Scale AI

Scale AI is an American artificial intelligence data, evaluation, and reinforcement-learning-environment company headquartered in San Francisco, founded in 2016 by Alexandr Wang and Lucy Guo. The company operates a global expert-labeling workforce and the Scale Forge and Scale Donovan platforms, supplying frontier AI labs and US government customers with the training data, RLHF feedback, expert evaluations, and synthetic data that underpin large foundation models. Scale AI's commercial customer base includes OpenAI, Anthropic, Google DeepMind, Meta, and the US Department of Defense. As of April 2026, Scale AI is the principal commercial AI data infrastructure company globally, operating under the strategic-partner relationship with Meta following the June 2025 transaction in which Meta acquired a 49 percent non-voting stake at a $14.3 billion valuation and Wang departed to become Meta's Chief AI Officer.

At a glance

• Founded: 2016 in San Francisco by Alexandr Wang and Lucy Guo.
• Status: Private. Strategic partnership with Meta since June 2025 (Meta holds 49 percent non-voting stake; Wang departed to lead Meta Superintelligence Labs).
• Funding: Approximately $1.6 billion in cumulative private capital before the Meta transaction. The June 2025 Meta investment provided $14.3 billion in capital, the largest single AI-infrastructure transaction at the time.
• CEO: Jason Droege, Chief Executive Officer (interim, since June 2025). Former Chief Operating Officer; previously a senior Uber executive.
• Other notable leadership: Alexandr Wang, Co-Founder (departed June 2025 to become Meta Chief AI Officer; remains on the Scale board). Lucy Guo, Co-Founder (departed 2018; founded the consumer-creator platform Passes).
• Open weights: N/A. Scale AI is a data and platform company, not a model producer.
• Flagship products: Scale Forge (RL environments and synthetic-data platform); Scale Donovan (US government and defense AI platform); Scale Data Engine and Studio (labeling and evaluation infrastructure); the Scale Generative AI Platform (model evaluation, fine-tuning, RLHF).

Origins

Scale AI was founded in 2016 by Alexandr Wang and Lucy Guo, both then nineteen years old. Wang, an MIT dropout who had been a software engineer at Quora, identified the bottleneck in early machine-learning systems as the absence of high-quality labeled training data at the volumes that frontier models would require. The original product, branded Scale Data Engine, supplied human-labeled image and sensor data to autonomous-vehicle programs at Waymo, Cruise, and the early Tesla Autopilot organization. The autonomous-driving customer base produced sustained early revenue through 2017 to 2020.

The 2020 to 2023 period saw Scale's commercial expansion across the federal-government market, with the launch of Scale Donovan in 2023 as the principal AI platform for the US Department of Defense and intelligence-community customers. The company's role as the labeling and RLHF supplier to leading frontier-AI labs accelerated through 2022 and 2023, with OpenAI, Anthropic, and Google DeepMind all reporting Scale relationships in commercial filings or executive comments. Scale's revenue scale through 2024 was reported in industry coverage as approaching $1 billion in annualized revenue, the highest among AI-data companies.

The June 2025 transaction with Meta restructured the company. Meta acquired a 49 percent non-voting equity position for $14.3 billion, valuing Scale at approximately $29 billion. Wang departed Scale to become Meta's Chief AI Officer and the lead of Meta Superintelligence Labs. Lucy Guo had previously departed in 2018 to start the consumer-creator company Passes. Jason Droege, the Chief Operating Officer, took the interim CEO role. Industry coverage characterized the transaction as Meta acquihiring Wang and the senior Scale leadership while leaving the operating company in place to continue serving non-Meta frontier-AI customers under separate-entity governance.

Mission and strategy

Scale AI's stated mission is to accelerate the development of AI applications by providing the data infrastructure that frontier-AI development requires. The strategy combines four threads. First, the labeling and RLHF capacity for frontier-AI labs, with the human-expert workforce as the operational moat. Second, the Scale Forge platform for synthetic data and RL environments, repositioning the company toward post-supervised-data training methods. Third, the Scale Donovan platform for US government and defense AI deployments, providing a regulated-customer revenue stream with structurally different margin and contracting dynamics from commercial AI customers. Fourth, the Scale Generative AI Platform for enterprise model evaluation and fine-tuning, targeting Fortune-500 customers building AI applications on top of frontier model APIs.

The competitive premise after the Meta transaction is that Scale can continue operating as a vendor-neutral data infrastructure layer for the broader AI industry while Meta's strategic investment provides capital and scale for continued platform investment. Industry coverage has questioned whether non-Meta frontier customers will reduce reliance on Scale given the Meta ownership stake; the early commercial-relationship reporting through 2025 to 2026 indicates that customer relationships have been preserved, with the non-voting structure of Meta's stake intended to address governance concerns.

