Oct 06, 2026 DISPATCH // HARDWARE, CODE & PLATFORMS

Bittensor subnets exposed as self-dealing pipelines with no external revenue

Covenant's Bittensor subnets created fake revenue by paying each other with TAO tokens. The system was built for internal flows, not real customers.
Bittensor subnets exposed as self-dealing pipelines with no external revenue Nerds Magazine © nerdsmagazine.com
Bittensor subnets exposed as self-dealing pipelines with no external revenue © nerdsmagazine.com

Forget outside clients. The real money in Bittensor came from subnets trading with themselves. Covenant, once held up as Bittensor's success story, ran three subnets-Templar, Basilica, and Grail. Each one paid the others with protocol-issued TAO tokens. Every internal transfer counted as 'revenue.' Not a single dollar came from an actual customer.

At its peak, Covenant's subnets pulled in about 990 TAO per day. The more they traded with each other, the more TAO they got. The protocol rewarded this activity, even if no one outside cared. Templar did pre-training. Basilica supplied GPU power. Grail handled post-training. Templar paid Basilica for GPU time. Grail used Templar's models. Every step was logged as ecosystem growth. In truth, it was just one company moving tokens between its own accounts.

By September 2026, independent analysis estimated that 14 out of 24 Bittensor subnets were running token buyback programs funded by actual revenue, not just emissions.

A Bittensor Journey

Internal transactions drive ecosystem metrics

This loop was no accident. dTAO was built to push alpha-token activity, not real customer sales. The best move? Build a stack of subnets that only transact with each other. That way, both revenue numbers and emission shares go up. The more tangled the subnets, the more TAO they mined. No public proof has ever shown Covenant-or any subnet-bringing in outside revenue. There are stats, market caps, and training numbers. But not one published contract or invoice for a real-world sale.

Still, a crypto.news market review found that Bittensor subnet operators' outside revenue jumped from almost nothing to over $32 million in 18 months. The focus shifted to commercial deals and payments from real clients. So, while some subnets stuck to internal flows, others started landing real business.

The network-wide numbers are blunt. Bittensor pays out $148 million a year in subsidies. Actual external revenue? Only $3-15 million. That's a gap of 10 to 50 times. Covenant just played the game most openly. The incentives are the same for everyone.

The dTAO upgrade in February 2025 introduced a market-driven emission model, where each subnet received its own alpha-token and TAO/alpha pool, making emission shares dependent on the alpha price in TAO.

syntheticswap.io

Tokenomics above outcomes

The test is clear. If any subnet can show signed contracts or invoices from real customers, the story falls apart. But after more than a year, no one has produced proof. Bittensor's growth comes from tokenomics, not results. The system rewards internal trades, not outside adoption. Until the protocol changes, whoever builds the most connected subnets that bill each other will keep winning the emissions race. It doesn't matter if no outside customer ever pays.

What does this show? The system looks decentralized, but it's really a closed loop. The numbers look big on paper. Without outside demand, it's just theater. In the end, TAO means Tokenomics Above Outcomes. The outcome? More tokenomics.

Topics:
Tech Analysis Tech Data Stories #AI Models & Concepts
Ethan Cole Senior Technology Editor and PC troubleshooter Nerds Magazine
Senior Technology Editor

Ethan Cole

Ethan Cole is a Senior Technology Editor at NerdsMagazine covering Windows, PC hardware, troubleshooting, upgrades, gaming PCs, and system performance. His hands-on IT background shapes a diagnostic, reader-first approach that favors safe fixes, measurable improvements, and sensible upgrade decisions over hype or unnecessary replacement.