TAO's Free-Rider Dilemma: Crypto Speculators Pay, AI Researchers Leave

Author: Momir Amidzic

Translation: Deep Tide TechFlow

Deep Tide Guide: IOSG Ventures Managing Partner Momir Amidzic provides a calm analysis of Bittensor. His core argument is straightforward: TAO is essentially a funding program for AI research without any obligation of returns, and subnets can leave at any time after receiving funds. Optimistically, AI’s eternal hunger for resources might keep subnets around; pessimistically, this is a wealth transfer from token speculators to AI developers. The article is short but clearly explains Bittensor’s structural contradictions.

Bittensor has a very refined narrative: a decentralized AI intelligence marketplace that allocates funds to the most influential research through market forces. TAO is the coordination layer, subnets are laboratories, and the market is the grant committee.

Removing the narrative, what remains is less appealing.

Bittensor is a funding scheme where crypto speculators pay for AI research, and the funded parties have no obligation to return value to TAO.

Imagine TAO as Elon Musk, the first investor in “non-profit” OpenAI. The subnets are like Sam Altman, who takes the money, develops products, and there’s not a single line in the contract requiring them to share profits. Ultimately, they might privatize the profits and give nothing back to the original investors.

Bittensor releases TAO to subnet operators and miners based on the price of subnet tokens. Once a subnet receives its TAO allocation, there are no enforced mechanisms requiring the AI models, datasets, or services produced to stay within the Bittensor ecosystem. Subnet operators can create valuable products, take the TAO emission rewards, and deploy their products elsewhere—whether on centralized clouds, independent APIs, or packaging them as regular SaaS companies.

TAO has no equity, no licensing rights. The only tether is the subnet token; only if the token price performs well can it continue to access resources. But this only works before the subnet reaches escape velocity. Once the product is good enough to operate independently outside Bittensor, that tether is broken. The relationship between Bittensor and subnets is more akin to research funding than venture capital.

From this perspective, Bittensor is a wealth transfer from token speculators to AI researchers. To put it plainly, from speculators to technically skilled farmers.

The mechanism is simple:

TAO investors support the market price of TAO to provide capital.

Subnet operators earn TAO inflation rewards by “proving performance,” which essentially means maintaining the subnet token price.

AI products built with this capital can leave at any time; the only constraint is whether they still need resources.

This is a VC nightmare scenario: you invest, the company produces something, and then it owes you nothing. Only a token emission plan and a prayer remain.

Optimistic interpretation

Now flip the perspective. The optimistic argument rests on two pillars:

Persistent resource hunger. AI companies are always short of money. Computing power, data, talent—all are expensive. If Bittensor can stably provide these resources at scale, subnets have rational motivation to stay—not because they are locked in, but because leaving means losing access to resource pipelines. The soft guarantee is that AI always needs more resources, and TAO can provide a scale unmatched by standalone financing. Logically, subnet teams will actively maintain their token valuation, creating a positive feedback loop for TAO’s economy without any enforced mechanisms.

Crypto has a unique ability to aggregate resources. Bitcoin has aggregated enormous computing power solely through token incentives. Ethereum’s proof-of-work is an extremely successful magnet for computing power. Bittensor applies the same approach to AI. The “enforcement mechanism” is the token game itself—if TAO has value, participation is incentivized.

Running 1,000 future simulations of Bittensor’s paths, the distribution would be heavily skewed.

In most paths, Bittensor remains a niche funding scheme. Subnets produce marginalized AI results. The best-performing ones gain some traction, but after earning rewards, they shift to a closed mode, giving no returns to TAO. Emissions exceed the value created, causing the token to decline.

In a few paths, something clicks. A subnet develops a truly competitive AI service. Network effects begin to compound. TAO becomes a meaningful coordination layer for decentralized AI infrastructure—not through enforcement, but by attracting value as a reserve asset within an active AI economy.

In very rare paths, TAO becomes a category-defining asset.

Potential issues

The bear case is simple:

Lack of stickiness. Once subnets no longer need emission rewards, they will leave. Bittensor is a stepping stone, not the destination.

Centralized AI wins. OpenAI, Google, and Anthropic control vastly more computing power and talent. TAO cannot compete with the depth of VC and PE markets. Top talent will follow conventional routes.

Inflation is a tax. TAO’s emission plan dilutes holders to fund subnets. If subnets do not generate corresponding value, this is a slow bleed disguised as a growth mechanism.

Frankly, the optimistic view is more wishful thinking than a realistic path to success.

Conclusion

Most of the funds invested in TAO will ultimately fund development that never returns value to token holders. But the crypto industry has repeatedly shown that token incentive mechanisms can produce outcomes that defy rational models. Bitcoin shouldn’t work, but it does. Still, this is a weak argument, as the industry uses it to endorse countless ideas that cannot stand up to first principles.

The problem with TAO isn’t about the absence of enforcement mechanisms—dTAO’s efforts can’t change that. The issue is whether game-theoretic incentives are strong enough to keep the best subnets on track. If you buy TAO, you’re betting on a hardcore world where a soft guarantee can hold.

Either that’s naive or visionary.