> For the complete documentation index, see [llms.txt](https://docs.thetpa.ai/deai-tokenized-processing-architecture/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.thetpa.ai/deai-tokenized-processing-architecture/executive-summary.md).

# Executive summary

**TPAI Chain: a next-generation blockchain with built-in AI compute.** Bitcoin made money programmable and Ethereum made agreements programmable; the chains that followed made them faster and cheaper. None can run AI, because a blockchain needs every node to repeat every computation and get the same result, while AI takes billions of GPU calculations and is not bit-identical from one GPU to the next **\[Analysis]**. TPAI Chain therefore puts a datacenter GPU pool under the protocol and a blockchain layer on top: the GPUs do the AI work, and the chain decides who may run what, proves the work and pays everyone. Ownership, control, payment, verification and governance are decentralized across 1,000 independent validator nodes, while the compute runs in real datacenters under caps and hardware proof. This is, by design, as decentralized as real AI compute can be with today's technology **\[Analysis]**.

DeAI TPA is how people use it: an AI service its builder can own, sell and earn from when others use it. It extends infrastructure that Stage Technology Inc. (Toronto, 2020) built for its own spatial-computing AI models; the same addressing, hosting and payment stack now serves any AI model. It is an extension of the original technology, not a pivot.

## Works today vs planned

| Works today                                                            | Status                      | Next                                           | Date            |
| ---------------------------------------------------------------------- | --------------------------- | ---------------------------------------------- | --------------- |
| TPA registry on Ethereum mainnet: about 1,220 TPAs held by 417 wallets | **\[Verified]**             | TPC listing as ERC-20 (1% of supply)           | 29 March 2027   |
| More than 20 models reachable through TPAs                             | **\[Working]**              | TPAI Chain public testnet                      | 11 June 2027    |
| S-LoRA serving of customer adapters on open base models                | **\[Working]**              | TPAI Chain mainnet with 1,000 validator nodes  | 11 October 2027 |
| TPC contract on the Sepolia testnet with its full fixed supply         | **\[Verified]**             | Approved partner datacenters join the GPU pool | Year one        |
| Clients in hospitality and entertainment in Mexico                     | **\[Company announcement]** | AI-powered dApps and fine-tuning               | Phase 2         |
|                                                                        |                             | Long-term allocations start vesting            | 11 October 2028 |

Everything in the right-hand column is designed but not yet live; status labels throughout the paper show which is which.

