The Real Cost of Using AI in 2026
The article discusses the cost of using AI in 2026, highlighting the different ways to access intelligence, including subscriptions, pay-per-token APIs, cloud GPU rental, and owning hardware. The author analyzes their own token consumption and calculates the costs of each method, finding that owning hardware can be a cost-effective option. The author's analysis is based on their moderate and mixed usage pattern, including coding, writing, and research, and they provide estimates of the annual costs for each method.
- ▪There are four ways to buy intelligence: subscription, pay-per-token API, cloud GPU rental, and owning hardware.
- ▪The author's token spend for open models that can be self-hosted was around $30 over two months.
- ▪The estimated annual costs for the author's usage pattern are: pay-per-token API (€130), cloud GPU rental (€2,300), and owning hardware (€2,900 upfront plus €30 per year in electricity).
Opening excerpt (first ~120 words) tap to expand
@adlrocha - The Real Cost of Using AI in 2026The math says that I shouldn’t be building my own AI rig, and why I am building one anywayadlrochaJun 28, 2026ShareA few weeks ago I wrote about the shift from GPU-poor to token-poor. Since this post, and the ones I wrote about my recent obsession with AI independence, a lot of people have asked me for advice about how they should access intelligence: “fine, but what should I actually do? Buy a subscription? Pay per token? Build a rig? Rent one?” I dodged the economics in the token-poor post, so let’s do them properly now.The first thing I did before writing this post, is to pull my own token bill for the last 60 days (which have actually been slower than usual) in order to model my own token consumption (sidenote: BI built a really cool tool…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Substack.
Discussion
0 commentsMore from Substack
