Bittensor TAO After Covenant AI: Revenue Gaps, ETF Catalyst, and the $2.4B Valuation Question

Bittensor TAO sits at one of the most important crossroads in decentralized AI. In April 2026, the token carried a market cap of roughly $2.4 billion, while identifiable demand-side revenue across its active subnets appeared to remain between $3 million and $15 million annually. That gap was already difficult to ignore, but Covenant AI’s exit turned it into the central question facing TAO investors.

The issue is not whether Bittensor has a powerful narrative. It does. The issue is whether the network’s subnet economy, governance structure, and external revenue can grow fast enough to support the valuation implied by TAO’s market price. With Grayscale pursuing a TAO ETF, emissions falling after the first halving, and governance concerns moving into the open, the investment thesis now depends on more than decentralized AI hype.

What Is Bittensor TAO and How Do Subnets Work?

Bittensor operates as a decentralized marketplace where specialized AI networks, called subnets, compete for TAO token emissions based on the value they deliver. Since the Dynamic TAO (dTAO) upgrade launched in February 2025, each subnet maintains its own token through an automated market maker that pairs the subnet’s Alpha token with TAO. When participants stake TAO into a subnet, they swap TAO for that subnet’s Alpha token, and the subnet’s share of daily emissions rises with net staking inflows.

The network distributes 3,600 TAO per day following the first halving in December 2025, which cut daily issuance in half from 7,200 TAO. Emissions split three ways: 41% to miners who perform AI work, 41% to validators who score output quality, and 18% to subnet owners. At current prices, annualized miner emissions alone represent roughly $135 million in value flowing to subnet participants.

TAO functions as the base currency for every network interaction. Mining registration, validator staking, subnet token purchases, and service payments all require TAO, creating what should be structural demand for the token. The top 10 subnets currently control approximately 56% of all emissions, concentrating significant economic weight in a small number of teams.

Why Did Covenant AI Leave Bittensor?

On April 10, 2026, Covenant AI founder Sam Dare published an open letter announcing his team’s departure from Bittensor, calling the network’s governance structure a facade. Covenant operated three subnets, Templar (SN3), Basilica (SN39), and Grail (SN81), and had delivered what many considered the decentralized AI sector’s most significant technical achievement: Covenant-72B, a 72-billion-parameter language model trained across 70+ independent contributors using commodity hardware. The model scored 67.1 on the MMLU benchmark, placing it in competitive range with Meta’s LLaMA-2-70B, and its March 2026 release had driven TAO’s price up roughly 90% in the weeks that followed.

Dare’s allegations centered on Bittensor co-founder Jacob Steeves, known on-chain as Const. According to Dare, Steeves suspended emissions to Covenant’s subnets, overrode the team’s moderation authority in community channels, and applied economic pressure through timed token sales. A companion website called Tao Papers published on-chain forensics claiming that of 41 network upgrades between 2023 and 2026, 38 originated from infrastructure Steeves controlled, with co-signers approving within minutes and no public deliberation recorded.

Covenant simultaneously sold approximately 37,000 TAO worth over $10 million, triggering cascading liquidations. TAO fell from $337 to a low of $254 within hours, erasing nearly $900 million in market capitalization. Steeves responded by denying unilateral control and proposing a “Locked Stake” mechanism that would require subnet owners to lock tokens for defined periods as a measurable commitment signal.

The episode matters for a reason beyond the immediate price impact. Covenant was the living proof that Bittensor’s model could produce competitive AI. Its departure removed the network’s strongest narrative anchor and validated a critique that had been circulating for months: the free-rider problem.

Can TAO Revenue Support a $2.4B Valuation?

The gap between what Bittensor’s supply side promises and what its demand side delivers is the central tension in any TAO investment thesis.

On the supply side, the economics are transparent and compelling. TAO’s 21-million hard cap mirrors Bitcoin. The December 2025 halving created genuine scarcity, and with approximately 70% of circulating supply locked in staking, the effective float is limited to around 30% of tokens. These mechanics have historically correlated with price appreciation in crypto assets.

