Author｜ Momir @IOSG
The bullish logic for TAO requires you to believe that a game theory miracle can work. But such miracles have happened before in the cryptocurrency industry.
Bittensor possesses one of the most elegant narratives in the cryptocurrency space: a decentralized AI marketplace where market mechanisms allocate funds to the most impactful research. TAO is the coordinating layer, subnets are the labs, and the marketplace is the funding committee.
Strip away the narrative, and you'll find something more unsettling.
Bittensor is a funding program where cryptocurrency speculators fund AI research—and those funded are not obligated to return any value to TAO.
Bittensor is a funding program where cryptocurrency speculators provide funding for AI research—and those funded are not obligated to return any value to TAO.
Think of TAO as Elon Musk—the first investor in OpenAI, a “non-profit” company. Subnets are like Sam Altman—they are the builders who receive funding and deliver the product, but have no contractual obligation to share the profits. They may eventually choose to privatize the profits without returning any value to their original funding sources. Bittensor distributes TAO tokens to subnet operators and miners based on the price of the subnet token. Once a subnet receives a TAO allocation, there is no mandatory mechanism requiring that the AI models, datasets, or services it generates remain within the Bittensor ecosystem. Subnet operators can use Bittensor's TAO incentives to generate revenue and then move the real product elsewhere—to centralized cloud servers, package it as a standalone API, or simply sell it as a SaaS. TAO has no equity or licensing agreement. The only binding is the subnet token—the token price must be maintained to ensure access to resources. But this only works before the subnet "flies out": once the product is strong enough to stand on its own outside the Bittensor system, this tether breaks. The relationship between Bittensor and the subnet is less like venture capital and more like research funding—it gives you start-up capital, but you don't get shares. To put it bluntly, Bittensor is essentially a wealth transfer: from the pockets of token speculators to the accounts of AI researchers—or more bluntly, from retail investors to tech-savvy "miners." The principle is simple: TAO investors are essentially underwriting the entire ecosystem. They buy and hold TAO, supporting the price, which itself serves as a conduit for funds to flow into the subnet incentive system. Subnet operators earn TAO inflation rewards by "demonstrating performance"—but in reality, this "performance demonstration" largely means maintaining a favorable price for their subnet's token. AI products built with this funding can be abandoned at any time—the only constraint is that they still need to continuously acquire network resources. This is the nightmare VCs fear most: you provide the funding, they develop the product, but they owe you nothing. All that's left is a token issuance schedule and a prayer. Now let's look at it from a different angle. The optimistic view rests on two pillars: The persistent need for resources means AI companies always face funding shortages. Computation, data, and talent are expensive. If Bittensor can reliably provide these resources at scale, subnets have a legitimate incentive to remain—not because they are locked up, but because leaving means losing access to these resources. Logically, there's a soft support: AI's demand for resources is endless, and TAO's capacity is unattainable through self-financing alone. Following this logic, the subnet team will proactively maintain its token valuation; without any coercive mechanisms, the TAO economy will spontaneously form a positive flywheel. Cryptocurrencies excel at resource aggregation. Bitcoin aggregates massive computing power solely through token incentives. Ethereum's proof-of-work mechanism has also achieved tremendous success, becoming a powerful magnet for computing resources. Bittensor is applying the same strategy to the field of artificial intelligence. The "enforcement mechanism" is precisely the token game itself—as long as TAO has value, the incentive to participate will continuously increase. If 1000 simulations of Bittensor's future were performed, the distribution of the results would be extremely skewed. In most simulated scenarios, Bittensor will remain a niche funded project. The AI output from subnets will be insignificant. The best-performing subnets will gain significant attention, reap rewards, and then revert to closed-source mode, leaving no value for TAO. When the token supply exceeds the value created, the TAO token will depreciate. In a few simulated paths, something actually works. A subnet creates a truly competitive AI service, and the network effect begins to snowball. TAO becomes a true coordinating layer for decentralized AI infrastructure—not by forcibly capturing value, but by the inherent attraction of being a reserve asset in a functioning AI economy. In very rare cases, TAO becomes the defining asset class of an entirely new entity. Where might problems arise? The bearish logic is simple: There's no stickiness. Once the subnet no longer needs TAO token incentives, it will leave. Bittensor is a transitional phase, not the final destination. Centralized AI has an overwhelming advantage. Companies like OpenAI, Google, and Anthropic possess orders of magnitude in computing power and talent. TAO cannot match the strength of the venture capital and private equity markets. Therefore, the most talented individuals will choose the traditional development path. Issuance is taxation. TAO's issuance plan subsidizes subnets by diluting holders. If the value created by a subnet doesn't match this dilution, it's a slow, painful drain disguised as a "growth mechanism." The optimistic scenario, frankly, is more wishful thinking than a truly viable path to success. The majority of capital invested in TAO will ultimately subsidize development activities that don't return value to token holders. However, Crypto has repeatedly demonstrated that token-incentivized coordinated games can produce outcomes unpredictable by any rational model. Bitcoin shouldn't have succeeded, but it did—although this argument itself is weak, and the industry has used it to endorse numerous projects that don't stand up to first-principles scrutiny. The core problem with TAO isn't whether a coercive mechanism exists—it doesn't, and dTAO's efforts haven't changed that. The core problem is: are the game-theoretic incentives strong enough to keep the best subnets on track? Buying TAO is betting that a "soft guarantee" will hold true in harsh reality. This is either naiveté or foresight.
Anais
Catherine
Kikyo
Weatherly