Meta Takes a Major Leap Forward with Scale AI
A year ago, the social media giant led by Mark Zuckerberg made a significant investment in the American tech startup Scale AI, participating in a funding round that raised one billion dollars. Since then, their partnership has only deepened, culminating in a substantial new funding agreement that values Scale AI at approximately $29 billion—double its valuation following last year’s Series F funding round.
This recent deal involves Meta investing an impressive $14.3 billion in exchange for around 49% ownership stake, although it appears that Meta does not hold voting rights in the company. The deal’s valuation places Scale AI among the most highly valued AI companies globally, reflecting Meta’s strong interest in harnessing the startup’s capabilities for advancing its own AI and data processing strategies.
Previously, the company was mostly backed by traditional venture capital firms such as Accel Ventures, Coatue, Founders Fund, Index Ventures, and Y Combinator. Notably, NVIDIA was also a key investor, alongside major tech giants like Amazon, AMD, Cisco, Intel, Qualcomm, and ServiceNow, who have all entered Scale AI’s capital during its growth phase.
From a Promising Startup to a Key Player in AI Since 2017
Scale AI’s journey began in the summer of 2017 with its Series A funding of $4.5 million led by Accel Ventures, when the company was still in its infancy. At that time, it offered a modest range of API tools designed for creating labeled datasets crucial for training artificial intelligence models—covering data collection, image annotation, Optical Character Recognition (OCR), and categorization.
Over the subsequent years, the company experienced rapid growth through annual funding rounds, attracting notable figures from the tech world—such as Adam D’Angelo, founder of Quora and Facebook’s first CTO; Drew Houston, CEO of Dropbox; Justin Kan, co-founder of Twitch; along with Instagram founders Mike Krieger and Kevin Systrom. During this period, Scale AI expanded through acquisitions, acquiring companies like Helia AI, which specialized in real-time video processing, and SiaSearch, focusing on data management solutions.
The company also engaged in strategic partnerships, such as one with MIT Media Lab to analyze and mitigate biases in dermatological datasets, emphasizing its commitment to improving AI fairness and robustness.
Collaborations and Expanding Influence in 2023
In 2023, Scale AI’s relationship with Meta became more prominent through a partnership involving Meta’s Llama 2 large language model (LLM). Just weeks prior, the startup joined forces with Anthropic to integrate Claude, a competitive AI model, into its platform—a platform designed for training and evaluating AI models. The U.S. Army also signed a contract to utilize Scale AI’s technology in its program for robotic combat vehicles.
That same year, OpenAI identified Scale AI as its preferred partner for fine-tuning GPT-3.5, after initially employing the startup for early reinforcement learning experiments on GPT-2. Following this, Amazon Web Services (AWS) collaborated with Scale AI to customize and evaluate models on Amazon Bedrock, illustrating the company’s growing influence within the industry’s leading AI platforms. Simultaneously, Scale AI developed its own large language models, especially targeting applications within the automotive sector.
Growing Contracts and Focus on Defense and Government Projects
Today, Scale AI employs around 900 people and is projected to generate nearly one billion dollars in revenue in 2024. Its product offerings are structured into three primary divisions:
- Scale Data Engine: Focuses on data collection, preparation, training, and evaluation of models, serving clients such as Instacart, Microsoft, Pinterest, and Square.
- Scale GenAI Platform: Develops applications based on their models, utilized by companies like Cisco and media organizations such as TIME.
- Scale Donovan: Specializes in deploying AI agents, notably used by the U.S. Army and U.S. Air Force for defense-related projects.
In recent months, Scale AI has expanded its international footprint by signing agreements with Qatar’s Ministry of Telecommunications & Information Technology and the Emirati holding company G42. The latter involves creating sector-specific evaluation datasets and assisting in the post-training fine-tuning of large language models for G42’s AI subsidiary, Inception.
Meanwhile, its defense collaborations within the United States have intensified. The company has joined the Thunderforge program, which focuses on integrating AI into military planning, and has developed a specialized version of its Llama 3 model called Defense Llama, tailored for military applications.
The company’s co-founder and key executive, Alexandr Wong—who hails from the Massachusetts Institute of Technology—has joined Meta’s "superintelligence" team dedicated to the development of Artificial General Intelligence (AGI), underscoring the strategic importance of Scale AI in the broader AI ecosystem.
