Decentralized AI investors are venture funds, crypto-native firms and token ecosystems that finance AI built on open networks: distributed GPU compute, crowdsourced data, on-chain verification and permissionless model training.
TL;DR
AI Funding tracked three decentralized AI (DeAI) rounds between June and August 2026 (Ornn, Vangrid and Perceptron) totalling $48.5 million, and every one was backed mainly by crypto-native investors rather than traditional AI venture funds. a16z crypto and Galaxy Ventures co-led Ornn's $33 million seed for an AI compute marketplace, the largest of the three. The DeAI investor base is a different crowd from mainstream AI venture capital: Animoca Brands, HashKey, Crypto.com Capital, Robot Ventures, Paradigm and Delphi Ventures show up repeatedly, while Sequoia, Accel and Lightspeed are absent.
Key Takeaways
- Ornn raised a $33 million seed (June 2026) for a GPU compute marketplace and price index. It is the largest DeAI-adjacent round in AI Funding's 2026 data, co-led by a16z crypto and Galaxy Ventures.
- Vangrid raised a $9 million seed (August 2026) for a smartphone-based decentralised spatial mapping network. Six crypto-native backers took part: HashKey, Borderless, Crypto.com Capital, Animoca Brands, Gate Labs and Mapleblock.
- Perceptron closed a $6.5 million strategic round (July 2026) for a decentralised AI data network that reports more than 807,000 nodes. All 15 named participants are Web3 funds, trading firms or protocol foundations, such as Walrus Foundation and Aethir.
- Prime Intellect, which ran the first globally distributed training of a 10B-parameter model (INTELLECT-1, October 2024), raised a $130 million Series A in July 2026. Radical Ventures led it, according to TechCrunch.
- Nous Research, which coordinates decentralized training on Solana through its Psyche network, was raising at least $75 million in July 2026, led by Robot Ventures, according to TechCrunch. Earlier backers include Paradigm, Delphi Ventures and North Island Ventures.
- Crypto funds increasingly write cheques into AI companies with no token at all. Canonical Crypto joined Synthefy's $6.5 million seed, and Paradigm co-led Zipline's $1.5 billion round in November 2026.
What is decentralized AI and why does it need different investors?
Decentralized AI (often shortened to DeAI) covers startups that replace one centralized part of the AI stack with an open network. The four parts it targets most often are:
- Compute: pooling idle or fragmented GPUs instead of renting from one hyperscaler (Gensyn, Ornn's pooled capacity)
- Data: paying a crowd of contributors to supply and verify training data (Perceptron, Vangrid)
- Training: splitting a single training run across poorly connected machines (Prime Intellect's INTELLECT-1, Nous Research's Psyche)
- Verification: recording provenance or coordination state on a blockchain (Vangrid's on-chain capture verification, Psyche's Solana coordinator)
These companies often use incentive networks, node operators or token-adjacent mechanics. Traditional AI funds struggle to underwrite those. Crypto-native funds already know how to value networks, run validators and model token economics, so they have become the default lead investors for the category.
Who are the most active decentralized AI investors in 2026?
The table below lists every 2026 round in AI Funding's dataset where the company builds on a decentralized network, along with the investors named by the primary source.
| Company | Round | Amount | Date | What it decentralizes | Named investors (source) |
|---|---|---|---|---|---|
| Ornn | Seed | $33.0M | Jun 2026 | GPU compute trading | a16z crypto, Galaxy Ventures (co-leads); Nordstar, SV Angel, Crucible Capital, Vine Ventures, Link Ventures, Box Group |
| Vangrid | Seed | $9.0M | Aug 2026 | Spatial/mapping data | HashKey, Borderless, Crypto.com Capital, Animoca Brands, Gate Labs, Mapleblock |
| Perceptron | Strategic | $6.5M | Jul 2026 | AI training data | Sigma Capital, Selini Capital, QCP Capital, P2 Ventures, CoinDCX Ventures, Momentum6, DeFi Capital, Walrus Foundation, Aethir, Colosseum, and others |
| Total | $48.5M |
Three investor types stand out:
- Crypto venture arms of large firms. a16z crypto is the most prominent example. It led Gensyn's Series A in June 2023 and describes Gensyn as "a blockchain-based marketplace protocol" for machine-learning compute. In 2026 it reappeared on the cap table of Ornn, which turns AI compute into a tradeable commodity.