## Summary

* **What it is.** Each TPA is a three-letter, three-digit address such as `ABC-042` (17,576,000 possible) that points to an AI model, LoRA adapter, agent or API. One TPA is at once a name, an endpoint, a revenue channel for its owner and a key to shared compute. TPC is the single coin used to pay for AI usage across all TPAs. A TPA is an AI endpoint, not a wallet; wallets are separate.
* **What is new.** OpenRouter and Hugging Face offer many models behind one account; x402 lets agents pay per request; Together AI and Fireworks host many LoRA adapters cheaply; ENS sells owned names; Virtuals Protocol turns agents into co-owned tokens; GPU marketplaces match buyers and sellers of compute. DeAI TPA puts an owned, tradable address for any AI service on a chain whose protocol includes the datacenter GPUs that run it. See [How it compares](https://docs.thetpa.ai/deai-tokenized-processing-architecture/the-solution.md) **\[Analysis]**.
* **How TPAI Chain works.** Users fund a session balance in TPC from post-quantum TPAI21- wallets. TPAs call AI contracts; the chain sends the job to the GPU pool or to an outside AI API, the result goes straight to the user, and calls are settled on chain in batches with their receipts. AI runs off the consensus path, so consensus never waits for AI work, and blocks target finality of about 1–2 seconds, to be confirmed on testnet **\[Planned]**. See [Architecture](https://docs.thetpa.ai/deai-tokenized-processing-architecture/architecture.md) and [TPAI Chain](https://docs.thetpa.ai/deai-tokenized-processing-architecture/tpai-chain.md).
* **Who runs it.** Exactly 1,000 licensed, KYC-verified validator nodes, all running from mainnet, in four tiers (micro, standard, super and super max). Every node has equal voting weight, one node one vote: an attacker would need 334 nodes. Hosting operators and datacenters are registered on chain, and the chain refuses any node that would take an operator above 5% or a datacenter above 10%; the 10-node owner cap is enforced through KYC of each node's real owner. Nodes are paid for the work they do, from every revenue stream the chain earns **\[Planned]**.
* **Proof and privacy.** Every GPU in the pool proves its hardware and approved software with NVIDIA attestation before it can take jobs. That software records a fingerprint of the exact model and adapter it ran in each receipt, signed with the attested key. From launch, super nodes re-run a random sample of jobs, and proven cheating costs the provider its bond. User data never goes on chain, only a receipt with a hash. Fake or harmful TPAs are reported, paused and removed by validator vote **\[Planned]**.
* **What runs today.** The TPA registry on Ethereum mainnet since February 2023: about 1,220 TPAs held by 417 wallets, some of them created by Stage for users who sign up by email or held by Stage itself **\[Verified]**. More than 20 models reachable through TPAs: frontier models such as GPT through provider APIs, and open base models such as Llama, Qwen and DeepSeek served by Stage with S-LoRA, which accept customer LoRA adapters **\[Working]**. The TPC contract on the Sepolia testnet with its full fixed supply **\[Verified]**.
* **Who uses it.** Hospitality and entertainment businesses in Mexico. Stage has announced Grupo Divertido and its Zoo Aventuras (Mexico City), Ripley's Museum Cancún and Sky Wheel **\[Company announcement]**, and names Delphinus (Grupo Xcaret, TPA DLP-000) and SLS Cancún (SLS-000) as clients **\[Founder statement]**. These clients use TPAs for spatial-computing experiences; AI usage figures are not yet published. See [Evidence](https://docs.thetpa.ai/deai-tokenized-processing-architecture/evidence.md).
* **Who is behind it.** Co-founders Vahid (Vito) Akbari and Azam Mohabatian, who previously co-founded Nekso. Investors include Right Side Capital Management, Kevin Dick and Dave Lambert, with strategic angel investors Mohammad Galadari, Roland Alexander and Alejandra Luis Lépiz. TheBlock (Dubai) is the operational partner for the launch. TPAI Chain is built in-house by Stage's engineering team. Public profiles are linked in [Team](https://docs.thetpa.ai/deai-tokenized-processing-architecture/team.md).
* **The coin.** TPC has a fixed supply of 17,576,000,000: never increased, never burned. Nothing is released upfront from the three long-term allocations (Founders and team, AI Computing Power, and Community & Ecosystem); they vest monthly over 10 years from 11 October 2028, against milestones confirmed by validator vote. The exceptions are stated in full on the TPCoin page: the public allocation (1% on 29 March 2027, 9% at mainnet), staking rewards from Community & Ecosystem from day one within a hard cap, and early adopters' TPC on their SAFT terms, none of it tradable before mainnet **\[Planned]**. See [TPCoin](https://docs.thetpa.ai/deai-tokenized-processing-architecture/tpcoin.md).
* **How the economics work.** AI is priced in US dollars and paid in TPC. Today, before TPAI Chain, the TPA owner receives 80% and Stage 20%. On TPAI Chain, from mainnet, every AI payment is split on chain between the TPA owner, the GPUs and the validators, and Stage takes no protocol fee; the full split is on the [TPCoin](https://docs.thetpa.ai/deai-tokenized-processing-architecture/tpcoin.md) page **\[Planned]**. Each TPA pays a flat $10 a month on TPAI Chain, half to the validators and half to Community & Ecosystem; TPAs bought before 29 March 2027 get their first 12 months free **\[Founder statement]**. Meta Energy (PING) is a direct, single-level referral: when a new TPA is bought, it can ping an existing one, which earns TPC from Community & Ecosystem, never from the buyer's payment, and better access to the GPU pool when it is busy **\[Founder statement]**.
* **What comes next.** Phase one of TPC on Ethereum mainnet on Easter Monday, 29 March 2027: 1% of supply (175,760,000 TPC) at a listing price of $0.31, with TheBlock as operational partner. The TPAI Chain public testnet opens on 11 June 2027, and mainnet goes live on 11 October 2027; both dates are fixed by the founders. At mainnet the other 9% of the public allocation is released natively and ERC-20 TPC can be swapped 1:1 for native TPC. AI-powered dApps and fine-tuning follow in Phase 2 **\[Planned]**. See [Roadmap](https://docs.thetpa.ai/deai-tokenized-processing-architecture/roadmap.md).
* **Why it matters.** PwC estimated in 2017 that AI could add up to $15.7 trillion to world GDP by 2030 ([PwC, 2017 estimate](https://www.pwc.com/gx/en/issues/data-and-analytics/publications/artificial-intelligence-study.html)), and McKinsey estimates that generative AI alone could add $2.6–4.4 trillion a year ([McKinsey, 2023](https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier)) **\[Verified source]**. The starting market is far smaller: software agents paid about $73 million on-chain in the year to April 2026. If a growing share of AI is built by many developers and called by agents, the owned address that earns from each call, on a chain that delivers the compute itself, becomes core infrastructure for that economy **\[Hypothesis]**.
* **Main open points.** Usage, latency and cost figures are not yet published, and no public inference API is documented yet; provider terms for reselling API access; the path away from a single cloud; TPAI Chain's code is private until its public testnet; the 1–2 second finality target must be confirmed on testnet; the Community & Ecosystem budget model and the price check against AI APIs are due before launch; node licensing and Meta Energy rewards are under legal review; the contracts are not yet verified or audited; and the legal entity that issues TPC will be named before any public sale. See [Risks and limitations](https://docs.thetpa.ai/deai-tokenized-processing-architecture/risks-and-limitations.md) and [Open questions](https://docs.thetpa.ai/deai-tokenized-processing-architecture/open-questions.md).
* **Where it leads.** Stage's own models are open source and hosted only on TPAI Chain, so the validators can stop any model that becomes dangerous, on a network governed by 1,000 independent validators, with no single operator above 5%. Stage plans its own AI compute datacenters in Canada, starting in Alberta **\[Planned]**. The direction is ASI **\[Vision]**.

