I’ve worked long enough in the power sector (20-plus years) to remember the cycles when we aspired to provide electricity too cheap to meter to everyone. It’s safe to say that was a pipe dream, given where we currently stand on both the input side (energy prices) and the end-use product (power prices rising for consumers).
There are a few reasons for this; surging power demand from artificial intelligence and industrial reshoring, capacity market spikes in places like PJM, and aging grid infrastructure and the capital requirements are among the most obvious. So when a similar promise about AI surfaced during the period when ChatGPT became the most popular AI chatbot in 2022, it felt familiar. Abundant artificial intelligence. On tap. At a low, low price. For everyone. Every time. But that is not the reality we are experiencing.
While the price of a token has gone down about 60 to 70 percent, according to Goldman Sachs, we are spending more on AI. Expenditure on tokens, the building block of text that AI uses to read and write, has continued to go up as more advanced AI models are released and AI agents take over from simple chatbots. The reason for this increase has little to do with the technology itself and, as is the case with power, has everything to do with who is capturing the value the technology creates.
Despite my desire to move on from this topic, (I’ve told colleagues and friends the conversation about the convergence of AI and power has become stupid, a topic for another post), I cannot put it down. The stories of abundance (cheap power, cheap intelligence for all!) are the same story. It will likely end the same way if we do not understand why and create policies that ensure the little guys (you and me) do not get left holding the bag.