- Exchange and trading-firm venture arms. Crypto.com Capital, Gate Labs, HashKey, CoinDCX Ventures, QCP Capital and Selini Capital all appear in 2026 DeAI rounds. These firms profit when on-chain activity grows, so data and compute networks that pay contributors fit their strategy.
- Protocol foundations and ecosystem funds. Walrus Foundation (decentralized storage), Aethir (decentralized GPU cloud) and Colosseum (the Solana accelerator) all invested in Perceptron. Foundations invest to attract applications to their own networks. In practice this is distribution capital more than financial capital.
How do decentralized AI rounds compare with mainstream AI funding?
DeAI rounds are small. The three rounds in the table total $48.5 million, less than half of the $130 million Prime Intellect raised in a single Series A, and a rounding error next to the multi-billion-dollar foundation-model rounds that dominate AI Funding's leaderboard.
| Metric | Decentralized AI (2026 tracked) | Mainstream AI example |
|---|---|---|
| Largest round | Ornn, $33.0M seed | Prime Intellect, $130M Series A |
| Typical stage | Seed / strategic | Series A and later |
| Typical lead | Crypto VC or none named | Generalist or AI-specialist VC |
| Investors per round | 6 to 15+ | Usually 3 to 8 |
| Lead disclosed? | Often not (Vangrid, Perceptron) | Almost always |
Two structural differences explain the gap:
- Syndicates are wide and leaderless. Perceptron's round named 15 participants and no lead, and Vangrid's announcement listed "key stakeholders" without a lead. In a traditional Series A the lead sets the price and takes a board seat. In a DeAI strategic round, each participant often brings a network integration, liquidity or exchange distribution instead.
- Revenue arrives later. Perceptron reports node counts (807,000+) rather than revenue. Prime Intellect, which began as a distributed-training lab, now reports a $100 million annualized revenue run rate from enterprise agent tooling, and that commercial pivot is what drew a mainstream lead.
Which companies show the path from decentralized training to venture scale?
Prime Intellect: from distributed research to enterprise platform
Prime Intellect launched INTELLECT-1 on October 11, 2024 as "the first globally-distributed training run of a 10-billion-parameter model". It used the DiLoCo method, which lets training run on "islands of poorly connected devices". The company's blog reports 98% compute utilization across distributed workers, with workers synchronizing every 100 steps.
By July 2026 the company was selling a "full stack" for AI agent development: compute access, a reinforcement learning framework and evaluation tools. Customers include Ramp and Zapier. TechCrunch reported that the $130 million Series A was led by Radical Ventures, with Nvidia Ventures, Intel Capital, Dell Technologies Capital and Iconiq participating, at a $1 billion valuation. The investor list has nothing crypto-native in it. Once a decentralized-training company sells conventional enterprise software, conventional AI funds take over the cap table.
Nous Research: the crypto-native cap table that stayed crypto-native
Nous Research is a contrasting case. Its Psyche network coordinates training "on the Solana blockchain", and a smart contract acts as the authority for each training run. Its first Psyche run targets a 40B-parameter model called Consilience. TechCrunch reported in July 2026 that Nous was finalizing at least $75 million led by Robot Ventures, with USV as a significant participant. The same report said Nous had raised about $70 million before that from investors including Paradigm, Robot Ventures, North Island Ventures, Delphi Ventures, OSS Capital and Balaji Srinivasan. The round was still being finalized when that report was published, so treat the figures as reported rather than closed.
Why are crypto funds investing in AI companies without tokens?
