Neuron registration trends
The alpha price in τ, day by day, and what moved it each way.
Log change over the last 30 days −27.05%: protocol buys +6.31%, root sells −25.84%, users net −17.80%, cross-subnet +10.28%, unattributed −0.000000000000000000035%.
The chart follows the alpha price in τ. Each period starts at its opening price and steps through the mechanisms below, in a fixed order, to its closing price: a step that rises pushed the price up, one that falls pushed it down, and its height is how far it moved the price. A thin line carries each close into the next period’s open; it is dashed where the two differ. The faint candle behind each period is the market’s own open, high, low and close, for context.
With many periods on screen (90 days, or hours), the steps would be too thin to read, so the chart draws the price as a line instead and colours each period by the mechanism that moved it most. Hover a period for the full step-by-step breakdown.
Not here, because they do not move the spot price: the per-block TAO injection into the pool, and burns.
All figures are log changes, not simple percentage changes, so the parts add up exactly to the total; on big moves the two differ (a 10% log fall is a 9.5% price fall).
2026-11-04 UTC, as of ≈00:32 UTC (block 9,429,055)
The day isn’t over yet, so these figures keep changing. They are in no daily total, average or rank.
Organizations running sensitive AI workloads—inference on proprietary models, fine-tuning on confidential datasets, or multi-step ML pipelines—face a hard choice: centralized cloud providers give up data privacy and audit control; on-premises infrastructure is capital-intensive and operationally complex; public GPU marketplaces offer neither privacy guarantees nor hardware security.
A decentralized network of independently operated GPU clusters, each attested by cryptographic proof that data is isolated inside hardware-secured Trusted Execution Environments, can provide the privacy and verifiability of on-premises compute without the capital or operational burden. Participants (miners) earn emissions for providing attested capacity; consumers pay for the compute they use; validators score miners on infrastructure readiness and competitive pricing, creating incentives for both supply and demand to stay efficient.
KubeTEE (SN90 on Bittensor) implements this as a multi-cluster Kubernetes infrastructure where every workload runs inside Intel TDX or AMD SEV-SNP TEEs verified by CoCo remote attestation.
Miners register clusters (3+ nodes per cluster, one hotkey per cluster, nodes co-located in one data center), deploy Kata Containers + Confidential Containers with GPU passthrough, and earn emissions based on attestation + uptime + competitive pricing against Targon, Lium, and Chutes. Consumers submit jobs via OpenAI-compatible API, get cryptographic proof that code and data ran inside a verified TEE, and pay at a published resources price per hour (Alpha or TAO on-chain). All east-west traffic (LiteLLM to inference backends, Armada to workers) uses Trustee-issued mTLS certificates, not Kubernetes Secrets.
Auto-generated from the GitHub repository and may not reflect the latest changes.
KubeTEE AI Factory is a decentralized confidential computing subnet on Bittensor that provides SOTA AI services (inference, fine-tuning, batch jobs) running inside Trusted Execution Environments (Intel TDX/SGX, NVIDIA CC) using Kata Containers and Confidential Containers. Miners host multi-cluster Kubernetes infrastructure with hardware-secured GPU nodes; validators score miners on infrastructure readiness, TEE attestation, competitive pricing against other compute subnets, and job execution quality.
KubeTEE operates a multi-cluster Kubernetes RKE2 infrastructure spanning a subnet-owner control plane (validator, Armada Server, Trustee/KBS, staging cluster) and multiple miner clusters (one per hotkey, each in a single data center). Miners deploy 3+ TDX/SEV-SNP GPU nodes with 8-GPU passthrough per node; every pod runs under a confidential Kata + CoCo runtime class (guest debug off, Trustee attestation). LiteLLM gateway (llm.kubetee.ai, 3 TDX CPU replicas) routes inference to either in-cluster TEE-served NeMo/NIM models (GLM, Ornith, DeepSeek, MiMo, etc.) or remote miner backends via attestation-gated mTLS (Trustee-issued certificates after quote verification).
All east-west traffic is encrypted by default: LiteLLM client mTLS to NIM terminator, NIM inbound HAProxy to SGLang loopback. Secrets (TLS keys, NGC/HF tokens, sealed LUKS keys) live in Trustee KBS, never in Kubernetes Secrets or git. Batch jobs (fine-tune, eval, pipelines) are submitted to Armada queues (Armada Server on control plane, Executor on each miner cluster) with a confidential runtimeClassName; Armada scheduler dispatches them across clusters with fair-use queuing and gang scheduling.
Validator reads metagraph each epoch, queries Rancher for miner cluster state, fetches price signals (Taostats, Targon, Hippius S3), computes infrastructure + price-competitiveness scores, and sets on-chain weights once per epoch. Consumers can submit jobs via LiteLLM /fine_tuning/jobs (OpenAI-compatible API), Airflow/Metaflow connectors (Phase 1), or MCP server (Phase 3) with credentials/budgets inherited from the gateway. All workloads run inside hardware TEE (Intel TDX/SGX, NVIDIA CC) with remote attestation (CoCo Trustee + Intel Trust Authority) proving code and configuration are unmodified.
