Future Stories

It is a procurement metric, not a sales metric. There are three separate stacks and the industry deliberately blurs them. CoreWeave’s 3.5 GW contracted power and its $99.4B backlog are two different books. How tightly is the second mapped onto the first. They are very tightly, by design, because layer 3 is the collateral for ... Read more

Semianalysis agrees with my analysis of SpaceX getting up to $300 Billion per year by end of 2027. Spot prices are higher. The AI lease rates make sense. Coreweave and Nebius are spreading out fixed premium GPUs annual prices but requiring 5 year lockin. 5 years for $100 billion but not available during the next ... Read more

SpaceX Will push to new technology to improve the most important part of TERAFAB. These are the secrets that will enable 50 times more chips by 2030-2035 and the thousands more by 2040-2045. ASML technology will be taken to the next level. They only replace and improve the light source, which is about half of ... Read more

Terafab will close the gap between today’s chip production and the future’s demand — a future among the stars. Terafab Texas will be the largest and most valuable building on Earth by far. And it will be stunningly beautiful. pic.twitter.com/4NweOqTL7y — Elon Musk (@elonmusk) August 6, 2026 Sequoia's @shaunmmaguire explains why @elonmusk's Terafab is massively ... Read more

Recursive Language Models (RLMs) are a general inference paradigm that treats long prompts as part of an external environment and allows the LLM to programmatically examine, decompose, and recursively call itself over snippets of the prompt. RLMs can successfully process inputs more than an order of magnitude beyond model context window limits and, even for ... Read more

What a 23% kernel speedup tells us about whether AI can improve itself and how to quantify the gains. In 2025, Google DeepMind’s AlphaEvolve, an AI system that discovers algorithms, found a way to make a FlashAttention kernel run 23% faster. This was a great result but it was not the beginning of a runaway ... Read more

When does revenue growth and profits overtake AI capex spending ? I have run numbers on the fastest AI data center builder. The faster you can build, then the faster you can monetize and get to profits.

The high $50 Billion per GW lease rates for xAI/SpaceX Colossus and similar premium, immediately available AI clusters are for scarce of ready-to-use, large-scale coherent GPU capacity. Kevin Pathrath’s is wrong about a conspiracy. He is comparing against different types of data center. The $2.7M/MW ($2.7 billion per gigwatt build figure is shell-only and brownfield. ... Read more

Prime Agent is an open-source self-improving coding and research agent (harness) released by Prime Intellect on August 5, 2026. It is built for coding workflows and long-running autonomous tasks. It combines a Recursive Language Model (RLM) runtime with a Continual Harness, emphasizing token efficiency and expressiveness through programmatic tool calling, treating context as a variable, ... Read more