The clearest 2026 trend is crypto-native investors moving into AI companies that have no token at all:
| Crypto-native investor | 2026 AI-related deal | Role |
|---|---|---|
| a16z crypto | Ornn, $33M seed (compute marketplace) | Co-lead |
| Galaxy Ventures | Ornn, $33M seed | Co-lead |
| Canonical Crypto | Synthefy, $6.5M seed (numerical-data foundation models) | Participant (Wing Venture Capital led) |
| Paradigm | Zipline, $1.5B round (autonomous drone delivery) | Co-lead |
| Coinbase Ventures | Cycles, $6.4M seed (crypto financial infrastructure) | Participant |
Synthefy is the clearest example. It builds "Structured Data Foundation Models" trained on numbers rather than text, for fraud detection, pricing and demand forecasting, and has no blockchain component. Canonical Crypto joined Wing Venture Capital, Samsung Next, Haystack and Lightscape in the round. Ornn is similar: SiliconANGLE describes a compute price index and a proposed ICE-cleared futures market, with no on-chain mechanics. Its only crypto connection is its investors.
Three things are driving the crossover:
- Overlapping skill sets. Crypto funds have spent years backing marketplaces for scarce, commoditized resources such as block space, storage and bandwidth. GPU hours are the next resource of that kind.
- Overlapping founders. Many distributed-systems engineers who built blockchain networks now build distributed training and inference systems.
- Portfolio diversification. Crypto token markets are cyclical, so funds whose LPs committed to "frontier technology" can deploy into AI without leaving their mandate.
What should founders know before raising from decentralized AI investors?
- Ask who leads. Wide syndicates with no lead (Perceptron, Vangrid) bring integrations but rarely bring a board member who prices the next round.
- Separate token and equity. Know whether an investor expects token warrants, equity or both. That decision affects which mainstream funds can follow later.
- Plan the crossover. Prime Intellect shows that a decentralized-training company can attract Radical Ventures, Nvidia Ventures and Intel Capital, but only after it built a product enterprises pay for.
- Use foundation capital for distribution. Protocol foundations such as Walrus Foundation and Aethir are most valuable for ecosystem access, not for follow-on capital.
FAQ
Who are the biggest decentralized AI investors?
The most active decentralized AI investors in 2026 are a16z crypto (Gensyn's Series A lead, Ornn co-lead), Galaxy Ventures (Ornn co-lead), Paradigm and Delphi Ventures (Nous Research backers), Robot Ventures (leading Nous Research's 2026 raise), and exchange-linked funds such as Crypto.com Capital, HashKey, Gate Labs and Animoca Brands (Vangrid).
How much money did decentralized AI startups raise in 2026?
AI Funding tracked three explicitly decentralized AI rounds between June and August 2026 (Ornn $33 million, Vangrid $9 million, Perceptron $6.5 million), totalling $48.5 million. Larger AI companies with decentralized-training roots, such as Prime Intellect ($130 million Series A), raised from mainstream AI funds instead.
Do traditional VCs invest in decentralized AI?
Rarely at the seed stage. In AI Funding's 2026 data, none of the three explicitly decentralized AI rounds named Sequoia, Accel, Lightspeed or Index. Traditional AI funds appear once a company has conventional enterprise revenue, as Radical Ventures did when it led Prime Intellect's Series A.
What is the difference between decentralized AI and open-source AI?
Open-source AI publishes model weights or code. Decentralized AI distributes the infrastructure that produces or runs models: compute, data collection, training coordination or verification. Nous Research does both, releasing open models and training them over the Solana-coordinated Psyche network.
Sources
- SiliconANGLE: Ornn raises $33M to help companies buy and sell AI compute
- Tech.eu: Vangrid raises $9M seed for a decentralised spatial intelligence network
- Tech.eu: Perceptron raises $6.5M to build decentralised AI data network
- TechCrunch: Prime Intellect raises $130M Series A
- TechCrunch: Nous Research in talks for new funding
- Prime Intellect: INTELLECT-1, the first decentralized training of a 10B model
- Nous Research: Nous Psyche
- a16z: Investing in Gensyn
- SiliconANGLE: Synthefy raises $6.5M for models trained on numerical data
- Tech Funding News: Zipline's $1.5B raise
Every round in this analysis is listed with its investors and source on the AI Funding investor and company pages at aifunding.me, which update as new decentralized AI deals are ingested.
