AI’s Coming Glut: What Crypto’s Bubble Cycle Teaches Tech Investors
Blockchain Capital researchers have drawn a striking parallel between the boom-and-bust dynamics of the crypto market and the current artificial intelligence frenzy, warning that AI investors may be walking into a familiar trap: the commoditization of a once-scarce resource.
The comparison centers on a structural lesson from crypto’s first cycle. Between 2017 and 2019, dozens of layer-1 blockchains raised billions on the promise of superior throughput, only to discover that block space itself was not a durable moat. As chains proliferated, supply of blockspace ballooned far faster than demand, and most of those networks — and their tokens — collapsed under the weight of their own overcapacity.
From Block Space to Compute
The report argues that AI is now running the same playbook. Foundation model developers have raised unprecedented sums on the premise that frontier models constitute a defensible, high-margin asset. But open-weight models are closing the performance gap at remarkable speed, inference costs are falling sharply, and the number of capable model providers keeps growing. The result, the researchers contend, is that models themselves are on a path toward commoditization — much like blockspace before them.
That does not mean AI is a bubble destined to pop in the same shape. It means the value is likely to migrate away from the raw output — tokens, inference, generic model access — and toward the layers that capture durable demand: proprietary data, distribution, vertical integration, and the applications that wrap models into products customers actually pay for.
Why This Matters for Crypto
For crypto investors, the parallel cuts both ways. On one hand, decentralized compute and GPU networks marketed as “the AI trade” face the same overcapacity risk as the layer-1 tokens of 2018. Token valuations tied to raw compute supply may prove as fragile as valuations tied to raw blockspace.
On the other hand, the report suggests that projects capturing the scarce complements to commoditized models — verifiable data, provenance, agent coordination, and settlement rails — could be the cycle’s real winners. That is where crypto’s core competencies, such as trust-minimized coordination and open data markets, intersect with AI’s genuine bottlenecks.
The Forward View
The lesson from crypto is not that AI will fail. It is that infrastructure booms tend to overshoot, and that the market eventually re-prices scarcity. Investors who mistake abundant supply for a durable moat — whether in blockspace or in models — tend to learn the same lesson twice. The question for both crypto and AI allocators is simple: which layer of the stack actually captures the value when the underlying resource becomes cheap?
