The frontier AI market is picking teams like a schoolyard : Salesforce chose Anthropic, OpenAI expelled Cursor, & governments ration who may use the frontier. Even the open weights now segment.
AI model companies buy wholesale electricity by the megawatt and resell it as cognitive work. The unit economics of turning power into intelligence.
NVIDIA guided Q3 to $108b, but hyperscaler growth slowed to 13% sequentially. The buyers making up the difference need financing, so DSO jumped 15 days to 60 & the commitment stack reached $581b including a $101b equity portfolio in its own customers.
Should a calendar agent run for your entire five-year company tenure or reset every day? Why perpetual sessions fail & how to design agent lifespans.
AI infrastructure shortages do not hit simultaneously. They cascade in multi-year waves across the server rack & into the physical grid : GPUs in 2023, memory & flash in 2024–2025, CPUs in late 2025, spinning disk in 2026, & data centers at $20b per gigawatt.
Local models now answer 89% of everyday chat & reasoning queries as well as frontier models, & their efficiency per watt has improved 5.3x in two years.
Software's hidden cost is learning its grammar. AI lets a founder speak English & use CAD to make a dress once previously unmanufacturable.
A local model that generates tokens 2.2x faster than the incumbent finished later in wall-clock, because it emitted 3.1x more tokens. Tokens per second is the wrong metric. Time to answer is the right one.
Today's AI doesn't learn after it's trained. Test-time training changes that, & the tradeoff it creates, one model per user instead of one model for everyone, may be the more efficient architecture despite what it costs a GPU provider to serve.
84% of tokens on OpenRouter aren't state of the art. The six models carrying the supermajority deliver 77% of frontier performance at 2.5% of the price.
