Hub
Hub collects real-world egocentric video to sell as training data for physical AI and robots.
On the record
Checkpoint: Window to prove Hub for business scales into repeatable, retained installs before Build AI widens its dataset lead
The call
Hub is trying to turn scattered camera captures into a recurring business channel before Scale AI's shipped platform and Build AI's faster pace close the gap.
Hub records video from cameras worn by people doing physical tasks, then sells that data to companies building AI robots that learn from human movements. The problem is that Scale AI (a much bigger competitor) already has a similar product and close ties to Meta, while Build AI is collecting videos ten times faster than Hub. The only thing that might save Hub is if their camera system catches on with small business owners as a standard tool, but that has not been proven to work at a large scale yet.
Why
01
AI pressure: NVIDIA Isaac-Sim synthetic data generation. Isaac-Sim can generate synthetic egocentric footage, cutting demand for the human-captured video Hub sells. Could AI do the core job? Only partly.
02
Bigger players: Scale AI. Scale AI already runs a live Physical AI Data Engine and has a Meta equity and data tie-up, covering most of the market Hub is chasing. Could a giant copy it? Yes, easily.
03
What protects it. Its 'Hub for business' workplace camera installs could give repeat access rivals lack, but so far it is one restaurant deal, not a program.
What would change this call
Whether Hub names any frontier lab or robotics client in a signed multi-year contract by mid-2027. Would show Hub has real enterprise traction beyond vendor-claimed unnamed deals.
Whether Hub for business expands from the single restaurant case to multiple documented multi-site installs with retention data. Would show the camera-mount distribution channel is a repeatable moat, not an anecdote.
Whether Build AI publicly reports dataset hours beyond 1M or a new milestone by end of 2026. Would confirm Build AI continues outscaling Hub's data collection pace.
Whether Hub for business expands from the single restaurant case to multiple documented multi-site installs with retention data. Would show the camera-mount distribution channel is a repeatable moat, not an anecdote.
Whether Build AI publicly reports dataset hours beyond 1M or a new milestone by end of 2026. Would confirm Build AI continues outscaling Hub's data collection pace.
Key numbers
- Build AI scaled from 10K to 1M egocentric hours in 5 months
- Hub reports 540,000+ hours captured across 10k+ sites
- Scale AI, Surge AI, Mercor, Handshake hold over 75% of the data market
Named risks
- Core capture business faces execution-speed competition from Build AI and Scale AI with no structural moat.
- Three business-model pivots in under 18 months (Web3 marketing, bandwidth network, physical-AI data) signal strategy instability.
- Self-reported contributor figures inconsistent (500,000+ vs 150,000+ within months), undermining traction credibility.
- Biometric/privacy regulation (Illinois BIPA, EU AI Act Article 10 phasing through August 2027) creates liability for face capture.
- Only $1.7M to $2M raised against Big Four rivals with far greater capital and Meta-level distribution.
Rivals, as read in July 2026
| Rival | Direction | Evidence |
|---|---|---|
| Scale AI | losing to | Already shipped a dedicated Physical AI Data Engine and holds a Meta equity/data deal reported mid-2025 |
| Build AI | losing to | Scaled egocentric dataset from 10K to 1M hours between Nov 2025 and April 8, 2026, far outpacing Hub |
| Handshake | losing to | Revenue reportedly grew 100-fold within nine months per July 24, 2026 sector report |
| Mercor | losing to | Revenue reportedly tripled within nine months per July 24, 2026 sector report |
| Ego4D / EgoVerse (free datasets) | losing to | Open academic egocentric datasets undercut Hub's differentiation at zero cost |
| Surge AI | contesting | Named among the Big Four controlling over 75% of AI training-data sector revenue as of July 2026 |
Questions people ask
- Is Hub (hub.xyz) profitable in 2026?
- No profitability data is public. Hub claims a 7-figure revenue agreement reached weeks before its June 15, 2026 YC launch and says it is scaling multiple 7 to 8 figure contracts as of July 25, 2026, but these are vendor-claimed and unaudited with no named clients.
- Could AI or synthetic data replace Hub's egocentric data collection?
- Partly. NVIDIA Isaac-Sim synthetic data pressures some use cases, but a July 2026 HumanScale paper found egocentric human video beat real-robot data (24% lower validation loss), sustaining demand for Hub's category even as rivals commoditize the raw capture itself.
- Who is Hub's biggest competitor?
- Scale AI, which already shipped a dedicated Physical AI Data Engine and holds a Meta equity/data relationship reported mid-2025. Build AI is also outpacing Hub, scaling its Egocentric dataset from 10,000 to 1,000,000 hours between November 2025 and April 8, 2026.
- How much funding has Hub raised?
- SwissBorg led Hub's pre-seed and joined its seed round, bringing total funding to $1.7M as of September 25, 2025. A mid-2026 job posting states $2M raised alongside its Y Combinator P26 spot, suggesting a small incremental increase since.
- What does Hub actually sell?
- As of its June 15, 2026 YC launch, Hub sells multimodal egocentric RGB-D data (stereo video, depth, IMU, hand pose) captured via its HubApp and a B2B 'Hub for business' camera-mount program, for frontier AI labs and robotics companies.
- Who owns Hub.xyz?
- Hub is privately held and venture-backed, with SwissBorg as a lead investor since its September 25, 2025 funding round and Y Combinator as an accelerator investor via the P26 batch (Demo Day June 16, 2026). No acquisition or IPO has occurred.
Timeline
- 2026-07-25Hub job posting claims scaling of multiple 7 to 8 figure contracts and 150,000+ users
- 2026-06-24Report cites Build AI's Egocentric-1M dataset released April 8, 2026, up from 10K in Nov 2025
- 2026-06-16Y Combinator P26 batch Demo Day
- 2026-06-15Hub publicly launches on YC as egocentric data lab for frontier AI and robotics, claims 7-figure revenue deal
- 2026-04-08Build AI releases Egocentric-1M dataset, roughly 100x its Nov 2025 release
- 2026-03-14Hub job posting cites $2M raised and 500,000+ contributors, 300,000+ community members
- 2025-10-21Hub opens formal Palo Alto HQ, hires Head of Sales Jeff Kanai (ex-Google/Bright Data)
- 2025-09-25Hub closes SwissBorg-led round, total funding reaches $1.7M, positioned as bandwidth network
- 2025-02-24Hub.xyz positioned as AI-driven Web3 marketing platform with $HUB token airdrop
Revision history
27 July 2026
First assessed. Losing its protection. The text above is fixed as of this date and is not edited afterwards.
In the same band
Sources
- current-scene.md (2026-07-08)
- 10 Egocentric Datasets Reshaping Robotics and AI in 2026 (2026-06-24)
- The AI Training Data Sector Spawns a Near-$100 Billion Valuation Business (2026-07-24)
- The Economics of AI Training Data: A Research Agenda (2026-04-20)
- Launch YC: Hub (2026-06-15)
- Founding Engineer at Hub (2026-07-25)
- Hub · Multimodal data lab for physical AI
- Founding Full Stack AI Engineer
- Get paid for the work your team already does, Hub for business
- Hub.xyz Closes Major Funding Round with SwissBorg to Scale The First Distributed AI & Data Infrastructure Netw (2025-09-25)
- Jobs at Hub.xyz
- Hub.xyz Expands into the U.S. with New Headquarters in Palo Alto, California (2025-10-21)