Last week I attended the first Tokenomicon in Amsterdam. While there’s a great tradition of hosting a FinOps X conference in Amsterdam for the past few years, the focus on AI made this one something I was especially looking forward to. For myself, but also for you if you want to get a quick insight in the conference, I tried to highlight the things that stood out to me most.
Energy entered the chat with a bang.
I am putting this one up top, because it is easy to still get burried under the other amazing insights that is coming out of Amsterdam and the State of Tokenomics (SoT) released there. The SoT revealed that 38% are considering energy consumption in their Tokenomics. That to me is huge news. For years I feel like there has been a quiet fight within FinOps to get sustainability on equal footing with cost and AI Economics seem to have just seen energy rush through the wall like the Kool-Aid man.
It shows that AI Economics goes beyond analysis of the usage numbers through cloud bill or vendor provided logs. There’s a broader chain at work here and in that chain energy (and water!) drive a large part of the cost and value equasion.
AI economics is layered, FinOps needs to expand to all 5 layers.
Right away, JR Storment gave me food for thought in his keynote. If you look at the entire chain of cost and value of AI, it goes far beyond what FinOps traditionally looks at. While we traditionally look at the operational cloud part as it applies to our applications, projects, business logic… AI economics speak about the cost of data centers, resources, people and so much more. I intend to dive deeper into this by reading the paper on the different layers of tokenomics, which seems highly interesting.
Two tracks of spend: SLM/LLMs vs Developer Endpoints.
Another big take-away for me was that there seem to be two big tracks on-going in the current push to establish Tokenomics practices. On the one hand there’s the need to bring clarity to the value of using SLM/LLMs in infrastructure integration and business projects such as workflow automation, chatbots, operational activities… While on the other hand there’s the AI costs driven through developer usage in tools like VSCode, Cursor… I think there’s also a case to be made to include non-developer tools such as Microsoft Copilot in this group.
In short, I think two tracks are forming with some measure of clarity: Economics of AI in Infrastructure Integration and Economics of AI in Work Execution.
It really is not about tokens, and everyone knows.
I have to come clean: I was sceptical. And I am open about that, I told JR as much when I spoke to him in Amsterdam and I will own up to my initial prejudice when I heard the name ‘Tokenomics‘.
The conference blew me away and convinced me that actually everyone knows it’s not about tokens. Hearing people share their stories about identifying value and structuring costs, no one was focusing on consuming less tokens. What’s more, one speaker even indicated lowering tokens increased costs. It feels like AI economics/Tokenomics are off to a really healthy start and I am excited to see the field grow.