AI Is Already a $175B Industry — And It's Growing 3× Faster Than the Internet Did
A new report dropped this week that puts hard numbers on something everyone in tech has been feeling but struggling to quantify: AI is not a hype cycle. It's a revenue wave.
The State of the AI Economy report from @exponentialview tracked de-duplicated consumer AI spending and landed on $110 billion in actual AI revenue over the past 12 months. The annualized run rate is now above $175 billion — and the speed at which that number is climbing is unlike anything we've seen before.
Here's what the data actually says, broken down plainly.
The Numbers That Matter
The headline is $175B annualized, but the more interesting figure is the acceleration.
In 2023, generating $1 billion in new AI revenue took 180 days. Today, it takes less than two days.
That's not incremental growth. That's a fundamentally different kind of adoption curve. For comparison, the mobile and internet waves of the 2000s and 2010s grew fast — but AI is growing approximately three times faster than either of them did at this stage.
Enterprises Are Moving — But Only Barely
Despite the headline numbers, enterprise adoption is still in its early innings.
31% of S&P 500 companies mentioned AI in their earnings calls last quarter. But only 20% of those companies actually quantified the impact. The rest are mentioning it because they feel like they have to — not because they've fully integrated it into how they work.
The report puts it simply: enterprise AI has moved out of pilot phase, but broad rollout is still in its early stages. The gap between "we're exploring AI" and "AI is how we do work now" is where most large companies still live.
The Price-to-Demand Relationship Is Unusually Strong
Here's something the report found that's worth flagging: for every 10% drop in token pricing, AI usage increases by 12–18%.
That's unusually elastic demand. It means as AI gets cheaper — and it is getting cheaper, fast — actual usage isn't just growing proportionally. It's outpacing the price drop. People aren't just using AI more because it's cheaper; they're using it for new things they couldn't justify at old prices.
This has big implications for where AI ends up. If pricing keeps falling and demand keeps responding this strongly, the ceiling is much higher than the current numbers suggest.
The Infrastructure Bet Is Real, But Risky
On the supply side, the report notes that hyperscaler cloud companies (the Microsofts, Googles, and Amazons of the world) are currently earning enough from AI to roughly cover their infrastructure depreciation — but only if you assume a six-year useful life for their GPU hardware.
That's not a comfortable margin. Electricity costs and data center capacity are the named constraints on future scaling. The companies betting on AI growth are betting they can keep building fast enough to stay ahead of demand — and that their hardware holds up long enough to justify the cost.
What the OpenAI Codex Data Adds
Separately this week, OpenAI published an internal report that adds a striking ground-level data point to the macro picture.
At OpenAI itself, AI agents — specifically Codex — have almost entirely replaced ChatGPT as the primary work tool. Codex now accounts for 99.8% of their internal output tokens (up from less than 10% less than a year ago). Legal, Finance, and Recruiting teams crossed the 50% Codex usage threshold in April. The average lawyer or recruiter at the company now gets more than 85% of their output from agents, not from manual work.
That's not a prediction. That's already happening, at one of the most-watched companies in tech.
What This Means If You Use OpenClaw
The report describes a world where AI is moving from "something teams experiment with" to "how work actually gets done." That transition is already complete at the leading edge — and it's driven by agents, not chat.
That's exactly what OpenClaw is built for.
When you connect OpenClaw to your tools and workflows, you're not building a ChatGPT wrapper. You're building something closer to what OpenAI's internal teams are using: an agent that runs background tasks, handles recurring work automatically, and gets better context with every session.
The $175B industry the report describes isn't built on chatbots. It's built on systems that can take a task, run it, and hand back a result — without someone having to babysit the process. That's what agent platforms do.
If you want to see what this looks like in practice — not as a stat in a report, but as something you can actually use — ClawWorld is a good place to start.