The AI Bubble Is A Myth: Anthropic Prints $16B In Six Months
Critics say AI is a 5 trillion dollar bubble waiting to pop. The actual financial data from 2026 shows massive profits and stable valuations.
Business · Source: Hacker News
What happened
Everyone assumes the AI industry is a massive bubble about to burst. Companies have committed at least three trillion dollars to infrastructure and already spent another 1.9 trillion. Skeptics point to this five trillion dollar price tag as proof the math does not work. They argue the debt will crush the entire economy. They are wrong.
These massive costs are one-time capital expenditures for data centers. Once the servers are plugged in, the operating costs drop. Anthropic recently filed its S-1. The numbers show they make $2.08 for every single dollar spent on compute. They are already making more than their compute and infrastructure investments combined. If needed, AI labs can instantly cut costs by ending free tiers and pausing expensive research.
The revenue growth is staggering. Anthropic went from 41 million dollars in 2023 to 16.3 billion dollars in just the first half of 2026. Tech giants like Alphabet, Meta, and NVIDIA are seeing similar massive revenue jumps. Google Services grew by 37.8 billion dollars, proving they monetize AI directly to consumers. Their price-to-earnings ratios remain stable in the 20s. This looks nothing like the dot-com crash where companies had zero profits and ratios in the hundreds.
Key facts
- 5 trillion — Estimated total expense of the AI industry so far in dollars
- $2.08 — Revenue Anthropic makes for every $1 of compute
- 16.3 billion — Anthropic revenue in dollars for the first 6 months of 2026
- 37.8 billion — Growth in Google Services revenue in dollars
- 72 billion — Loss Anthropic recorded in dollars in the form of shares due to lenders
Why it matters
Founders need to stop waiting for the AI crash. The infrastructure is being paid for by real revenue, not just venture capital hype. Institutional lock-in is already happening. Large corporations sign massive token contracts and integrate these models into their daily operations. Transitioning away from an integrated AI system is not costless. Switching costs are high. The companies building B2B AI tools will reap the rewards of this inertia.
The lack of a technical moat does not mean a lack of profits. A leaked Google memo noted that open source models keep pace with frontier labs. That is fine. Fast food chains do not have moats for chicken sandwiches, but they still print money. Normal users do not want the hassle of hosting local models. Brand loyalty wins. People stick with ChatGPT simply because it was their first tool. The real war is for distribution and enterprise contracts, not just model superiority.
For builders
Stop worrying about the technical moat
Google admitted nobody has a moat in AI. Fast food chains do not have moats for chicken sandwiches, but they still print money. Build a great product and rely on brand loyalty to keep normal users paying.
Target enterprise institutional lock-in
Big companies are signing massive contracts for token usage. Once they integrate an AI tool into their workflow, they will not switch easily. Focus on B2B integrations where high switching costs force corporations to keep paying you.
Prepare for infrastructure and regulatory backlash
The biggest risk to AI is not financial. It is public anger over data centers and new regulatory barriers. Builders relying on cheap compute will lose heavily if an autonomous AI hacker causes public panic and triggers regulatory crackdowns.
My take
I am tired of the bubble talk. The math is right there in the filings. When Anthropic makes two dollars for every dollar of compute, that is a printing press, not a fad. Stop predicting the crash and start building.
Original reporting: Hacker News. This is my rewrite and opinion.