NVIDIA CEO Speaks on Bittensor | Weekly Bittensor Update
NVIDIA CEO Jensen Huang publicly acknowledged Bittensor for the first time at GTC 2026, calling it “our modern version of Folding@home.” $TAO surged past $301 in response, and the ecosystem delivered across the board. Chutes (SN64) powered inference for two more subnets, Score (SN44) landed a partnership with major football leagues, Targon (SN4) joined the NVIDIA Inception program, and BitMind (SN34) open-sourced autonomous deepfake detection research. Read the full weekly breakdown of everything that happened in the Bittensor ecosystem this week.
Big Events in the Bittensor Ecosystem
NVIDIA CEO Jensen Huang Acknowledges Bittensor on the All-In Podcast
The biggest Bittensor moment of the week came straight from GTC 2026 in San Jose. During a live conversation with Chamath Palihapitiya on the All-In Podcast, NVIDIA CEO Jensen Huang publicly acknowledged the network for the first time. Chamath described the Covenant-72B decentralized training run on Templar (SN3), and Jensen responded with five words that instantly framed Bittensor for a mainstream audience: “Our modern version of Folding@home.”
The market reacted fast. $TAO surged over 17%, daily trading volume hit $677 million, and three Bittensor subnet tokens ranked among the top eight gainers on CoinGecko. Shortly after, the Opentensor Foundation posted a correction noting the model actually has 72 billion parameters, not the 4 billion Chamath mentioned on air. As a result, the real achievement turned out to be 18x larger than described.
This marks the first time the CEO of the world’s most valuable semiconductor company has referenced Bittensor by name in a major public forum. For a full breakdown of the conversation, the technical details behind Covenant-72B, and what it means for the ecosystem, read our deep dive here.
Subnet Updates
Chutes (SN64) Powers Inference for Two More Subnets
Chutes (SN64) expanded its role as the go-to inference layer in the Bittensor ecosystem this week with two new integrations. First, RESI Labs (SN46) went live on Chutes infrastructure. The subnet builds institutional-grade property appraisals for US real estate and now benefits from lowest-cost compute, end-to-end encryption, and increased distribution to agents. In other words, one Bittensor subnet is building directly on top of another.
On top of that, Vocence (SN102) joined the fold as well. The team builds decentralized voice AI and has started deploying its PromptTTS models as private chutes. Voice remains one of the fastest-growing frontiers in open-source AI, and it now runs on decentralized compute through the SN102 x SN64 collaboration.
Both integrations highlight what the Bittensor flywheel looks like in practice. Subnets do not just coexist on the network. Instead, they actively build on each other, creating compounding value across the ecosystem.
Score (SN44) Partners With Eyeball to Bring Vision AI to Youth Football
Score (SN44) landed a real-world partnership this week that puts Bittensor-powered Vision AI on football pitches worldwide. Eyeball, a youth-football technology company working with clubs across the Premier League, Ligue 1, Serie A, Bundesliga, LaLiga, MLS, and the Saudi Pro League, announced a research collaboration with Score to bring computer vision directly to grassroots player development.
The partnership centers on Manako, a platform powered by Score’s Vision AI models. Through Manako, Eyeball will build football-specific models capable of analyzing player behaviors and game intelligence directly from camera feeds. As a result, the system aims to work anywhere, from elite academies to local parks, giving every young player access to insights once reserved for top professionals.
This is one of the clearest examples yet of a Bittensor subnet driving adoption outside the crypto ecosystem entirely. Score’s decentralized computer vision infrastructure now solves a real problem in professional sports: turning raw match footage into objective, actionable insight at scale.
BitMind (SN34) Open-Sources Autonomous Deepfake Detection Research
BitMind (SN34) released dfresearch, an open-source framework that turns AI agents into autonomous deepfake detection researchers. The concept is simple but powerful: give an AI agent a training setup, let it experiment on its own, and wake up to a better model. Each experiment runs for 10 minutes, so the agent can test roughly 50 ideas overnight without any human input.
