Why smarter AI models could drive up compute prices 10x
As AI labs like Anthropic scale revenue 10x year-over-year while compute capacity only grows 3x, compute prices must rise significantly to bridge the gap. The speaker argues that as AI models become more capable, they can monetize the same compute at much higher rates, potentially driving prices up 10-15x, creating winner-take-most dynamics and pricing out current AI applications.
Summary
The speaker analyzes the computational economics facing AI labs over the next few years. Anthropic has 10xed revenue for three consecutive years (ending last year at $9B, projected to reach $100-150B this year), but compute capacity only grows 3x annually. For labs to maintain 10x revenue growth with only 3x compute growth, three mechanisms must occur: increasing margins, raising compute prices, or shifting more compute toward inference rather than training.
All three trends are already happening: Anthropic's inference margins have risen from 40% to 80%, spot compute prices are 40% higher than February 2024 lows, and the inference-to-training compute ratio is increasing (from 25% inference in 2024 to approximately 50% now). However, labs prefer not to shift heavily toward inference, as this signals AI progress has stalled and positions them as cloud providers rather than AGI builders.
The speaker argues that margins cannot realistically exceed 90% without being competed away, leaving compute price increases as the likely escape valve. He cites Google and Anthropic's arrangements with SpaceX, where they pay 2x spot prices for dedicated GPU capacity. A key insight: if an AI model achieved human-level software engineering capability, the compute (H100 equivalent) should theoretically rent for $250K annually—15x current spot prices—based on software engineer salaries.
The speaker acknowledges this echoes the Simon Ehrlich-Julian Simon debate about resource scarcity but argues compute supply is less elastic than commodity extraction. He identifies three components of the 3x annual compute growth: 1.4x from Moore's Law (likely unsustainable), 1.2x from new fab construction (bottlenecked by ASML EUV machine availability through 2030+), and 1.8x from AI absorbing wafer allocation from smartphones/PCs (hitting a wall when leading-edge nodes reach 86% AI allocation). He concludes this 3x growth rate may be difficult to maintain and certainly cannot accelerate significantly.
Key Insights
- For labs to maintain 10x revenue growth while compute only 3xes, either margins must increase beyond 90% (unsustainable due to competition) or compute prices must rise—and the speaker argues price increases are the more likely outcome.
- A human-level AI software engineer running on an H100 should theoretically rent for over $250K annually based on software engineer salaries—more than 15x the current spot price—demonstrating how smarter AI models enable much higher compute monetization.
- The 3x annual compute growth rate is decomposed into three components (1.4x Moore's Law, 1.2x new fabs, 1.8x AI absorbing smartphone/PC wafer capacity), and each component faces fundamental constraints that make acceleration difficult or impossible.
- As compute becomes more expensive, labs with more efficient models can charge significantly higher margins because using a weaker model becomes economically irrational on scarce, expensive compute—creating a powerful incentive for model efficiency.
- AI labs strongly prefer to allocate most compute to training rather than inference, because high inference spending signals that AI progress has stalled and positions them as cloud providers rather than AGI builders seeking investor funding.
Topics
Transcript
[0:00] Today I want to talk about what the comput situation for the labs will look like over the next few years. For the last three consecutive years, Anthropics revenue has 10xed year-over-year and it's likely to do so again this year. So they ended last year with 9 billion in revenue. I think they'll probably end this year with somewhere between 100 billion to $150 billion in revenue. Now for this trend to continue, Enthropic would need to make $1 trillion in revenue by the end of next year. Of course, there's no deep reason why this has to be true. It's a very wild conclusion, and it's ultimately a question of AI capabilities. Does AI get that useful…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Dwarkesh Patel
Grant Sanderson's Advice for Students
Grant Sanderson advises students pursuing mathematics to think critically about the value they create and where money flows in their chosen career path, rather than simply following their aptitude. He argues that teaching is one of the most stable post-AGI careers due to its relational, mentoring nature that goes beyond just explaining concepts.
The Problem With How LLMs Generate Text - Grant Sanderson
Grant Sanderson explains that autoregressive text generation in LLMs is fundamentally misaligned with human writing processes. Rather than composing with foresight and deep connections, LLMs are forced to predict tokens sequentially with memory wiped between predictions, making it difficult for them to form the unlikely but substantive connections that characterize good writing.
Why General Relativity Is the Most Beautiful Idea in Physics - Adam Brown
Adam Brown celebrates Einstein's theory of General Relativity as the most beautiful intellectual achievement of a single mind, highlighting how it elegantly describes phenomena across multiple orders of magnitude—from elevator thought experiments to the fate of the universe.
China's Belt and Road Problem - Sarah Paine
Sarah Paine argues that China's Belt and Road Initiative is strategically flawed, citing the economic and security superiority of maritime shipping over land-based routes and China's vulnerability to naval blockade in wartime due to its geographical position surrounded by shallow seas and islands.
Einstein's happiest thought: General Relativity from scratch – Adam Brown
Adam Brown explains Einstein's general relativity from first principles, starting with the equivalence principle and how gravity curves spacetime. He demonstrates why black holes exist as inevitable consequences of general relativity and discusses how falling into a black hole feels different depending on your reference frame.