Scale AI

In the era of rapid development of artificial intelligence, high-quality data has become the core element to promote model training and application. As a company focusing on AI data annotation and infrastructure, Scale AI has become a key player in the global AI ecosystem since its establishment. It not only provides data support for giants such as OpenAI and Meta, but also promotes innovation in autonomous driving, robotics and generative AI. As of October 2025, Scale AI's valuation has exceeded $29 billion, but its rapid expansion has also been accompanied by leadership changes and market challenges. This article will comprehensively introduce this AI unicorn from the aspects of company background, founder story, products and services, achievements and impact, and recent developments.

Company background and history

Scale AI was founded on June 1, 2016 and is headquartered in San Francisco, USA. It is a private enterprise focusing on the field of information technology. The company originated from the founder's deep insight into the bottleneck of AI data: at that time, there was a lack of sufficient annotated data for AI model training, resulting in low development efficiency. Scale AI's mission is to "accelerate the development of AI applications" and help enterprises bridge human intelligence and machine learning capabilities by building data center infrastructure.

The company focused on the field of autonomous driving in its early days, because autonomous driving requires massive amounts of human-labeled image data to train the AI ​​system. In 2016, Scale AI was selected into the Y Combinator accelerator and received US$120,000 in seed round financing. Subsequently, it quickly expanded into areas such as satellite imagery, e-commerce, and robotics. In 2018, the company’s founder was named to Forbes’ “30 Under 30” list, marking its early success. By 2024, Scale AI's annual revenue has reached $870 million, and is expected to double to $2 billion in 2025.

In its financing history, Scale AI has raised a total of US$15.9 billion and gone through 9 rounds of financing. The E round in 2021 is US$325 million, with a valuation of US$7.3 billion; the F round in 2024 is US$1 billion, with a valuation of US$13.8 billion; in June 2025, Meta invested US$14.3 billion and acquired 49% of the shares (without voting rights), pushing the company's valuation to nearly US$29 billion. This round of financing not only injected huge funds, but also triggered a huge change in leadership: founder Alexandr Wang left Scale AI and joined Meta to lead the AI ​​super intelligence laboratory, while chief strategy officer Jason Droege was promoted to CEO.

Scale AI's customer base covers three major areas: generative AI companies (such as OpenAI, Nvidia, Cohere), the US government (such as the Army, Air Force, Defense Innovation Unit) and enterprises (such as General Motors Cruise, Zoox, Toyota, Airbnb, Lyft). These customers rely on Scale AI's data services to train models and drive innovation from autonomous driving to defense applications.

Founder’s Story: From Campus to AI Pioneer

The souls of Scale AI are co-founders Alexandr Wang and Lucy Guo. Wang was born in 1997. He entered the Massachusetts Institute of Technology (MIT) to study computer science in 2015. He dropped out of school to start a business after only one year due to perfect grades. His early projects included a refrigerator camera to detect milk levels, but he found that insufficient data annotation was an AI bottleneck, which led to the idea of ​​Scale AI. Guo is from Carnegie Mellon University. In 2014, he received US$100,000 from the Thiel Fellowship to drop out and start a business. She has interned at Facebook and worked as a product designer at Quora and Snapchat. The two met while working together at Quora and founded the company together.

Wang is known for his keen sense of AI trends. He has studied the success of CEOs such as Jobs and Musk, and predicted that AI will face the challenge of "data walls". After joining Meta in 2025, he continued to serve as a director of Scale AI's board of directors to promote the company's transformation into a deeper AI infrastructure. Guo left in 2018 to found Backend Capital venture capital fund.

Products and Services: Building the full life cycle of AI

As of October 2025, Scale AI's product line is divided into three pillars: Build AI (build AI), Apply AI (application AI) and Evaluate AI (evaluate AI), covering the entire process of machine learning models from data collection to deployment. The company supports multiple data types such as images, videos, texts, audios and maps, and uses automated annotation, pure manual or human-machine integration (HITL) methods to ensure high-precision "ground truth" data.

Build AI: At the core is Scale Data Engine, an end-to-end data platform for generative AI, government and automotive applications. It manages a global workforce of 240,000 (through subsidiary RemoTasks, which operates in Kenya, the Philippines and Venezuela) and provides data collection, collation and annotation services. Sub-products include:

Apply AI: Focus on actual deployment, including Scale Donovan (government-specific AI suite) and Scale GenAI Platform (generative AI platform).

Evaluate AI: Solve the pain points of model evaluation, such as data shortage and bias detection, through the Scale Evaluation tool. SEAL (Security, Assessment and Analysis Lab) uses thousands of red team members to test for vulnerabilities. SEAL Leaderboards, launched in May 2024, uses private data sets to rank cutting-edge LLM (such as coding and mathematics fields), and was selected by the White House as a public AI assessment partner.

These products adopt enterprise-customized pricing and self-service payment models. The first 1K marked units are free, and the gross profit margin is about 50-60%. Scale AI has also experimented with Synthetic, Document AI, etc., but some of them have been discontinued.

Achievements, Impact and Challenges

Scale AI occupies the leading position in the AI data market (USD 4.9 billion in 2025, expected to be US$13.8 billion in 2030), helping customers such as Toyota and OpenAI improve model performance and promote autonomous driving and defense innovation. In 2025, it was selected for the CNBC Disruptor 50 list and cooperated with NVIDIA and others for LLM fine-tuning.

However, after Meta's investment in 2025, the company faced challenges: Google (the largest customer) transferred contracts, and Microsoft, OpenAI, etc. suspended cooperation, raising concerns about neutrality. In July, Scale AI laid off 14% of its workforce (about 200 people), mainly affecting the generative AI team. It also laid off 20% of its employees in 2023. Additionally, the U.S. Department of Labor investigated labor standards violations and worker class action lawsuits alleged misclassification, unpaid overtime and exposure to harmful content.

Conclusion: AI’s future data guardians

Scale AI’s journey from a Y Combinator start-up to an AI infrastructure giant valued at US$29 billion reflects the central role of data in the AI revolution. Despite the recent turmoil, the company will still lead the AI ​​data ecosystem with its strong product line and global customer base. Looking to the future, Scale AI may further integrate generative AI and enterprise applications under the influence of Wang's Meta to promote intelligent collaboration between humans and machines. As Wang said, the next six months of AI will witness the breakthrough of data walls—Scale AI is at the forefront.