TAO is Elon Musk, who invested in OpenAI; Subnet is Sam Altman.

Author: Momir, IOSG

TAO’s bullish logic requires you to believe that a game theory miracle can happen. But the crypto industry has seen such miracles before.

Bittensor has one of the most elegant narratives in the cryptocurrency space: a decentralized AI intelligent market, where market mechanisms allocate funds to the most influential research. TAO is the coordination layer, subnetworks are laboratories, and the market is the funding committee.

Removing the narrative veneer, you’ll find something more unsettling.

Bittensor is a funding scheme where crypto speculators provide capital for AI research—while the funded parties have no obligation to return any value to TAO.

You can think of TAO as Elon Musk—he’s the first investor in OpenAI, a “non-profit” enterprise. Subnetworks are like Sam Altman—they’re builders who receive funding and deliver products, but have no contractual obligation to share profits. They might ultimately privatize the gains without returning any value to the original funding source.

Bittensor distributes TAO tokens to subnet operators and miners based on the price of the subnet tokens. Once a subnet receives a TAO allocation, there’s no enforced mechanism requiring its AI models, datasets, or services to stay within the Bittensor ecosystem. Subnet operators can use Bittensor’s TAO incentives to milk the system, then move the real products elsewhere—host on centralized cloud servers, package as standalone APIs, or directly sell as SaaS.

TAO has no equity stake, nor does it have licensing agreements. The only binding factor is the subnet token—its value must hold up to maintain access to resources. But this only works before the subnet “flies out”: once the product is robust enough to stand outside the Bittensor system, that tether breaks. Bittensor and the subnet relationship is less like venture capital and more like research funding—giving you startup capital, but not your equity.

To put it bluntly, Bittensor is essentially a wealth transfer: from token speculators’ pockets to AI researchers’ accounts—or more plainly, from retail investors to technically savvy “miners.”

The principle is simple:

TAO investors are footing the bill for the entire ecosystem. They buy and hold TAO, supporting the token price, which in turn becomes the pipeline for capital inflows into the subnet incentive system.

Subnet operators earn TAO inflation rewards by “showing performance”—but in reality, “performance” largely means keeping their subnet token prices looking good.

The AI products built with this capital can walk away at any time—the only constraint is that they still need to keep acquiring network resources.

This is the nightmare venture capitalists fear most: you fund the project, the product gets built, but they owe you nothing. All that remains is a token issuance schedule and a prayer.

1. The Optimists’ Perspective

Now, from a different angle. The optimistic view rests on two pillars:

Persistent resource demand keeps AI companies in a constant state of funding shortage. Computing, data, and talent costs are high. If Bittensor can reliably provide these resources at scale, subnetworks have a strong incentive to stay—not because they’re locked in, but because leaving means losing their resource supply channel. Logically, there’s a soft support: AI’s endless resource needs, and TAO’s scale that can’t be achieved through self-funding alone. Following this logic, subnet teams will proactively maintain their token valuation—no coercion needed—allowing the TAO economy to self-sustain and generate a positive feedback loop.

Cryptocurrencies excel at resource aggregation. Bitcoin, through token incentives alone, has aggregated massive computational power. Ethereum’s proof-of-work mechanism has also been hugely successful, acting as a powerful magnet for compute resources. Bittensor is applying a similar strategy to AI. The “enforcement mechanism” is the token game itself—so long as TAO remains valuable, participation will keep increasing.

If we run 1,000 simulations of Bittensor’s future, the distribution of outcomes will be highly skewed.

In most simulations, Bittensor remains a niche funding project. The AI results produced by subnetworks are insignificant. The best-performing subnetworks garner attention, claim rewards, then shift to closed-source models, leaving no value for TAO. When token issuance exceeds the value created, TAO will depreciate.

In a few simulations, something truly takes off. A subnet develops a genuinely competitive AI service, network effects snowball, and TAO becomes a true coordination layer for decentralized AI infrastructure—not through forced constraints, but by becoming a magnet as a reserve asset of a functioning AI economy.

In very rare cases, TAO becomes the defining asset class of a whole new category.

2. Where Things Might Go Wrong

The bearish case is straightforward:

Lack of stickiness. Once a subnet no longer needs TAO incentives, it will leave. Bittensor is a transitional phase, not the final destination.

Centralized AI giants dominate. Companies like OpenAI, Google, and Anthropic possess orders of magnitude more compute capacity and talent. TAO cannot compete with the deep pockets of venture capital and private equity markets. As a result, top talent will choose traditional career paths.

Token inflation is a form of tax. TAO’s inflation schedule dilutes holders to subsidize subnetworks. If the value created by the subnet doesn’t match this dilution, it’s a slow bleed disguised as a “growth mechanism.”

The optimistic scenario, frankly, is more wishful thinking than a viable success path.

3. Conclusion

Most capital invested in TAO will ultimately subsidize development activities that do not return value to token holders. But crypto has repeatedly shown that token-incentivized coordination games can produce outcomes no rational model can predict. Bitcoin arguably shouldn’t have succeeded, yet it did—despite the fact that this argument alone is insufficient, and the industry has used it to endorse many projects that fail basic first principles.

The core issue with TAO isn’t whether coercive mechanisms exist—because they don’t, and dTAO’s efforts haven’t changed that. The real question is whether game-theoretic incentives are strong enough to keep the best subnetworks on track. Buying TAO is betting on a “soft guarantee” holding up in a brutal reality.

Either way, it’s either naivety or foresight.