The demand side is opaque by design. AI service delivery, including inference requests, compute jobs, and training tasks, happens off-chain and is not recorded on the blockchain. No aggregated dashboard tracks external revenue across subnets. Investors rely on proxy metrics like staking flows, subnet token prices, and self-reported figures from individual teams.

The most complete demand-side picture available shows the network generating between $3 million and $15 million in annual external revenue across all 128 subnets. Chutes (SN64), the highest-traffic subnet operated by Rayon Labs, processes over 100 billion tokens daily and reports approximately $2.4 million in annual revenue against roughly $52 million in annualized emission subsidies it receives. That means for every dollar customers pay for Chutes’ inference services, the network contributes $22 to $40 in TAO emissions to subsidize operations. Targon (SN4), operated by Manifold Labs, projects approximately $10.4 million in annualized revenue, though this figure has not been independently audited.

Against a $2.4 billion market cap, even the generous end of the revenue range implies a multiple above 150x. Centralized AI infrastructure companies that raised capital in 2025 and early 2026 were valued at 15x to 25x forward revenue. The gap suggests TAO’s price reflects supply-side scarcity and sentiment rather than demonstrated economic productivity from AI services.

How Important Is the Grayscale TAO ETF Filing?

Grayscale filed an amended S-1 registration statement with the SEC on April 2, 2026, advancing its effort to convert the over-the-counter Bittensor Trust (GTAO) into a spot ETF listed on NYSE Arca. The trust currently holds approximately 2 million shares outstanding and uses Coinbase and BitGo as custodians. Just days before the Covenant AI crisis, Grayscale had increased TAO’s allocation to 43% of its AI-focused fund, the largest single-asset allocation increase for any holding in that fund.

The ETF application represents the first regulated vehicle of its kind for a decentralized AI token in the U.S. If approved, it would open TAO to financial advisors, institutional allocators, and retirement platforms that cannot hold crypto directly. The GBTC precedent, where Grayscale’s Bitcoin Trust conversion to a spot ETF in January 2024 preceded significant price appreciation, offers a template for what institutional access could mean for TAO’s liquidity and price floor.

The timing, however, is complicated. The governance crisis has introduced reputational risk at precisely the moment institutional credibility matters most. SEC reviewers evaluating whether TAO meets listing standards will weigh the concentration allegations and governance questions alongside the product’s technical merits.

What Are Bittensor’s Biggest Competitive Risks?

Bittensor subnets face margin pressure from two directions simultaneously. Self-hosting open-source models caps pricing from above, as every model deployed on Bittensor is available on Hugging Face, and a single H100 GPU can serve a 70B model at roughly $40 to $50 per day. Hyperscalers compress from below, with Microsoft, Google, Amazon, and Meta collectively investing over $200 billion in AI infrastructure during 2025 alone. Bittensor’s entire annual emission budget at current prices represents less than what a single major cloud provider spends on AI in a week.

The moat challenge compounds this pressure. Covenant-72B is Apache-licensed. The SparseLoCo training methodology was published on arXiv. Any competitor can replicate the approach without participating in the TAO economy. Traditional competitive advantages like proprietary technology, switching costs, and network effects do not apply in the same way when the core innovation is open source by design.

The counterargument draws on a historical parallel that crypto investors find persuasive. Bitcoin aggregated massive computing power through token incentives alone, with no company coordinating the effort. Ethereum’s proof-of-work era achieved similar results. Bittensor applies this coordination mechanism to AI resources. If AI’s hunger for compute, data, and talent remains insatiable, and if Bittensor can reliably supply these at scale, subnets may stay in the ecosystem voluntarily because leaving means losing access to a resource pipeline they cannot replicate on their own.

What Should TAO Investors Watch Next?

The weeks following the Covenant exit will test several hypotheses simultaneously. The most immediate signal is whether other prominent subnet teams follow Covenant out of the ecosystem. As of mid-April 2026, 129 subnets continue operating, and community members have organized to maintain work on the three subnets Covenant abandoned. If the network absorbs the departure without further defections, the event may be repriced as a temporary sentiment shock rather than a structural failure.

