Research Archive
Gigawattonomics
Power Sets the Deployment Schedule If you have followed our research, we have tried to maintain as close a view as possible on returns on invested watts. Power availability determines when capacity can be deployed, while revenue per watt determines whether that capacity earns an adequate return. We have consistently framed infrastructure decisions through…
Meta Is an Ads Company Building a Token Factory
Start With the Business Model Business model dictates strategy. That has always been one of the more useful observations in technology analysis, and it provides the way to interpret Meta’s AI buildout. Meta is an advertising company. It distributes its products to billions of people at no direct cost, monetizes their attention, and uses technology…
GPT-5.6, the Diligence Stack Agent Bench, and Why Practical Agent Benchmarks Need to Get Real
Most AI benchmarks still do a mediocre job of explaining how useful models are in the kinds of agentic workflows businesses actually care about. That is not because benchmarks are useless. They are useful. But many of them are still optimized around narrow tasks, coding-heavy environments, academic-style reasoning problems, or synthetic prompts that do not…
The GPU Tsunami: TSMC, Intel, and Samsung Foundry
We continue our series on the GPU Tsunami within the framing that the GPU (thanks to AI) was the spark that set the entire industry on fire. Our first report in this series looked at how power semiconductors and analog control are benefiting from this dynamic. Higher rack density turns electricity delivery into a larger…
Where AI Constraints Become Pricing Leverage
This is part 2 of our series on the semiconductor industry constraints. Part 1 below: Counting Real AI Capacity Ben Bajarin · Jun 30 Read full story For most of the last twenty-five years, unit shipments ran well ahead of industry revenue. The industry scaled unit shipments faster than dollars. Then AI hit. The chart…
Counting Real AI Capacity
This report is our first full constraint map for AI infrastructure. We have been tracking a constraint stack that now runs much deeper than accelerator allocation. The map in this report pulls those constraints into one framework and asks a simple question: which layers determine when announced capacity becomes billable compute? Capacity has to be…