The framework covers all three media types that matter in deepfake detection: images, video, and audio. It ships with six baseline models out of the box, including EfficientNet-B4, CLIP ViT-L/14, VideoMAE, and Wav2Vec2. On top of that, datasets are pulled directly from BitMind’s gasbench benchmark at runtime. As a result, training data stays automatically in sync with the competition evaluation. The current dataset library spans over 60 image datasets, 45+ video datasets, and 35+ audio datasets covering real, synthetic, and semisynthetic content.
For Bittensor miners on Subnet 34, dfresearch lowers the barrier to competition significantly. The entire workflow fits into four commands: install, download, train, and export. Models that score above 80% accuracy pass the entrance exam and can be pushed directly to the network. The repository is available on GitHub.
BitMind (SN34) Mind Security Brings Deepfake Detection to AI Agents
2026 is the year of agents. As autonomous AI systems become the biggest digital customers, old cybersecurity assumptions break down fast. Prompt injection, malicious URLs, and fake media all become trivial exploits. BitMind (SN34) responded this week with Mind Security, an open-source skill published on ClawHub that builds verification and detection directly into the agent itself.
The toolkit uses BitMind’s deepfake detection API powered by Bittensor Subnet 34 as its core module. On top of that, it includes prompt injection scanning, malware/phishing URL detection, and AI-generated text identification. Agents can verify whether images or videos are real or synthetic in real time. The skill supports direct file uploads as well as URLs from YouTube, X, and TikTok.
For the Bittensor ecosystem, this signals a new category of demand. Subnets are no longer just serving human users. Instead, they are becoming infrastructure that other AI systems depend on. Mind Security is open-source under an MIT license and available on ClawHub.
Targon (SN4) Accepted Into NVIDIA Inception Program
Targon (SN4) announced this week that it has been accepted into the NVIDIA Inception program for startups. The collaboration will help Targon grow and improve its Confidential NVIDIA GPU experience on targon.com. As a result, this is another direct link between a Bittensor subnet and one of the biggest hardware companies in the world.
TAO Market Update
Price: $243.91 – $301.54
Weekly: +23,63%
Ranking: #30
Market Cap: $3.07B
24h Volume: $463M
Top Gainer Subnet: AlphaCore (SN66)
AlphaCore (SN66) is building a decentralized marketplace for autonomous DevOps agents on Bittensor. Miners compete by completing real infrastructure tasks like Terraform provisioning on Google Cloud, while validators verify outcomes inside Firecracker microVM sandboxes. The subnet plans to expand to additional clouds and task types over time.
Weekly change: +64.93%
Price: $1.948049
Market Cap: $7.69M
Volume (24h): $1.50M
FAQ:
During a live conversation at GTC 2026, NVIDIA CEO Jensen Huang called Bittensor “our modern version of Folding@home.” He made the comment after Chamath Palihapitiya described the Covenant-72B decentralized training run on Templar (SN3). This marks the first time the CEO of the world’s most valuable semiconductor company has referenced the network by name in a major public forum.
$TAO rallied sharply after the Jensen Huang mention on the All-In Podcast. The token briefly touched $301 and trading volume spiked to over $727 million in a single day. As a result, three Bittensor subnet tokens also ranked among the top daily gainers on CoinGecko.
Covenant-72B is the largest decentralized LLM pre-training run ever completed. Over 70 independent contributors trained a 72-billion-parameter model on 1.1 trillion tokens using regular home internet connections. The training happened entirely on Bittensor Subnet 3 (Templar) and finished on March 10, 2026. For a full breakdown, read our deep dive here.
This week showed several examples of the Bittensor flywheel in action. Chutes (SN64) now powers inference for both RESI Labs (SN46) and Vocence (SN102). On top of that, BitMind (SN34) released tools that other subnets and agents can use for deepfake detection. Instead of operating in isolation, Bittensor subnets increasingly build on each other’s infrastructure.
Score (SN44) partnered with Eyeball, a youth-football technology company working with clubs across the Premier League, Bundesliga, LaLiga, and other major leagues. Additionally, Targon (SN4) was accepted into the NVIDIA Inception program for startups. Both deals show Bittensor subnets gaining traction outside the crypto ecosystem.
