Fork detected. Volatility imminent.
Anthropic, the AI safety company that built Claude, has leaked a revenue target for 2028: $190 billion to $200 billion. This is not a prediction. This is a contract being written in the market's mempool, and the gas fees are the attention spans of every investor, VC, and competing lab.
Let’s dissect the block. The source is anonymous, the timing is August 2024, coinciding with a new funding round. The number is astronomical. Even for a sector that has normalized exponential curves, this is a 100x jump from a projected $5-10 billion in annualized revenue at the midpoint of 2024.
Context: Why Now?
This is a classic financing narrative, a signal sent to the market to manipulate the order book. The core logic is simple: a 1,900% increase in 4 years requires a compound annual growth rate (CAGR) of roughly 276%. In software history, this is unprecedented. To put it in blockchain terms, it's like a L2 promising to handle 10x the throughput of Ethereum mainnet within a single upgrade cycle, without any new sharding or ZK proofs. The math is aggressive, but the narrative is the real asset.
The context is the AI arms race, specifically the battle between Anthropic and OpenAI. OpenAIs annualized revenue is 3-5x larger. They have ChatGPT, the consumer gateway. Anthropic has a brand built on safety, interpretability, and constitutional AI. This is a bet on the enterprise market, not the mass market. The target implies that by 2028, Claude will be running the backend of the global economy, not just assisting with code generation.
Core: The Data Does Not Lie, But It Can Be Recursive
Let’s run the numbers. If $200B is real, half from API, that’s $100B. At current Claude 3.5 Sonnet pricing ($3/$15 per million tokens), that’s roughly 33.3 trillion input tokens per day. That’s a mempool congestion event of global proportions. The current global internet traffic is estimated at 1-2 trillion tokens per day. This means Anthropic would need to process orders of magnitude more data than the entire internet generates today.
This is not a scaling problem. This is a physics problem. The only way this works is if the token price drops by a factor of 100-1000, or if the model’s value per token explodes. The latter implies a shift from text generation to autonomous agent execution. The model must become a full-stack replacement for a white-collar worker, not just a copilot.
Based on my audit experience with EigenLayer’s slasher contract, I can tell you that forecasting a system’s capacity without understanding the underlying logic is dangerous. The same applies here. The revenue target is a “bull case” scenario, likely a high-end estimate for a funding pitch deck. The internal models probably have a base case at $50-80B and a bear case at $10-20B. The leak is the bull case, designed to set the valuation floor.
Contrarian: The Unreported Angle – The Security vs. Growth Betrayal
Here’s the blind spot everyone is missing: Anthropic is built on a promise of safety-first AI. Their entire public mission is “ensuring AGI benefits humanity.” A $200B revenue target by 2028 is a direct contradiction of that mission. It implies a radical acceleration of deployment, a push towards high-autonomy agents that can make decisions involving money, law, and medical care. The Responsible Scaling Policy (RSP) becomes a liability, not a feature.
This leak is a signal that the “growth faction” inside Anthropic is winning. The company is preparing to move from “safe AI” to “safe-ish AI at scale.” The investors are demanding a return on their capital, and the answer is to deploy agents into the wild. This is a recursive risk. The faster they deploy, the higher the chance of a catastrophic failure. A single high-profile agent error could trigger a global regulatory response that kills the entire sector’s growth trajectory.
Think of it like a L2 chain that promises to scale infinitely but has a known bug in the slasher logic. The testnet is safe. The mainnet is where the capital is. The revenue target is the mainnet launch. The bug is the lack of a truly robust safety framework for autonomous agents.
Takeaway: The Next Watch
The real question is not whether Anthropic hits $200B. It’s whether the market will price in the risk of failure before the next funding round closes. The signal to watch is the next Claude model – Claude 4. If it shows a demonstrable leap in task completion and autonomy, the narrative is strong. If it’s a marginal improvement, the bubble in the valuation will deflate faster than a TerraUSD peg.
This is a bet on the future of AI as an economic infrastructure. It’s a bet that the market will accept a 10-20x forward revenue multiple on a company that hasn’t proven it can scale. The smart money is watching the on-chain data: the GitHub commits, the model evaluation benchmarks, the enterprise contract signings. The narrative is the hook. The code is the truth.
I’ve seen this playbook before. The 2020 Uniswap fork sprint. The 2022 Terra collapse debate. The 2023 EigenLayer audit. The 2024 Bitcoin ETF positioning. The 2025 AI-Agent economy framework. The pattern is always the same: a narrative is created to extract value from the market. The truth is in the execution. The question is whether the market will understand the technical gap before the funding round closes.
