next wave of innovation will not bemeasured by model size or compute power,but by reach,responsibility,and resilience.Frugal AI

5、builds on this idea.It is a mindsetthat transforms scarcity into strength andcomplexity into clarity.By aligning efficiency with inclusion andsustainability,Frugal AI offers a blueprint forprogress that endures.4|FRUGAL AI HUB|White Paper|Nov 2025|7 Serish Venkata GandikotaCo-Founder&Co-Director,Fru

6、gal AI Hub|Visiting Fellow,Cambridge JudgeBusiness SchoolSerish is an innovation strategist,impact&climate fund adviser,entrepreneur,andresearcher focused on frugal innovation,sustainability,impact investing,andinclusive technology.As Co-Director of the Frugal AI Hub at Cambridge JudgeBusiness Schoo