Intel Engineers Said It Works
On March 23, 2026, a paper dropped that most of the Bittensor community missed: "Decentralized Compute on Untrusted Hardware Using Intel TDX and Encrypted CVMs" — co-authored by Intel engineers Haidong Xia and Sathi Nair alongside the Manifold Labs team behind SN4 Targon.
This is not a blog post or a tweet thread. This is a published technical paper on Intel's official platform. Intel put their name on it.
The implication: decentralized AI compute can be trusted, even when running on hardware you do not control. That is the single biggest barrier to enterprise adoption of decentralized infrastructure, and Intel just said it is solved.
What the Paper Actually Proves
The core innovation is using Intel TDX (Trust Domain Extensions) with encrypted Confidential Virtual Machines (CVMs) to create a trusted execution environment on untrusted hardware. In practical terms:
- A hedge fund can run AI inference on a miner's GPU without the miner seeing the data
- A government agency can process classified workloads on decentralized infrastructure
- A healthcare company can run models on patient data without HIPAA exposure
The paper demonstrates this working on SN4 Targon's infrastructure — 1,500+ H200 GPUs, $10.5M in Series A funding, and a live incentivized compute marketplace.
Two Subnets, Two Approaches to the Same Problem
Here is where it gets interesting for Bittensor investors. Two subnets are independently solving the same trust problem with the same Intel technology — and they are approaching it very differently.
| Metric | SN4 Targon | SN64 Chutes |
|---|---|---|
| TEE Technology | Intel TDX + CVMs | Intel TDX + NVIDIA CC |
| Pool TAO | 131,635 | 205,117 |
| EMA Price | 0.056831 | 0.078120 |
| Revenue Model | Compute marketplace | Inference-as-a-service |
| GPU Fleet | 1,500+ H200s | 700+ H200s |
| Daily Revenue | Growing | ~$22K/day |
| Intel Validated? | YES — co-authored paper | Uses same tech, no paper |
| Smart Contract Vault | No | Yes (Trust Security audited) |
SN4 Targon is the compute marketplace — you bring your workload, they match you with hardware. Intel validated their approach. They have the academic credibility.
SN64 Chutes is the production beast — $22K/day in real revenue, smart-contract vault to prevent rug pulls, and the largest pool depth on the network at 205K TAO. They use the same Intel TDX technology but have not published a paper.
Why This Matters for the Entire Network
When Intel co-authors a paper validating your technology, it signals something to enterprise buyers: this is not experimental anymore. The two largest objections to decentralized compute have always been:
- "Can I trust the hardware?" — Intel says yes.
- "Is it production-ready?" — SN64 doing $22K/day says yes.
Between SN4 and SN64, Bittensor now has both the academic validation and the production proof that decentralized AI compute works, is secure, and is commercially viable.
That is what NVIDIA saw when they deployed $420M in TAO in Q1 2026. That is what Barry Silbert saw when DCG launched Yuma. The smart money is not speculating on whether this works — they are deploying capital because Intel already proved it does.
Track Both Subnets
We monitor SN4 and SN64 in real-time on our Live Dashboard. Check their pool depth, EMA prices, emission share, and deregistration risk on our Subnet Liquidation Risk page.