## Contents

1. [Origin: Stage Technology and StageMeta](https://docs.thetpa.ai/deai-tokenized-processing-architecture/origin.md)
2. [The problem: paying for AI today](https://docs.thetpa.ai/deai-tokenized-processing-architecture/the-problem.md)
3. [The solution: TPA addresses, TPAI Chain and TPCoin](https://docs.thetpa.ai/deai-tokenized-processing-architecture/the-solution.md)
4. [Architecture and how a request works](https://docs.thetpa.ai/deai-tokenized-processing-architecture/architecture.md)
5. [Addressing, networking and isolation](https://docs.thetpa.ai/deai-tokenized-processing-architecture/addressing-networking-isolation.md)
6. [TPCoin (TPC)](https://docs.thetpa.ai/deai-tokenized-processing-architecture/tpcoin.md)
7. [TPAI Chain](https://docs.thetpa.ai/deai-tokenized-processing-architecture/tpai-chain.md)
8. [Evidence: what is verified on-chain](https://docs.thetpa.ai/deai-tokenized-processing-architecture/evidence.md)
9. [Roadmap](https://docs.thetpa.ai/deai-tokenized-processing-architecture/roadmap.md)
10. [Team](https://docs.thetpa.ai/deai-tokenized-processing-architecture/team.md)
11. [Risks and limitations](https://docs.thetpa.ai/deai-tokenized-processing-architecture/risks-and-limitations.md)
12. [Open questions](https://docs.thetpa.ai/deai-tokenized-processing-architecture/open-questions.md)
13. [Glossary and sources](https://docs.thetpa.ai/deai-tokenized-processing-architecture/glossary-and-sources.md)
14. [Revision history](https://docs.thetpa.ai/deai-tokenized-processing-architecture/revision-history.md)
