By Craxel Founder and CEO David Enga
May 24, 2025
The All-In Podcast is an incredibly interesting window into tech and investor thinking. On today's episode, David Friedberg very delicately suggested that if we are able to add a terawatt of electricity per year to power AI, he would stop talking about the national debt. Presumably because the corresponding productivity gains for our economy would allow us to outgrow the debt.
Therefore, isn't it axiomatic that if we are able to find much more efficient AI algorithms that only need a tiny fraction of the computation of today's algorithms, that the corresponding productivity gains for our economy would also allow us to outgrow the national debt - without consuming such an extraordinary amount of resources?
I find this topic so interesting because it is a brute-force approach versus algorithmic innovation and its a parallel to what we've seen at Craxel with data. Brute-force techniques that scan large quantities of data and waste many resources to perform a query are prevalent today. ...or you can use Craxel's O(1) multidimensional indexing innovation to rapidly and efficiently query massive and complex data sets with a tiny fraction of the resources.
Since algorithmic innovation has provided a step change for what we can do with data, will someone invent something that does the same for AI? ...or is the future of AI really tied to an unprecedented build out of data centers and energy production? I'm going to cheat and say I think its going to be both; and if human ingenuity does come through with an important algorithmic innovation in AI, not quite as much will be built and the rest of the capacity will get used for even higher value purposes.
