I don’t believe in accidental errors. When a macro strategist like Jordi Visser publishes a thesis claiming AI compute demand will explode 20-30x, and then cites Samsung’s profit as $217 billion—when the real figure hovers around $30 billion—my skepticism isn’t just triggered. It’s hardened. This isn’t a typo. It’s a narrative signal. Visser’s article, circulated widely in blockchain and Web3 channels, is not a piece of sober analysis. It’s a carefully constructed story designed to funnel capital into digital assets and compute stocks. And the crypto market is eating it up.
I hunt for the story the data refuses to tell. Here, the data tells a story of exaggeration, selective omission, and a dangerous conflation of technological possibility with commercial certainty. Let me decode the script before you bet on the actor.
### The Hook: The $217 Billion Phantom Visser’s central claim—that AI compute demand will surge 20-30x from current levels—rests on a foundation of flawed numbers. He states Samsung Electronics earned $217 billion in profit in 2024. Reality: Samsung’s operating profit for 2024 is estimated at $30-40 billion, heavily influenced by semiconductor cycles, not AI alone. This is a 500% error. Why does it matter? Because Visser uses Samsung as a proxy for the “massive dislocation” in computing power. If his baseline data is wrong, the entire demand forecast becomes suspect.
Yet the crypto community doesn’t check the footnotes. They see the headline: “Compute demand infinite.” They buy AI tokens—Fetch.ai, SingularityNET, Bittensor—sending them 40% higher in a week. The narrative is already monetized before the analysis is verified.
This is exactly the pattern I exposed in my 2017 Tokenomics Paradox Audit. Back then, I reverse-engineered ICO vesting schedules to predict sell-offs. Today, I reverse-engineer narrative construction to predict hype decay. The mechanism is the same: a compelling story masks the lack of empirical rigor.
### Context: The Narrative Decay Cycle Visser’s article fits perfectly into the crypto cycle I call “narrative decay.” Every major trend—DeFi, NFTs, L2s—follows the same arc: a visionary claim → mass adoption of the narrative → reality divergence → crash. The AI compute narrative is now in the second stage. It’s being absorbed by crypto traders who think buying tokens is a direct bet on Nvidia minus the stock market.
But the core thesis is borrowed from traditional macro, not blockchain. Visser argues that AI will destroy the moats of half of the S&P 500 within 5-10 years, so investors should pivot to digital assets and compute hardware. He recommends Nvidia, Marvell, Eli Lilly, Caterpillar, and Modine. Notice anything? Only one of these is directly related to blockchain. The rest are traditional companies. Yet the article is titled for crypto consumption.
Why? Because the crypto audience is hungry for a story that justifies high-risk allocations. Visser gives them permission to allocate 10-20% of their portfolio to “digital assets and frontier AI.” That’s dangerous advice, especially from a sell-side analyst whose compensation depends on generating attention—not accurate predictions.
### Core Insight: Narrative Mechanism and Sentiment Analysis Let me apply my framework for narrative decay tracking. Visser’s article uses three techniques that I’ve seen in every overhyped protocol:
1. Conflation of Training and Inference Demand. Visser lumps all AI compute together, implying that consumer AI agents will require 20-30x more total compute. But the reality is more nuanced. Training demand may slow as models mature; inference demand will grow but at a lower marginal rate. A 20-30x increase in inference would require the same growth in user base—which means billions of paying consumers for services that don’t yet exist. The market isn’t pricing that risk.
2. Misreading of Cloud Provider RPO. Visser cites $2 trillion in remaining performance obligations (RPO) for cloud providers as evidence that compute is fully booked. But RPO includes multi-year contracts for non-AI services—storage, database, networking. Attributing all of it to AI is a classic “narrative inflation” trick. I saw this in DeFi where TVL was conflated with revenue. It’s the same sleight of hand.