Validator scoring is a single Infrastructure mechanism (Phase 0): binary readiness gate (hotkey binding, Rancher API connectivity, HA topology, 8-GPU workers, confidential runtime handler) + USD compensation pricing. Miners with perfect attestation but priced 2× competitor average score low (weighted by price-competitiveness index). Miner/owner weight split is dynamic per epoch: miners earn a USD target, everything else is set as weight on owner UID which under recycle_or_burn=recycle returns to unissued supply (no owner reward unless owner runs a miner).
Miner deposit (10 TAO per node) is not slashable; enforcement is a reversible freeze: validators withhold score → emission stops → deposit draining stops → capital freezes on miner hotkey until scoring resumes. Phase 1 adds fresh TEE attestation gate (MRCONFIGID/RTMR verify), Armada job metrics (success rate, latency), and health scoring. Phase 2 adds USDC-on-BASE and TAO-on-BASE paid jobs at dynamic per-job-class price (resources per hour, bounded by competitor signals and queue depth).
Crossover point: when consumed Alpha revenue > emission spend, the subsidy ratio crosses zero and DePIN self-funds.
The sections above were auto-generated from the GitHub repository and may not reflect the latest changes. View repository (github.com, opens in a new tab, external site)
KubeTEE AI Factory: Confidential Computing TEE Multi-Cluster K8s
https://x.com/KubeTEEAI (x.com, opens in a new tab, external site) — Enterprise-Grade AIaaS with FIPS-140-3 Federal-Grade Security
This information was auto-extracted from the GitHub repository and may be incomplete or out of date. Always refer to the subnet's GitHub repository (github.com, opens in a new tab, external site) for the latest instructions.
Not yet verified: GPU count, Network, Notes
Not yet verified: GPU, OS, Network, Notes
Machine-extracted from the subnet's repository docs; may be wrong or out of date. Check the repository before buying or renting hardware. Open the repository (github.com, opens in a new tab, external site) · Extracted 2 Nov 2026
Procure confidential-compute-capable server hardware; enable TDX/SEV-SNP in BIOS and kernel
Prepare hardware: 3+ Intel TDX/AMD SEV-SNP nodes with 8x B200/B300/H200 GPUs each, same GPU type per cluster
Stand up Kubernetes infrastructure with TEE runtime support; see Node BIOS & OS Preparation and GPU Node Requirements docs
Deploy RKE2 cluster on all nodes with FIPS-140-2 validated baseline; install GPU Operator with Kata Containers and Confidential Containers
Apply Kubernetes labels that connect the cluster to your mining hotkey; one cluster per hotkey
Register nodes with KubeTEE: apply kubetee.ai/hotkey label binding and other platform labels per Cluster Naming Convention
KubeTEE platform team applies kubetee.ai/hotkey binding and onboards cluster to Fleet GitOps; hand-reviewed during Early Access, permissionless in Phase 1
Contact KubeTEE for cluster enrollment: provide Rancher API endpoint, hotkey, coldkey, cluster name
Enable multi-cluster batch job scheduling; submit jobs via Armada queues with confidential runtime class(optional)
Install Armada Executor on the cluster; configure to receive jobs from subnet-owner Armada Server
Optional: host inference services; idle GPU headroom can serve SN28 sayGM buyers(optional)
Deploy NeMo Microservices or SOTA model instances (GLM, Ornith, MiMo, etc.) in Kata TEE pods
On-chain registration collateral held on your hotkey; use btcli add-collateral command(optional)
Deposit 10 TAO per contributed node on your mining hotkey (Phase 1 enforcement)
Published to Hippius S3 every cycle; verify TEE attestation is passing and infrastructure-readiness gate is met
Monitor validator emissions and attestation metrics via KubeTEE Validator/Miners Dashboard
Set up Python development environment with Bittensor SDK and required dependencies
Install Bittensor subtensor SDK 11.0.1+; git clone the KubeTEE validator code from validator/ directory
