Bittensor’s decentralized AI network delivered its most concrete technical milestone yet this week, as a 72-billion-parameter language model trained entirely across its subnet infrastructure validated what skeptics said couldn’t be done, while institutional players moved to formalize TAO exposure through regulated financial products.
TAO surged 17% on March 20, breaking above $300 for the first time since January. The rally capped a V-shaped recovery from a February 11 low of $143, representing roughly 110% gains in five weeks. At press time, TAO traded at $268.56 with a market cap of approximately $2.58 billion.
Covenant-72B Proves Decentralized AI Training at Scale
The catalyst behind this week’s momentum is Covenant-72B, a 72-billion-parameter large language model trained permissionlessly on Bittensor’s Subnet 3, known as τemplar. More than 70 global contributors used commodity internet hardware to collectively train the model on a 1.1-trillion-token dataset.
Covenant-72B achieved a 67.1 MMLU score, a widely used benchmark for evaluating language model reasoning. A March 2026 arXiv paper confirmed it as the largest decentralized LLM pre-training run on record.
The distinction matters. Centralized AI labs like OpenAI and Google rely on proprietary clusters of high-end GPUs costing hundreds of millions of dollars. Covenant-72B demonstrated that a permissionless network of independent contributors can produce competitive results without centralized coordination or hardware monopolies.
Nvidia CEO Jensen Huang lent unexpected validation to the approach during an appearance on the All-In Podcast, endorsing both proprietary and open-source AI as complementary forces. He framed decentralized training as benefiting broader AI adoption rather than threatening incumbent players.
Chamath Palihapitiya, also on the podcast, was more direct. He called the Covenant-72B result “a pretty crazy technical accomplishment,” a characterization that resonated across crypto and AI communities.
The τemplar subnet token responded accordingly, surging approximately 194% in a single week. Bittensor’s broader network now operates 128 independent subnets, each focused on different AI tasks ranging from text generation to image processing and coding.
Institutional Infrastructure Around TAO Takes Shape
The technical breakthrough arrived alongside a parallel institutional buildout that has been developing since late 2025. Grayscale filed Form S-1 with the SEC on December 30, 2025, seeking to convert the Grayscale Bittensor Trust into a spot exchange-traded product listed on NYSE Arca under ticker GTAO.
The filing includes plans to stake the fund’s TAO holdings, a detail that distinguishes GTAO from most existing crypto ETF applications. A staked TAO ETP is already live on SIX Swiss Exchange, having launched in October 2025, giving European investors a head start on regulated TAO exposure.
Derivatives markets reflected the shift in positioning. Futures open interest climbed from $131.9 million on March 4 to $361.1 million by March 17, a 174% increase in under two weeks. That kind of open interest expansion, particularly when accompanied by spot price appreciation, typically signals new capital entering rather than existing positions being leveraged up.
The pattern echoes what played out with institutional Ethereum accumulation earlier this year, where regulated products and custody infrastructure preceded sustained capital inflows. Whether TAO follows the same trajectory depends on the SEC’s response to Grayscale’s application and the broader regulatory appetite for AI-sector crypto products.
What Comes Next for Bittensor’s Momentum
Three concrete data points will test whether this week’s developments mark a turning point or a temporary spike.
First, the Grayscale GTAO application. The SEC’s review timeline will determine whether U.S. institutional capital can access TAO through traditional brokerage accounts. Given the precedent set by spot Bitcoin and Ethereum ETF approvals, a decision or comment period is expected within the standard 240-day window.
Second, Covenant-72B’s real-world adoption. The model’s benchmark score positions it competitively, but decentralized AI must demonstrate practical utility beyond proof-of-concept. The arXiv paper’s peer reception and any downstream deployment announcements will matter more than the initial MMLU number.
Third, whether TAO’s price action can hold above the $250-$300 range. The token remains 64.6% below its all-time high of $757.60 set in March 2024. The broader crypto market’s Fear & Greed Index sits at 12, deep in “Extreme Fear” territory, a stark contrast to the meme coin volatility that has dominated recent sentiment cycles. TAO’s rally against that macro backdrop makes the futures positioning data more notable.
Similar to how large whale accumulation patterns have preceded sustained moves in other major tokens, the open interest surge from $131.9 million to $361.1 million suggests conviction among larger players positioning for continued upside.
Bittensor’s network of 128 subnets represents a fundamentally different architecture than competing AI tokens that rely on single-purpose smart contracts. The Covenant-72B result gives that architecture its first undeniable proof point. The Grayscale filing gives it a potential capital pipeline. What remains is execution on both fronts simultaneously.
Disclaimer: This article is for informational purposes only and does not constitute financial or investment advice. Cryptocurrency and digital asset markets carry significant risk. Always do your own research before making decisions.