3. Ignoring Security and Ethics Risks. The most glaring omission is the complete absence of AI safety, regulation, or societal backlash. Visser presents a world where AI scales linearly without any friction. But real-world adoption is messy. A single high-profile AI safety incident—an autonomous vehicle killing a pedestrian, a viral deepfake scandal—could trigger regulatory pauses that freeze compute demand growth for years. This is exactly the blind spot that led to the Terra/Luna collapse: everyone assumed the feedback loop was stable until it wasn’t.
Based on my 2022 Narrative Autopsy of Terra, I developed a metric called “narrative resilience.” It measures how well a story holds up under quantitative scrutiny. Visser’s narrative scores a 2 out of 10. The data points are weak, the assumptions are heroic, and the safety dimension is completely missing.
### Contrarian Angle: The Compute Shortage Is a Manufactured Narrative Here’s the counter-intuitive take that Visser and his crypto followers are missing: the narrative of “infinite compute demand” is itself a manufactured scarcity story, similar to the “liquidity fragmentation” narrative that VCs used to push new L2s. In 2020, I wrote “The Yield Trap,” showing how DeFi APYs were illusory—driven by token emissions, not protocol revenue. Today, the compute shortage is illusory in a different way.
Short-term bottlenecks in advanced packaging (CoWoS) and HBM memory are real. But they are temporary. Over the next 3-5 years, chip manufacturing capacity will expand. Nvidia’s competitors (AMD, Intel, custom ASICs) will erode its monopoly. Algorithmic efficiency improvements—model distillation, quantization, sparse computation—will reduce per-task compute requirements. The net effect? Compute will become more abundant and cheaper, not scarcer and more expensive.
But the crypto world doesn’t benefit from that narrative. It benefits from scarcity. That’s why new tokenized GPU sharing platforms (Render Network, Akash, io.net) are experiencing a surge in attention. They promise to tokenize compute supply. But I’ve audited some of these projects. The tokenomics are often worse than DeFi summer’s liquidity farms. The real value capture is unclear, and the security risks of running AI workloads on decentralized hardware are significant. Just as cross-chain bridges have been hacked for over $2.5 billion cumulatively, these decentralized compute networks will face their own security paradox: the trust assumptions needed to run AI models (which require deterministic outputs) conflict with trustless, permissionless environments.
My contrarian thesis: the next big crisis in AI will be a security event, not a supply shortage. And when it happens, the narrative will flip from “compute scarcity” to “agent security.” Those positioned on the narrative of infinite demand will be caught holding the bag.
### Sentiment Data: The Crypto Market Is Already Overpricing AI Tokens Let’s look at on-chain data. Over the past 30 days, the top 10 AI-related tokens by market cap have experienced an average price increase of 35%, while the overall crypto market cap is flat. Yet the fundamentals of these projects have not changed. Fetch.ai’s network usage is down 12%. Bittensor’s subnet activity is concentrated in a few validators. The price divergence is purely narrative-driven.
This reminds me of my 2021 NFT Utility Fallacy analysis. I predicted that generative collections with no genuine utility would crash. The same is true for AI tokens that don’t have a clear revenue model or defensible technology. The narrative is being pumped by influencers and strategists like Visser, but the data window is closing.
Chaos is just a pattern you haven’t recognized yet. The pattern here is clear: every time a macro narrative enters crypto, it gets distorted into a shorter, more extreme cycle. The AI compute narrative will burn out faster than it built up.
### Takeaway: Decode the Script Before You Bet on the Actor So where does this leave us? Visser’s article is useful as a thought experiment—a case study in how narratives are constructed to serve specific interests. But as an investment thesis, it’s dangerously overconfident. The core insight for crypto investors is not to bet on the compute shortage, but to track its narrative decay. The next narrative shift will be toward decentralized AI compute (not tokenized, but genuinely decentralized via zk-proofs or federated learning) and toward agent security. That is where the real alpha lies.
Position for the decay, not the hype. When the compute bubble bursts, the question won’t be “Did you believe the story?” but “Did you read the data?” I don’t write to comfort. I write to expose. Because the truth hides in the footnotes—and in this case, it’s a $217 billion phantom.
Decode the script before you bet on the actor.
