Eight companies in the data services competitive set. Where SmartOne sits among them, what each one is doing right now, and the three wedges that belong uniquely to SmartOne.
In June 2025, Meta acquired a 49% stake in Scale AI at a $29B valuation. Within weeks: Google announced plans to cut ties, OpenAI confirmed it had "already been winding down its work with Scale AI," and Microsoft and xAI scaled back. Scale then pivoted to government and defense. The enterprise AI lab segment that once drove most of Scale's commercial revenue is actively looking for alternatives.
| Company | Revenue 2025 | Model | Independence | Physical AI depth | Production base |
|---|---|---|---|---|---|
| Scale AI | $2B | Services + Gov pivot | Meta-owned (49%) | Moderate | Global crowd + contractors |
| Surge AI | $1.4B | Expert contractors | Bootstrapped (raising) | Low (RLHF/LLM focus) | 50K global contractors |
| Appen | $232M | Crowd platform | Public (ASX) | Low | Global crowd |
| iMerit | ~$34M | Full-time services | Series B (BI + Omidyar) | High (active pivot) | India (7,700+ FTE) |
| Snorkel AI | $148M ARR | Platform + services | VC-backed ($238M raised) | Moderate (growing) | US-focused |
| Labelbox | $33-114M est | Platform + services | VC-backed ($189M raised) | Moderate | US-focused platform |
| Toloka | Undisclosed | Crowd platform | Bezos-backed ($72M) | Low (RLHF/evaluation) | 200K global crowd |
| Sama | Undisclosed | Services | $70M Series B | Moderate | East Africa (contracting) |
| SmartOne | $6M | Full-time services | Bootstrapped (no VC) | High (in production) | Madagascar (1,000 FTE) |
$2B revenue, $29B valuation post-Meta investment. Largest player by revenue and government relationships. Physical AI Data Engine launched 2024 with named clients including Physical Intelligence.
For SmartOne: The neutrality claim has structural backing. Scale's largest investor is a direct competitor to the enterprise buyers SmartOne is targeting. That fact does not require a press release; it resolves itself in a due-diligence conversation.
$1.4B revenue, 130 full-time employees, 50,000 contractors globally. Primary beneficiary of Scale's client exodus: Google, OpenAI, and xAI moved significant volume to Surge in H2 2025. Expert-level annotation, 10x price premium vs. commodity, RLHF and frontier model focus.
For SmartOne: Surge does not compete in Physical AI. Different segment: expert RLHF and frontier model training versus physical-world sensor and temporal data. Once Surge closes its raise, their neutrality claim weakens. SmartOne's bootstrapped posture is a structural differentiator that Surge is in the process of giving up.
~$34M revenue, 7,765 FTE employees in India. Full-time institutional workforce, not gig. Physical AI depth in autonomous mobility and healthcare. CEO Radha Basu named in AIM Media's Next 15 AI CEOs in America 2026.
For SmartOne: The closest model analog. The differentiators: Madagascar vs. India (different labor economics, different cultural positioning), bootstrapped vs. Series B-funded, and SmartOne's French entity plus EU footprint. iMerit would describe SmartOne as a peer, not a subordinate.
B-Corp certified, East Africa-based, 5,000+ workforce peak. Built its entire post-Meta positioning around independence: "Sama has never retained its clients' data." April 2026: Meta terminated Sama's contract after Kenyan workers reported viewing sensitive footage from Ray-Ban Meta smart glasses.
For SmartOne: Amazon at ~60% of revenue is SmartOne's Sama parallel. Ethical positioning alone does not protect against a single-client dependency. The Sama incident is not a cautionary tale about Africa-based operations. It is a cautionary tale about what happens when independence is a marketing claim rather than a structural fact.
$189M raised, $1B+ valuation. Platform-first: software for running annotation workflows plus the Alignerr expert network of 1M+ domain experts. Trusted by "over 80% of leading AI labs in the US." 449 employees.
For SmartOne: Labelbox would describe SmartOne as a potential customer, not a competitor. That view is worth noting when positioning. For buyers who want to bring annotation in-house, Labelbox wins. For buyers who want managed services with Physical AI depth from a conflict-free partner, SmartOne wins.
$148M ARR, $238M raised, $1.3B valuation at Series D. Stanford AI Lab origin. Programmatic labeling: foundation models and labeling functions reduce manual annotation dependency. Launched Expert Data-as-a-Service May 2025 to enter managed services.
For SmartOne: Snorkel's programmatic approach is a future substitution risk for rule-bound annotation tasks. For Physical AI ground-truth work requiring human judgment in ambiguous physical environments, automated labeling functions do not substitute. The moat is in the edge cases, not the volume work.
Amsterdam-based, originally Yandex subsidiary. $72M Series C in May 2025 led by Bezos Expeditions. Shopify CTO as Executive Chairman. 200,000+ annotators, 40+ languages. Named clients include Anthropic, Amazon, Microsoft, Shopify.
For SmartOne: Toloka's crowd model is not Physical AI-focused. The Bezos money means Amazon-adjacent investor dynamics. Despite the Amsterdam HQ, Toloka does not have a French entity, French-speaking leadership, or French defense client history. The EU positioning gap is real.
ASX-listed, Sydney-based. $232M revenue in 2025, down from peak. Lost Google as major client in 2024. GenAI now 44% of Q4 2025 revenue, up from 35% a year prior. FY2026 guidance $270-300M represents a recovery target from post-Google turnaround.
For SmartOne: Appen is rebuilding around Chinese GenAI clients. Not a Physical AI specialist, not a conflict-free positioning play. The comparison is useful for revenue context: SmartOne at $6M is at 2.6% of Appen's revenue with a cleaner ownership story and stronger Physical AI depth.
The specific claim: bootstrapped, no hyperscaler entanglement, Physical AI depth already in production, SOC 2 Type II and ISO 27001, 14-year institutional workforce. This is the clearest wedge and the most time-bounded one. It will narrow as Scale rebuilds trust, Surge raises capital and completes its VC round, and new entrants arrive. It is open today.
The reliability story. Full-time employees trained over years on specific annotation domains produce tighter accuracy distributions than gig contractors. For physical-world data where an annotation error can cause a robot to misclassify an obstacle, the production model matters. This wedge is durable because it is structural, not a positioning choice.
No direct competitor in this set has a French entity, a French-speaking leadership team, and active French defense client history simultaneously. As European AI adoption accelerates and EU AI Act compliance creates demand for EU-based data provenance, this combination is a real opening. This wedge activates in Phase 2.