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

Author: Momir, IOSG

The bullish case for TAO hinges on your belief in a game-theoretic miracle—and while such miracles are rare, the crypto industry has witnessed them before.

Bittensor boasts one of the most elegant narratives in crypto: a decentralized AI intelligence market where market mechanisms allocate capital to the most impactful research. TAO is the coordination layer, subnets are labs, and the market functions as a funding committee.

Strip away the narrative, and you’ll find something far more unsettling.

Bittensor is a grant program—crypto speculators fund AI R&D, while grantees bear no obligation to return value to TAO holders.

Think of TAO as Elon Musk—the first investor in OpenAI, that “nonprofit” entity. Subnets are like Sam Altman: builders who receive funding and deliver products, yet operate under no contractual obligation to share profits. They may ultimately privatize those profits without returning any value to their original funders.

Bittensor distributes TAO tokens to subnet operators and miners based on their subnet token’s price. Once a subnet receives TAO allocations, there is no enforcement mechanism requiring its AI models, datasets, or services to remain within the Bittensor ecosystem. Subnet operators can exploit Bittensor’s TAO incentives—then migrate their actual products elsewhere: onto centralized cloud infrastructure, repackaged as standalone APIs, or wrapped into SaaS offerings for monetization.

TAO confers neither equity nor licensing rights. The sole binding mechanism is the subnet token—its price must hold up to sustain access to network resources. But this only works *before* the subnet “flies away”: once a product matures enough to stand independently outside Bittensor, that tether snaps. The relationship between Bittensor and its subnets resembles research grant funding—not venture capital. You provide seed capital, but get no equity stake.

Put bluntly, Bittensor is fundamentally a wealth transfer: from crypto speculators’ pockets into AI researchers’ bank accounts—or, even more plainly, from retail investors (“the韭菜”) to technically skilled “miners.”

The mechanics are simple:

• TAO investors backstop the entire ecosystem. By buying and holding TAO, they prop up its price—which itself serves as the pipeline funneling capital into the subnet incentive system.
• Subnet operators earn TAO inflation rewards by “demonstrating performance”—though in practice, much of that “performance” simply means maintaining an attractive subnet token price.
• AI products built with these funds can exit at any time—the only constraint being their continued need for network resources.

This is every VC’s worst nightmare: you fund the development, the product ships—but legally, they owe you nothing. All that remains is a token emission schedule and a prayer.

I. The Bullish Interpretation

Now consider the alternative view. The bullish case rests on two pillars:

1. Persistent resource scarcity ensures AI companies remain perpetually underfunded. Compute, data, and talent are prohibitively expensive. If Bittensor can reliably supply these resources at scale, subnets have strong economic incentives to stay—not because they’re locked in, but because exiting means losing access to critical infrastructure. There’s a soft logical anchor here: AI’s appetite for resources is insatiable, and the scale TAO can mobilize exceeds what any single team could raise independently. Under this logic, subnet teams will organically maintain their token valuations—no enforcement required—and the TAO economy will self-sustain a virtuous flywheel.
2. Crypto excels at resource aggregation. Bitcoin aggregated massive computational power using only token incentives. Ethereum’s proof-of-work mechanism likewise succeeded spectacularly as a magnet for compute resources. Bittensor applies this same playbook to AI. The “enforcement mechanism” is the token game itself—if TAO retains value, participation incentives compound relentlessly.

If you ran 1,000 simulations of Bittensor’s future, outcomes would be extremely skewed.

In most simulations, Bittensor remains a niche grant program. Subnet-generated AI outputs are insignificant. Top-performing subnets attract outsized attention and rewards—then pivot to closed-source models, leaving zero value for TAO holders. When token issuance outpaces value creation, TAO depreciates.

In a minority of paths, something genuinely takes off. A subnet launches a truly competitive AI service, triggering network effects that snowball. TAO evolves into the de facto coordination layer for decentralized AI infrastructure—not by enforcing value capture, but by becoming the reserve asset of a functioning AI economy, wielding gravitational pull through utility alone.

In rare cases, TAO defines an entirely new asset class.

II. Where Things Could Go Wrong

The bearish logic is straightforward:

• No stickiness. Once a subnet no longer needs TAO incentives, it departs. Bittensor is a transitional phase—not a final destination.
• Centralized AI holds overwhelming advantages. Companies like OpenAI, Google, and Anthropic command orders-of-magnitude greater compute capacity and talent depth. TAO cannot compete with the firepower of traditional VC and private equity markets. Consequently, top talent will choose conventional career paths.
• Emissions are taxation. TAO’s emission schedule subsidizes subnets by diluting existing holders. If subnets fail to generate commensurate value, this is slow bleeding disguised as a “growth mechanism.”

The optimistic scenario, frankly, resembles wishful thinking more than a credible path to success.

III. Conclusion

Most capital invested in TAO will ultimately subsidize development activities that yield no value back to token holders. Yet crypto has repeatedly demonstrated that token-incentivized coordination games can produce outcomes no rational model predicts. Bitcoin “shouldn’t” have succeeded—but it did. That argument alone isn’t sufficient; the industry has also used it to justify countless projects that collapse under first-principles scrutiny.

TAO’s core issue isn’t whether enforcement mechanisms exist—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 highest-quality subnets onboard. Buying TAO is betting that a “soft guarantee” holds up in a brutal reality.

That bet is either naive—or visionary.