- The SemiAnalysis AI Tokenomics Model connects the hardware inputs of AI to the software outputs of the services built on top of AI models. The model completes our end-to-end coverage of the AI and chip industry by providing users the tools and metrics to calculate the ROI on AI spend, adoption of AI by use case, and growth of new business models enabled by AI. The Tokenomics model is the first demand-side model of AI that will allow users to translate revenue and usage growth into future AI hardware demand for Nvidia GPUs, AMD GPUs, Google TPUs, Amazon Trainium and more. Investors, corporates, policy makers and others interested in following the money trail of AI can disentangle the complex economic relationships between the various players in the AI value chain to answer the multi-trillion dollar question: What is the ROI and profitability of AI business models?
- The AI Tokenomics Model incorporates the following
- Detailed forecast of AI hardware installed base and future investments by hyperscalers (ie Microsoft, Google, Amazon, Meta, and Oracle), leading foundation labs (ie OpenAI, Anthropic, DeepSeek), and rising neoclouds (ie Coreweave, Nebius, Crusoe).
- Detailed tracking of AI compute demand sources like OpenAI, Anthropic, Thinking Machines, DeepSeek and AI compute supply sources such as Coreweave, Nebius, Crusoe, and other quality neoclouds as measured by our ClusterMAX ratings
- Bottoms-up token throughput forecasts taking into account
- Hardware system (GB200 NVL72, VR144, CPX, TPU v7, Trainium 3)
- Model architecture (GPT 5, Sonnet 4, DeepSeek V3, Kimi K2)
- User behavior/workloads (coding, chat, document analysis, agentic)
- Track the disruption of traditional SAAS business models by the rise of new token consumption companies removing the ‘seat’ from software sales, affecting companies like Salesforce, Workday, Adobe, SAP, ServiceNow, Atlassian, Microsoft, and more
- Discover emerging token-consuming startups and understand the unit economics of the new SAAS model of token consumers
- AI compute supply and demand by hyperscalers (ie Microsoft, Google, Amazon, Meta, and Oracle), leading foundation labs (ie OpenAI, Anthropic, DeepSeek), and rising neoclouds (ie Coreweave, Nebius, Crusoe).
- Addressable market analysis of the token economy consisting of
- Existing application token usage (Google AI Overviews, ChatGPT, Grok, Meta AI,
- API inference endpoints (ChatGPT, Claude, Qwen, DeepSeek, Llama, etc.)
- Token consumption software companies (Cursor, Windsurf, Harvey, Perplexity, etc)
- ROIC calculation of AI deployments by (ie Microsoft, Google, Amazon, Meta, and Oracle), leading foundation labs (ie OpenAI, Anthropic, DeepSeek), and rising neoclouds (ie Coreweave, Nebius, Crusoe).
- Aggregate AI hardware demand forecast based on usage and revenue growth
- Inference demand based on adoption
- Training demand based on model architecture development and expected ROIC
- AI-related revenue and profit forecast by business and business model
- Rental revenue
- Model revenue
- Software revenue



