What Are 'Token Economics' in AI?
First, let's clear something up. When we talk about token economics for AI, we are not talking about cryptocurrency speculation. The term has been co-opted from the crypto world but given a new, more practical meaning. In AI, a 'token' is the basic unit
of data a model processes—think of it as a word or a piece of a word. AI token economics, or 'tokenomics', is the financial discipline of connecting the cost of processing these tokens to the business value they create. It's essentially the unit economics for intelligence. Every time an AI generates a response, it consumes tokens, and each token has a cost. A startup with a strong grasp of its tokenomics understands its core profitability, not just its technological flash.
Why It's a Litmus Test for Viability
At an event like Web Summit, you’ll hear a lot about groundbreaking features. But many AI startups are just thin wrappers around another company's API, like OpenAI or Anthropic. Their entire business exists because an API exists, making them incredibly fragile to price changes or new competition. A company that has deeply considered its tokenomics is different. They have a plan for managing costs, scaling efficiently, and creating a sustainable business model where the margins don't shrink as they grow. Asking a founder about their token economics forces a conversation about the business itself, not just the borrowed technology. It reveals whether they've built a real company or just a cool, unprofitable experiment.
Beyond Cost: Tokens as Ecosystem Fuel
While cost management is the foundation, a more advanced form of tokenomics uses digital tokens—similar to crypto assets—to build an entire ecosystem. This is where it gets really interesting. A startup might create its own token to incentivize specific behaviors that fuel its AI. For example, users could earn tokens for contributing high-quality data, verifying model outputs, or providing specialized computing power. This approach can solve the 'cold start' problem many platforms face, creating a powerful network effect where every new user adds value to the system. This kind of model suggests a company is thinking about long-term, defensible moats, not just short-term features.
Your 4-Question Web Summit Toolkit
As you walk the floor, here are four direct questions to help you evaluate any AI startup you meet.
1. Is Your Value in the Tech or the API Call?
This is the most important question. Ask the founder: 'What have you built that I can't build myself in a weekend with an API key?'. A strong answer will focus on proprietary data, unique workflow integrations, or a deep understanding of a specific niche. A weak answer will be about the features of the underlying model they use, like its massive context window. If their competitive advantage is just 'we use the best model', they don't have an advantage.
2. How Do Your Unit Economics Improve With Scale?
Many AI businesses have unit economics that get worse as they grow, because every new query costs them money. Ask: 'What is your gross margin per query, and how will you improve it?'. Smart founders will talk about strategic caching, routing tasks to cheaper models, or batching requests. They will have a clear plan for making the service more profitable over time, not just more popular. A vague answer on unit economics is a major red flag.
3. If You Have a Token, Does It Have a Real Job?
If the startup has a crypto-style token, its purpose must be crystal clear. Ask: 'What behavior does this token incentivize, and why is it essential for the platform to function?'. Good uses include rewarding data contributions, participating in governance, or paying for services within the ecosystem. If the token's main purpose seems to be fundraising or generating speculative hype, it's likely a distraction from a weak underlying business. The token should be a gear in the machine, not just a sticker on the outside.
4. Who Is on Your Founding Team?
Building a sustainable AI company requires more than just business acumen. Look at the founding team. Is there at least one founder with deep technical experience in AI or machine learning, not just a 'full-stack developer' who can integrate an API? A team of exclusively business-focused founders planning to outsource the tech is a warning sign. Real AI companies are built by people who understand how to design the engine, not just drive the car.