The “Locked Stake” governance proposal, which Steeves has attributed to work Dare himself completed before leaving the Opentensor Foundation, will go to community discussion. If implemented, it would create a time-locked commitment layer that makes sudden exits more costly and publicly visible. Whether this mechanism can substitute for the contractual obligations that traditional venture capital uses to retain portfolio company alignment remains an open question.

The Grayscale ETF timeline is uncertain but could provide a significant catalyst if it progresses. Any positive SEC communication during 2026 would add institutional buying pressure to a token with limited effective float. Conversely, a prolonged delay would remove the strongest near-term bullish catalyst from the thesis.

Finally, the demand-side revenue trajectory deserves the closest monitoring. Bittensor’s long-term value depends on whether subnets can transition from emission-subsidized operations to self-sustaining revenue models before the next halving further compresses incentive budgets. The current subsidy-to-revenue ratios suggest that transition is still in its earliest stages.

Key Takeaways

• Bittensor TAO’s April 2026 price crash of roughly 25% was triggered by Covenant AI’s departure and governance allegations, removing the network’s strongest narrative anchor while 129 subnets continue operating.
• Identifiable demand-side revenue across the entire Bittensor network falls between $3 million and $15 million annually, implying valuation multiples far above comparable centralized AI infrastructure companies.
• Grayscale’s amended S-1 filing for a GTAO spot ETF on NYSE Arca represents the first regulated institutional vehicle for a decentralized AI token, with approval potentially unlocking significant capital flows.
• The proposed “Locked Stake” mechanism aims to address the free-rider problem by requiring subnet owners to lock tokens, but its effectiveness as a substitute for contractual value-return obligations remains untested.
• TAO’s investment thesis splits into two distinct camps: one based on supply scarcity and institutional catalysts, and one based on Bittensor becoming a meaningful AI services network, and investors should be clear about which thesis they hold.

FAQ: Bittensor TAO Deep Dive

Why did Bittensor TAO crash in April 2026?

Covenant AI, the team behind Bittensor’s landmark Covenant-72B model (72 billion parameters, trained across 70+ independent contributors), announced its departure on April 10, 2026. Founder Sam Dare accused Bittensor co-founder Jacob Steeves of centralized control and sold approximately 37,000 TAO worth over $10 million, triggering cascading liquidations and a roughly 25% price drop.

How much revenue does Bittensor generate?

Identifiable external revenue across all 128 subnets falls between $3 million and $15 million annually as of early 2026. Chutes (SN64) reports roughly $2.4 million against approximately $52 million in annual emission subsidies. Targon (SN4) projects $10.4 million in annualized revenue, though independently audited figures have not been published.

What happened between Covenant AI and Bittensor?

Covenant AI founder Sam Dare alleged that Bittensor co-founder Jacob Steeves suspended emissions to Covenant’s subnets, overrode community moderation, and applied economic pressure through timed token sales. On-chain forensics published by the Tao Papers website claimed that 38 of 41 network upgrades between 2023 and 2026 originated from infrastructure Steeves controlled, with co-signers approving within minutes.

What is the Grayscale TAO ETF filing?

Grayscale submitted an amended S-1 registration with the SEC on April 2, 2026, seeking to convert its over-the-counter Bittensor Trust into a spot ETF listed on NYSE Arca under the ticker GTAO. If approved, it would be the first regulated ETF vehicle for a decentralized AI token in the U.S., potentially opening TAO to institutional allocators.

Is TAO overvalued compared with its revenue?

At a $2.4 billion market cap, even the upper end of the $3 million to $15 million revenue range implies a multiple exceeding 150x. Centralized AI infrastructure companies raised capital in 2025 and early 2026 at 15x to 25x forward revenue. The valuation gap suggests TAO is priced primarily on supply-side scarcity, institutional catalysts, and AI sector sentiment rather than demonstrated economic output.