Set KUBETEE_HOTKEY, KUBETEE_COLDKEY, KUBETEE_NETUID env vars; register on-chain if not already
Configure validator with hotkey, coldkey, subnet netuid (90), and chain network (finney or testnet)
Validator queries miner clusters via Rancher for infrastructure readiness; token is created by signing a Bittensor challenge
Configure Rancher v3 client credentials: set RANCHER_API_TOKEN and RANCHER_API_URL for read-only cluster access
Validator reads Taostats TAO/USD and on-chain alpha→TAO to compute miner compensation in USD; publishes Targon payout snapshot to Hippius S3
Configure price feeds: set KUBETEE_GPU_USD_PRICES (default card), TAOSTATS_API_KEY, HIPPIUS secrets for Targon snapshot publishing
Configure the on-chain registration collateral validator gate; Phase 0 runs measure-only (ENFORCE=false), Phase 1 enforces via score withholding
Set miner deposit gate parameters: KUBETEE_COLLATERAL_REQUIRED_TAO (10 × miner node count), KUBETEE_COLLATERAL_ENFORCE (false in Phase 0, true in Phase 1), grace and recovery margins
Main validator loop: each epoch reads metagraph, queries miner clusters, computes infrastructure-readiness + price-competitiveness scores, sets on-chain weights
python validator/run.py (or equivalent entry point); set weights once per epoch via Bittensor set_weights()
Publish internal metrics for operator observability; DePIN subsidy ratio (emission value ÷ total miner compensation) is the durable KPI
Monitor validator KPI metrics: kubetee_miner_collateral_coverage, kubetee_miner_collateral_status, kubetee_effective_usd_per_card, kubetee_targeting_precision via Prometheus
Referee attestation proves validator code is unmodified; public attestation endpoint allows anyone to verify scorer integrity(optional)
Phase 2 roadmap: move validator into a Kata + CoCo TEE pod on control plane with remote attestation endpoint
| Day (UTC) | Share |
|---|---|
| 2026-11-03 | 1.94% |
| 2026-11-02 | 1.99% |
| 2026-11-01 | 2.02% |
| 2026-10-31 | 2.03% |
| 2026-10-30 | 1.95% |
| 2026-10-29 | 1.92% |
| 2026-10-28 | 1.96% |
| 2026-10-27 | 1.86% |
| 2026-10-26 | 1.95% |
| 2026-10-25 | 2.15% |
| 2026-10-24 | 2.13% |
| 2026-10-23 | 2.22% |
| 2026-10-22 | 2.36% |
| 2026-10-21 | 2.51% |
| 2026-10-20 | 2.56% |
| 2026-10-19 | 2.41% |
| 2026-10-18 | 2.44% |
| 2026-10-17 | 2.53% |
| 2026-10-16 | 2.60% |
| 2026-10-15 | 2.70% |
| 2026-10-14 | 2.64% |
| 2026-10-13 | 2.45% |
| 2026-10-12 | 2.37% |
| 2026-10-11 | 2.49% |
| 2026-10-10 | 2.64% |
| 2026-10-09 | 2.68% |
| 2026-10-08 | 2.65% |
| 2026-10-07 | 2.55% |
| 2026-10-06 | 2.57% |
| 2026-10-05 | 2.60% |
| 2026-10-04 | 2.55% |
| 2026-10-03 | 2.65% |
| 2026-10-02 | 2.57% |
| 2026-10-01 | 2.53% |
| 2026-09-30 | 2.59% |
| 2026-09-29 | 2.66% |
| 2026-09-28 | 2.53% |
| 2026-09-27 | 2.46% |
| 2026-09-26 | 2.68% |
| 2026-09-25 | 2.86% |
| 2026-09-24 | 2.88% |
| 2026-09-23 | 2.73% |
| 2026-09-22 | 2.71% |
| 2026-09-21 | 2.77% |
| 2026-09-20 | 3.12% |
| 2026-09-19 | 2.88% |
| 2026-09-18 | 3.01% |
| 2026-09-17 | 3.22% |
| 2026-09-16 | 3.27% |
| 2026-09-15 | 1.72% |
| 2026-09-14 | 1.17% |
| 2026-09-13 | 0.56% |
| 2026-09-12 | 0.47% |
| 2026-09-11 | 0.42% |
| 2026-09-10 | 0.35% |
| 2026-09-09 | 0.17% |
| 2026-09-08 | 0.00% |
| 2026-09-07 | 0.00% |
| 2026-09-06 | 0.00% |
| 2026-09-05 | 0.00% |
| 2026-09-04 | 0.00% |
| 2026-09-03 | 0.00% |
| 2026-09-02 | 0.00% |
| 2026-09-01 | 0.00% |
| 2026-08-31 | 0.00% |
| 2026-08-30 | 0.00% |
| 2026-08-29 | 0.00% |
| 2026-08-28 | 0.00% |
| 2026-08-27 | 0.00% |
| 2026-08-26 | 0.00% |
| 2026-08-25 | 0.00% |
| 2026-08-24 | 0.00% |
| 2026-08-23 | 0.00% |
| 2026-08-22 | 0.00% |
| 2026-08-21 | 0.00% |
| 2026-08-20 | 0.00% |
| 2026-08-19 | 0.00% |
| 2026-08-18 | 0.00% |
| 2026-08-17 | 0.00% |
| 2026-08-16 | 0.00% |
| 2026-08-15 | 0.00% |
| 2026-08-14 | 0.00% |
| 2026-08-13 | 0.00% |
| 2026-08-12 | 0.00% |
| 2026-08-11 | 0.00% |
| 2026-08-10 | 0.00% |
| 2026-08-09 | 0.00% |
| 2026-08-08 | 0.00% |
| 2026-08-07 | 0.00% |
| 2026-08-06 | 0.00% |
