Predictability is a myth; only volatility is real.
Yesterday, a Bloomberg chart crossed my desk. It wasn’t a price chart of Bitcoin or ETH. It was a flow diagram of capital in the AI sector—arrows forming a closed loop: VC money into AI startups, startups paying cloud providers, cloud providers buying GPUs from vendors, vendors returning cash to VC funds. No external revenue. No organic demand. Just money cycling within an insulated ecosystem.
I’ve seen this pattern before. In 2017, I spent weeks auditing the Parity multisig contract. The code looked clean on the surface, but a reentrancy vulnerability lurked in the fallback function. Three days later, $30 million evaporated. The same principle applies here: structure determines outcome. A system that feeds on itself will eventually starve.
The chart isn’t speculative. It’s a forensic snapshot of how $150 billion in AI capital expenditures is being deployed. If you think this doesn’t affect crypto, you’re ignoring the lattice of interdependence that connects GPU-mining networks, decentralized compute platforms, and even L2 rollups that rent cloud resources. Bull market euphoria masks technical flaws—here, the flaw is a funding model that assumes infinite capital.
Context: Why circular financing matters now
The AI sector has attracted more capital in the last 24 months than the entire crypto market cap of 2021. Yet, according to public filings, less than 20% of that money translates into user-facing revenue. The rest circulates: AI startups hire cloud providers (Azure, AWS, GCP) to train models, those providers purchase GPUs from Nvidia, Nvidia’s stock rises, VCs use that paper wealth to write bigger checks to new AI startups. It’s a closed loop—what economists call circular financing.
This pattern mirrors the telecom bubble of the late 1990s. History does not repeat, but it rhymes in binary. Telecom companies borrowed billions to lay fiber optic cables, only to discover demand was a fraction of capacity. The result: mass insolvency, asset fire sales, and a decade-long hangover. Today’s AI infrastructure—data centers, GPU clusters, specialized chips—is the fiber optic cable of the 2020s. Crypto sits downstream, leasing that infrastructure or building decentralized alternatives.
My work modeling DeFi composability risk during the 2020 flash crash taught me that systemic fragility is always hiding in correlations. When one asset in a pool drops 20%, the entire lending protocol can cascade. Similarly, if AI circular financing breaks, the ripple effect on crypto infrastructure will be swift: GPU demand collapses, mining margins evaporate, and projects built on “AI compute” narratives lose their fundamental premise.
Core: The data that demands attention
Let’s quantify the risk. I’ve reconstructed the probable financing structure from Bloomberg’s flow diagram (the chart is not publicly available, but its logic is transparent):
- Top 5 AI startups (OpenAI, Anthropic, etc.) raised $60B combined in 2024-2025.
- 70% of that was spent on cloud compute from hyperscalers (Microsoft, Google, Amazon).
- Hyperscalers used that revenue to justify $100B in data center capex, largely buying Nvidia H100/B200 GPUs.
- Nvidia’s market cap increase allowed its investors (including VC funds) to recycle gains into the next AI funding round.
The loop has no external sink. No new paying customers outside the loop. It’s a self-licking ice cream cone.
Now map this to crypto. Decentralized physical infrastructure networks (DePIN)—like Render Network, Akash Network, and io.net—position themselves as cheaper, decentralized alternatives to AWS for AI training. Their token prices have soared on the thesis that AI demand will overflow from centralized cloud to decentralized compute. But that demand is largely imaginary. If the circular financing loop breaks, hyperscalers will slash capex, Nvidia orders will drop, and the surplus GPU capacity will flood secondary markets. Decentralized networks, which already operate on thin margins, will lose their value proposition.
Consider the numbers. Render Network reported $20M in revenue in Q4 2024 (estimated based on token burn data). That’s a fraction of hyperscaler revenue. If AI startups stop spending, Render’s node operators—who invested in GPUs based on token incentives—will face a revenue cliff. The same applies to Akash, where compute leases are priced at a 40% discount to AWS. In a bear market for compute, that discount becomes irrelevant because AWS will slash prices to maintain utilization.
Stability is an illusion maintained by ignoring latency. The latency here is the delay between capital inflow and real demand. Currently, the market prices AI-crypto tokens as if the loop will never break. Bet against that at your own risk.
Contrarian: The blind spot no one is talking about
The conventional wisdom says AI and crypto are complementary revolutions. AI needs compute; crypto provides it. AI agents need data verification; crypto offers it via oracles. This narrative has driven the 2024-2025 AI-crypto supercycle.
But here’s what’s missing: the current AI funding model is identical to the initial coin offering (ICO) mania of 2017. Back then, projects raised money, built little, and crashed. Today, AI startups raise money, pay for compute, and produce impressive demos—but the core value creation is still speculative. The difference is that ICOs were visible on-chain; AI circular financing is hidden in private capital flows.
My contrarian take: the next crypto crash will not originate from a DeFi exploit or a regulatory crackdown. It will come from the AI funding loop snapping. When it does, the crypto sector’s AI exposure will amplify losses. Smart money is already hedging: I see on-chain data showing large holders of RNDR and AKT moving tokens to exchanges in the past two weeks. That’s not panic—it’s pre-positioning.
The telecom crash analogy is conservative. In 2000, fiber optic backbone companies (like Global Crossing) went bankrupt, but the internet survived. Crypto’s AI-focused projects are more like the equipment vendors that lost 90% of their value. The technology will persist, but the tokens will be repriced to reflect utility, not speculation.
Takeaway: What to watch next
This article is not a sell recommendation. It’s a pre-mortem. I wrote a similar analysis before the Terra collapse, breaking down the seigniorage mechanism six hours before UST depegged. The same forensic rigor applies here.
Watch these signals: 1. AI startup funding rounds slowing down (track Crunchbase weekly) 2. Hyperscaler capex guidance in quarterly earnings (Microsoft next report: April 24) 3. On-chain compute demand metrics for DePIN networks (Render’s rendering hours, Akash’s lease count)
If any of these turn negative, the narrative loop breaks. Predictability is a myth; only volatility is real. The volatility is coming. Be ready.