We mapped 40 AI-crypto projects by GitHub commits, TVL, active users, and revenue. The sector splits cleanly into two cohorts: a small group that is shipping product and growing real usage, and a much larger group that is shipping tweets and growing market cap. Below: the names with the largest ship-vs-tweet gap, in both directions.
Key takeaways
- Top-5 by GitHub commit velocity: Bittensor (TAO), Render, Akash, Fetch.AI, Ocean Protocol.
- Top-5 by Twitter/X following relative to GitHub commits: deliberately not naming, see methodology.
- Average AI-crypto token is down 47% from Q1 peak; outperformers are protocols with measurable revenue.
- Three real revenue stories: Bittensor (subnets paying TAO for compute), Render (creators paying RNDR for GPU time), Akash (real cloud-rental revenue).
- The narrative-to-ship ratio is the most useful signal for sector-specific allocation.
The methodology
For each of 40 AI-crypto projects we collected:
- GitHub commits over the past 90 days (across all repos in the project’s primary org)
- Unique contributor count over the same period
- TVL where applicable, active users where measurable
- Token holder count and concentration
- Twitter/X following + 90-day engagement
The “narrative-to-ship ratio” is (Twitter follower count × engagement rate) / GitHub commits. High ratio = lots of mouth, less product. Low ratio = product-led growth.
The shippers
1. Bittensor (TAO)
Bittensor runs decentralised subnets where TAO holders allocate emission to subnet operators based on the model performance subnets deliver. Real GPU and inference work happens through these subnets — and operators are paid in TAO. Commit velocity: 320+ commits across the org in 90 days. 28 active contributors.
The risk is reflexivity — TAO price drives emission rewards, which drives subnet operator participation, which drives utility, which drives TAO price. If price falls hard, emission economics weaken.
2. Render Network (RNDR)
Render is GPU compute marketplace for creative rendering — Blender, Octane, etc. Real customers (animation studios, 3D artists, small game studios) pay in RNDR for render time. Commit velocity: 180+ commits. Token velocity high (customers buy RNDR, spend on render, RNDR cycles back to GPU operators).
The product works. Whether token-value capture works at scale is the long-term question.
3. Akash Network (AKT)
Akash is decentralised cloud-compute marketplace. Real revenue — measurable in USD-equivalent terms — has grown from $0 in 2023 to roughly $4.5M annualised today. Small relative to AWS, real relative to anyone else in crypto-cloud.
Commit velocity: 240+ commits. Notable: separates infrastructure work (the chain) from product work (the marketplace) clearly.
4. Fetch.AI (FET, now ASI alliance)
Fetch.AI merged into the ASI alliance (with Ocean and SingularityNET) in 2024. Real ML agent activity, though the alliance integration is still being completed. Commit velocity: 160+ commits across the alliance orgs.
5. Ocean Protocol
Data marketplace for ML training data. Slow but consistent commit cadence. Real datasets, real buyers.
| Project | GitHub commits (90d) | Contributors | Token Y/N | Revenue signal |
|---|---|---|---|---|
| Bittensor | 320+ | 28 | TAO | Strong (subnet payments) |
| Render | 180+ | 14 | RNDR | Strong (GPU marketplace) |
| Akash | 240+ | 22 | AKT | Strong ($4.5M ARR) |
| Fetch.AI | 160+ | 19 | FET/ASI | Moderate |
| Ocean Protocol | 90+ | 11 | OCEAN/ASI | Moderate |
The narrators
Twenty-three of the 40 projects in our sample have:
- Twitter/X followings above 50,000
- GitHub activity below 20 commits in 90 days
- No identifiable revenue or paying users
- Token launches that captured 8-figure or 9-figure market caps
We are choosing not to name these. The pattern is identifiable from public data — anyone replicating our methodology can produce the same list. Naming them invites both legal threats and a flood of community pushback that does not change the underlying facts.
“The AI-crypto sector is doing two different things at once. Some projects are building products. Most projects are running performances. The token price often does not distinguish.”
What this means for allocation
The lesson is the same one applies to every sector: revenue and product velocity matter more than narrative. The projects in the shipper list have measurable usage that can compound. The projects in the narrator list need new narratives every quarter to sustain price.
None of this guarantees the shippers outperform short-term. Reflexive cycles can lift narrators well above shippers for months. But on multi-year horizons, the gap between ship and tweet eventually shows up in price.
What to track next
- Bittensor subnet performance distribution. If a few subnets capture all emission, the network becomes less distributed than the marketing claims.
- Akash revenue growth rate. Sustained 30%+ YoY would put it in real-product category.
- Render customer concentration. Currently top-10 customers are a large share of revenue — diversification matters for the long-term thesis.
Why this matters
AI-crypto is the most-talked-about sector in crypto right now. Disentangling real product from narrative is the highest-value analytical task an investor or developer can do in the sector. Doing it requires looking at GitHub, not just X.