In 2024, a single AI company's revenue projection for 2028 is larger than the entire global cloud computing market in 2023. That number is not just a valuation tool—it's a signal that the world is betting on a future where autonomous agents handle the bulk of economic activity. For crypto, this is both an opportunity and a warning.
Context: The Anthropic Bet and Its Hidden Assumptions
Anthropic, the AI safety company behind Claude, has reportedly told investors it expects to generate $190 billion to $200 billion in annual revenue by 2028. This figure, leaked during a fundraising round in August 2024, has been dissected by analysts as a classic bull-case narrative. From my seat as a Digital Asset Fund Manager in Nairobi, I see a different layer beneath the noise. The forecast implies a world where AI agents are embedded in every critical business process—from legal contracts to medical diagnosis to financial trading. To reach that scale, Anthropic must deliver near-AGI-level models, secure enterprise trust, and deploy compute infrastructure that rivals the entire cloud industry.
The ledger remembers what the algorithm forgets. But the algorithm is evolving fast. Whether this forecast is achievable matters less than what it reveals about the direction of capital: massive compute, autonomous agents, and the need for a trust layer that is decentralized and programmable. That is where crypto enters the equation.
Core: The Infrastructure Gap and Crypto's Role
From my experience modeling AI-agent economies on ZK-proof networks in 2025, I know that the compute demands of a $200 billion AI revenue stream are staggering. At current pricing, serving that many tokens per day would require hundreds of thousands of high-end GPUs. The energy consumption would rival a mid-sized city. Anthropic will likely rely on Amazon and Google for compute, but that creates a centralization risk. For crypto, this is the opening.
Decentralized compute networks—like Akash, Render, or even Ethereum's upcoming ZK-rollup-based compute markets—could become the alternative. But there is a catch: most rollups and Layer 2s today cannot handle the data throughput needed for AI inference. The Data Availability layer is overhyped; 99% of rollups don't generate enough data to need dedicated DA. However, AI agents will generate massive data streams—millions of transactions per second, each requiring verification. This demands a new paradigm for data availability, one that is scalable, low-cost, and trustless. Projects like Celestia or Avail are positioning for this, but they must prove they can handle AI-grade workloads.
Stablecoins and payments are another critical junction. If AI agents are to transact autonomously, they need a medium of exchange that is censorship-resistant. USDC's compliance-first strategy is its biggest risk: Circle can freeze any address within 24 hours. How is that decentralized? For AI agents operating across jurisdictions, a freeze on their treasury could halt operations. This is where decentralized stablecoins—like DAI or new programmable money protocols—become essential. The market will demand a stablecoin that cannot be frozen by a single entity, especially when agents handle billions in value.
DeFi's interest rate models also need an overhaul. Aave and Compound's interest rate models are completely arbitrary—they have nothing to do with real market supply and demand. For AI agents that need to borrow and lend capital efficiently, these models create inefficiencies. A 2% rate gap due to a curve parameter can cause agents to miss profitable arbitrage opportunities. We need rate models that dynamically adjust to on-chain liquidity and external market conditions, perhaps using machine learning itself. The irony is that AI could help design better DeFi protocols, but it first needs a reliable DeFi base.
Contrarian: The Decoupling Thesis and the Risk of Centralized AI Dominance
The contrarian view is that Anthropic's forecast is pure fiction designed to raise capital. But even if it is only 10% accurate, the compute demand will be enormous. However, the crypto industry's response—building decentralized AI networks—may be premature. The risk is that centralized AI (like Anthropic or OpenAI) will dominate because they can achieve scale and reliability that decentralized networks cannot match in the short term.
Trust is borrowed; trust is never owned. The ledger remembers what the algorithm forgets, but algorithms are getting faster. Decentralized compute networks suffer from latency, variable quality, and lack of enterprise-grade SLAs. Anthropic will likely stick with AWS and Google Cloud for production, leaving decentralized networks for niche or speculative use cases. This is a blind spot for many crypto investors who assume that AI will inevitably decentralize. It may not, at least not in the next 4 years.
Furthermore, the autonomous agent risk is real. If AI agents execute millions of transactions on decentralized networks, any bug in the agent's logic could cause cascading failures. I have seen simulations where 10,000 agents trading on a single DEX caused a liquidity crisis. Safety is the only yield that compounds over time. We must build circuit breakers, human-in-the-loop checks, and insurance mechanisms before we let agents run wild. The current infrastructure is not ready.
Takeaway: Positioning for the Cycle
Whether Anthropic reaches $200 billion or not, the direction is clear: AI agents will require massive compute, and that compute must be trustless. Crypto's role is to provide the settlement layer, the data availability, and the programmable money for these agents. But the market is still early. The 2028 forecast is a North Star, but the path is through 2025's consolidation.
We build walls not to keep out, but to keep safe. In a sideways market, the best positioning is in infrastructure that can survive a bear market: decentralized compute, secure DA layers, and stablecoins that resist censorship. Avoid speculative AI tokens without real utility. Watch for signals like Anthropic's next funding round or Claude 4's capabilities. The ledger remembers, but it also waits. And when the next cycle begins, the protocols that have been quietly building will be the ones that catch the wave.