How Claude went from $9 billion to $45 billion in one year | CFO explains
Stop thinking linearly—start thinking exponentially. When your business or capabilities grow exponentially, small weekly changes compound into vastly different outcomes. Don't forecast once a quarter. Instead, build scenarios that account for a 'cone of uncertainty,' constantly update your assumptio
1h 22mKey Takeaway
Stop thinking linearly—start thinking exponentially. When your business or capabilities grow exponentially, small weekly changes compound into vastly different outcomes. Don't forecast once a quarter. Instead, build scenarios that account for a 'cone of uncertainty,' constantly update your assumptions, and maintain flexibility in your strategy. The world changes too fast for static plans.
Episode Overview
Krishna, a leader at Anthropic, discusses the company's approach to compute allocation, model development, and scaling laws. He explains how Anthropic navigates exponential growth, uses multiple chip platforms efficiently, and balances compute between model development, internal use, and customer demand while maintaining frontier intelligence.
Key Insights
Intelligence Is Multi-Dimensional, Not a Single IQ Score
Model intelligence isn't just a number that goes from 110 to 125. It encompasses long-horizon task capability, tool use, speed, and agentic performance. A model that completes tasks in a day versus a week can be seven times more valuable, even with equal raw capability. Real-world customer feedback matters more than saturated benchmarks.
Compute Is the Lifeblood—But Flexibility Is the Superpower
Anthropic uses three chip platforms (Amazon Tranium, Google TPUs, Nvidia GPUs) fungibly, allocating compute dynamically across training, internal use, and customer serving. This flexibility, built over years, allows them to maximize ROI from every dollar of compute. Buying too much compute risks bankruptcy; buying too little means losing the frontier.
The Cone of Uncertainty Requires Scenario Planning
Exponential growth creates a wide range of possible outcomes over 1-2 years. Anthropic plans for multiple scenarios, works backwards from them, and maintains a low bar for updating assumptions. Linear thinking breaks in exponential environments—you must revisit forecasts constantly, not quarterly.
Returns to Frontier Intelligence Are Extremely High in Enterprise
Each new model generation unlocks more TAM and enables new use cases. Anthropic went from $9B to $30B+ run rate revenue in four months, driven by model intelligence leaps. Customers invest heavily in newer models because they deliver better results, faster execution, and enable previously impossible products.
Recursive Self-Improvement Is Real—90%+ of Anthropic's Code Is Written by AI
Anthropic's models help build the next generation of models. Claude Code writes most of the company's code, and AI accelerates research, product development, and efficiency improvements. The models themselves are becoming key contributors to their own evolution, creating a flywheel of capability improvement.
Notable Quotes
"If you buy too much compute, you go out of business. If you buy too little compute, you can't serve your customers and you're not at the frontier."
"Intelligence for us is multi-dimensional. It's not just a score. What is the real world capability of this model?"
"I think humans mostly think linearly and incrementally. That's a paradigm I've had to break for myself—to stop thinking linearly and think exponentially."
"We started the year with about $9 billion of run rate revenue and we ended the quarter with north of $30 billion of run rate revenue. That kind of change is really enabled by these model intelligence leaps."
"Now within the company, 90 plus percent of our code is actually written by Claude Code. A lot of Claude Code is written by Claude Code."
Action Items
-
1
Adopt Scenario-Based Planning for Uncertainty
Stop relying on point estimates or quarterly forecasts. Instead, create 2-4 scenarios that represent different points in your 'cone of uncertainty' over the next 1-2 years. Work backwards from these scenarios to make decisions that give you flexibility across outcomes.
-
2
Build Flexibility Into Your Infrastructure and Contracts
Whether it's compute, suppliers, or partnerships, invest in building systems that allow you to adapt quickly. Anthropic's fungibility across chip platforms gives them optionality. Identify where flexibility matters most in your business and build it in proactively.
-
3
Lower Your Bar for Updating Assumptions
What was true a month ago may not be true today in fast-moving fields. Create a culture where people constantly challenge previously held assumptions. Review key metrics and beliefs weekly or monthly, not quarterly.
-
4
Invest in Capabilities That Create Compounding Returns
Anthropic allocates compute to internal use (like AI-assisted development) because it accelerates everything else. Identify investments that might reduce short-term revenue but create long-term leverage and speed—like tools that make your team 2-10x more productive.