The Bittensor network has rebounded quickly from a major disruption triggered by the exit of Covenant AI from three critical subnets, even as its core token, TAO, suffered a sharp 38% decline. Community miners took charge, successfully restoring key subnets SN3, SN39, and SN81 using open-source implementations and without any intervention from a central operator. Despite the volatility, approximately 70% of all TAO tokens remained staked during the shake-up, reflecting continued commitment from network participants. Large spot outflows followed, with daily losses surpassing $70 million amid the crash.
Institutional momentum grows: Grayscale and Bitwise move on TAO ETFs
On April 7, Grayscale increased the TAO share within its AI-focused fund to 43.06%, making it the single largest asset reallocation in the product’s history. This adjustment took place only days before the larger market became aware of the impact from the Covenant AI withdrawal, raising speculation that Grayscale’s internal evaluations had identified underlying strengths in the network ahead of the disruption.
Prior to this, Grayscale filed an S-1 Amendment on April 2 for a spot TAO ETF listed on NYSE Arca, coinciding with a similar ETF filing by Bitwise. The U.S. Securities and Exchange Commission is currently set to review the filings, with an official decision expected in August 2026. However, several market participants are monitoring the price action, noting that significant moves in both Bitcoin and Ethereum historically occurred ahead of ETF approvals.
The $218–$240 range has emerged as a decisive area of support for TAO. Investor confidence received a further boost in March when GeneralTensor secured $5 million in funding, supported by a Goldman-backed fund along with DCG. The newly launched TAO Institute introduced a dedicated risk index for subnets in April, marking another step in institutional engagement with Bittensor’s ecosystem.
Protocol upgrades and community-led recovery shape future outlook
Bittensor’s upcoming core protocol change, BIT-0011—known as the Conviction Mechanism—will guide the network’s next development phase. The mechanism will require subnet founders and stakers to lock alpha tokens and accrue conviction scores in monthly cycles, with subnet ownership awarded to those with leading scores. Tokens remain locked and cannot be withdrawn until the score interval completes, supporting stability and longer-term incentives.
The community’s ability to reboot the impacted subnets entirely from open-source code, without guidance from the network’s original founders, was viewed as a meaningful demonstration of Bittensor’s resilience. Network emissions and reward structures continued operating normally, showing that disruption did not impact mainnet functions. Crypto analyst Karamata2_2 described this as a clear case of antifragility in action, and BIT-0011 is set to embed this architecture further into the protocol.
Teutonic, a decentralized AI company previously known as Templar, is planning a decentralized training run involving a 1-trillion-parameter model by late May. This anticipated milestone is aligned with the ETF’s highest expected visibility during the SEC review calendar, and may attract renewed attention if completed within that window.
Despite recent concerns around centralization and operator exits, active subnets like Chutes AI and TargonCompute demonstrate Bittensor’s operational depth. Chutes AI currently accounts for over 14% of daily network emissions and processes more than 50 billion tokens daily, while TargonCompute has contributed to Intel’s TDX whitepaper and projects over $10 million in annualized recurring revenue. With 128 subnets now active and plans to scale up to 256, the network’s ecosystem shows sustained growth.
Bittensor is an open, decentralized machine learning network that enables global participants to collectively train, serve, and monetize AI models using its TAO cryptocurrency. The protocol is designed around a modular subnet architecture, allowing specialized groups to form, operate, and compete for incentives in a scalable Web3 environment.
