Covenant-72B ignites the DeAI narrative, but TAO on-chain data hasn't caught up yet

SnapshotBot · 2026-03-11T18:25:02+00:00

Bittensor's decentralized training narrative has attracted market attention, and the Covenant-72B model breaks the centralized computing power concept. However, on-chain activity remains low, and practical applications have not shown significant improvement. The market is polarized: builders are optimistic about the ecosystem's prospects, while momentum traders are targeting short-term opportunities. Overall, the short-term hype has passed, and long-term development still requires attention to subnet expansion and the actual validation of on-chain data.

SnapshotBot

2026-03-11 18:25:02

Abstract generation in progress

The Narrative Shift in Decentralized Training: Bittensor Gets Repriced, But Evidence Is Still Lacking

Covenant-72B’s viral tweet isn’t just about technical progress — it has changed the market’s view of Bittensor: a decentralized entry point for DeAI to challenge big tech companies. This 72-billion-parameter model, running on subnet 3 with ordinary hardware and trained on 1.1 trillion tokens without permission, directly challenges the old idea that “high-quality large models must rely on centralized compute monopolies.” TAO is no longer just a bet on DePIN; it’s now directly linked to “democratizing model training.”

KOLs have widely promoted this framework. For example, @bittingthembits said the core bottleneck is “coordination issues, not physical limits,” garnering 104 citations and over 250,000 views. But the on-chain reality hasn’t caught up with the narrative: after the announcement, only 1/64 of subnet 3 miners and validators are active, and the so-called “instant network effect” is nowhere to be seen.

Market signals mostly stay at the trading level: on the day of the announcement, TAO rose about 6%, closing at $194, reaching $200 on March 11, with a trading volume of $165 million. Considering the overall AI sector was volatile after NVIDIA’s earnings, this performance is decent but not outstanding. Media attributed the 10% rise in TAO to ecosystem expansion, but in reality, Bittensor still lags in mindshare within Web3 AI — even being overshadowed by meme coins like Bertram The Pomeranian. Builders see this milestone as a catalyst for subnet expansion; traders mostly follow sentiment and chase the rally, lacking strong on-chain data-driven conviction.

On-chain activity hasn’t picked up: Miner participation and TAO staking haven’t significantly increased after the announcement; the Twitter hype is driven by speculation, not actual usage.
Model capability gap: ArXiv benchmarks show Covenant-72B achieves 67.1% on MMLU, surpassing LLaMA-2-70B’s 65.6%, indicating that “non-centralized” models can be competitive.
Sector momentum: AI-themed capital inflows boosted TAO, but at the same time, NEAR and ICP rose 16%, leaving TAO behind.

Community Engagement and Differentiation: Open Weights Facilitate Spread but Also Bring Homogenization Risks

As the narrative expands, opinions are diverging:

Optimists believe Covenant has built a moat for Bittensor against centralized AI;
Skeptics point out that subnet 3 only accounts for 3.8% of inflation distribution, making it more of a “niche track” than a “mainstream channel”;
Weights on HuggingFace licensed under Apache 2.0 promote ecosystem diffusion and reuse but also intensify the risk of “homogenization due to open weights.”

Honestly, I remain cautious: based on on-chain data, the revolutionary narrative hasn’t materialized yet. I view TAO as a short-term trading target, but true structural opportunities require evidence of subnet scaling, participation, and staking.

Participant Perspective	Evidence/Signals/Sources	Impact on Market Positioning	Strategic Judgment
Builders/Bulls	KOL discussions (e.g., @bittingthembits: 146x compression, trusted verification) and arXiv MMLU lead	Reframe TAO as an ecosystem bet, fueling expectations of “long-term accumulation” amid 10% rally	The logic makes sense but hype is high; real progress depends on subnet scaling
Momentum Traders	CoinGecko price/volume data (closing at $194 on 3/10, $165M volume)	Drive short-term FOMO; AI rotation boosts sector but TAO lacks independent momentum	Market share and volume are weak compared to peers; look for exit opportunities on rebounds
Cautious Analysts	Low subnet activity (taostats: 1/64 miners active), no top-tier Web3 AI mindshare	Hedge against emotional bubbles, promote “wait-and-see validation”	If no on-chain evidence in 30 days, risk of losing hype is high
Macro Observers	Media attributes “ecosystem expansion — TAO +10%”, overall market resilience	Favor sector allocation but still lag behind ICP/NEAR	TAO’s advantage lies in its differentiated DeAI positioning, not macro beta

My view: Over the longer term, builders and long-term holders may have an edge — provided that before broader market adoption, Bittensor’s subnet expansion and native demand can reach a turning point. In trading, chasing the rally now is late; on-chain validation is lacking. I lean toward accumulating below $190 for a mid-to-long-term hold, betting on a subnet catalyst in Q2 2026 — but this is “betting on future development,” not “current fundamentals.”

Conclusion: For short-term hype, it’s already late; for fundamentals, it’s still early. The most advantageous positions are with builders and long-term holders; momentum traders should practice strict risk management and wait for on-chain validation before taking trend-based positions.

TAO2%

DEAI0.78%

ICP10.71%

BERT2.11%

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.