Models and products

• Scale Data Engine. The original labeling and data-annotation platform. Supports image, video, sensor, and text labeling at scale through the global expert workforce.
• Scale Generative AI Platform. Enterprise platform for model evaluation, fine-tuning, RLHF, and prompt engineering. Targets Fortune-500 customers building applications on frontier-model APIs.
• Scale Forge. RL-environment and synthetic-data platform launched 2024. Provides simulated environments for reinforcement-learning training of foundation models.
• Scale Donovan. US government AI platform. Deployed across the Department of Defense and intelligence community for AI-augmented analysis, decision support, and autonomous systems applications.
• Scale SEAL Leaderboard. Public foundation-model evaluation leaderboard maintained by Scale's evaluation team. Updated as new frontier models release.
• Scale Studio. The internal-tool front-end for the labeling workforce and the data-quality-management workflow.

The commercial-channel mix combines direct sales to frontier-AI labs (the core revenue base), federal contracting through the Donovan platform, and self-serve and field-sales engagement with Fortune-500 enterprise AI customers.

Benchmarks and standing

Scale AI does not produce foundation models and is not directly evaluated against horizontal benchmarks. The company's standing in the AI ecosystem is anchored on the customer relationships with frontier-AI labs and the operational scale of the labeling-and-RLHF workforce, with industry coverage consistently characterizing Scale as the principal commercial AI-data supplier globally.

The Scale SEAL Leaderboard is a recognized public-facing AI-evaluation resource, with periodic releases of frontier-model rankings across reasoning, coding, and instruction-following benchmarks. SEAL evaluations are independently maintained by Scale's evaluation team and have been cited in industry coverage of major frontier-model releases through 2024 to 2026.

Leadership

As of April 2026, Scale AI's senior leadership includes:

• Jason Droege, Chief Executive Officer (interim). Former Chief Operating Officer; previously a senior Uber Eats and ride-hail executive.
• Senior research, engineering, and federal-business leadership across the Forge, Donovan, and Generative AI Platform organizations.

Alexandr Wang, Co-Founder, departed in June 2025 to become Meta's Chief AI Officer following Meta's strategic investment. Wang remains on the Scale board. Lucy Guo, Co-Founder, departed in 2018 to start Passes. The post-Wang executive transition has continued under Droege, with senior platform leadership preserved through the Meta transaction.

Funding and backers

Cumulative private capital before the Meta transaction reached approximately $1.6 billion across multiple rounds, including the Series F at a $13.8 billion valuation in May 2024 led by Accel, Cisco Investments, DFJ Growth, Index Ventures, Tiger Global, and Y Combinator alumni. The June 2025 Meta strategic investment of $14.3 billion at a $29 billion valuation restructured the cap table. Existing pre-transaction investors retained their positions, with Meta entering as a 49 percent non-voting strategic partner.

Industry position

Scale AI occupies a structurally distinctive position as the principal commercial AI-data infrastructure company globally, with the labeling-and-RLHF workforce, the Forge synthetic-data and RL-environment platform, the Donovan US government platform, and the Generative AI Platform for enterprise customers all operating under unified senior leadership. Industry coverage has characterized Scale as the "picks-and-shovels" winner of the foundation-model commercialization wave, with revenue scale and customer-relationship breadth that no other AI-data vendor matches at the comparable combination of attributes.

The Meta partial ownership has reframed industry interpretation of Scale's position. Coverage has questioned whether competitor frontier-AI labs will diversify data-supplier relationships in response, with Surge AI, Invisible Technologies, and Mercor characterized in industry coverage as the principal Scale alternatives gaining customer traction in the post-transaction period.

Competitive landscape

• Surge AI. Direct labeling-and-RLHF competitor founded by former Scale and Meta engineers. Has gained customer traction with frontier-AI labs in the post-Meta-transaction period.
• Invisible Technologies. Enterprise-AI services and data-operations competitor. Different operating model with broader business-process coverage.
• Mercor. AI-talent-marketplace competitor founded 2022. Focused on expert-labeling and AI-research-talent placement.
• Snorkel AI. Programmatic-labeling research-product competitor with a different methodological approach (programmatic weak supervision rather than human labeling).
• Datology AI. Data-curation research-product competitor with focus on training-dataset selection and curation.
• Hugging Face Datasets, Common Crawl, LAION. Open-source training-data alternatives that reduce frontier-AI-lab dependence on commercial labeling vendors.

Outlook

• The trajectory of customer-retention with non-Meta frontier-AI labs through 2026 to 2027.
• The commercial scaling of the Scale Forge RL-environment and synthetic-data platform.
• Continued Scale Donovan US government and defense AI contracting.
• The post-Wang senior leadership transition under Droege, with potential permanent CEO appointment.
• The strategic-partner relationship with Meta and the operational-independence governance arrangements.

Sources

• Scale AI official site. Company reference.
• Alexandr Wang Wikipedia. Co-Founder reference.
• Scale Forge. RL-environment and synthetic-data platform.
• Scale Donovan. US government AI platform.
• Meta announcement. June 2025 strategic investment.