Keep pulling the thread on Tony Kim.
The cost of a high-end AI compute server is now approximately $1 million, a significant increase from the $10,000 price point of pre-AI era servers.
There are 30 to 40 potential IPOs for Chinese robotics companies expected this year.
The AI industry is projected to spend $1 trillion in CapEx this year and $10 trillion over the next 5 years.
Pre-AI data centers were designed for megawatt-level power consumption, whereas modern AI data centers require gigawatt-scale power.
The new AI technology stack, centered on massive compute and token-based sales, is eroding the high margins previously held by top-layer application (SaaS) companies.
AI model intelligence is improving at a rate of one order of magnitude (10x) per year.
Orbital data centers are expected to become a viable technology around the year 2030.
A utility-scale, logically error-corrected, million-qubit quantum computer is expected to be developed by 2030.
Small modular reactors (SMRs) and small fusion reactors are expected to receive regulatory approval by 2030.
Artificial General Intelligence (AGI) based on classical computing is predicted to be achieved around 2030.
Companies must position themselves within the "token flow" — creating, serving, or repackaging AI tokens — to remain relevant, while traditional application companies are struggling to find their role.
The cost of a compute server has increased from approximately $10,000 in the pre-AI era to $1 million for AI-specific hardware.