Loop.AI Hits $4.2B Powering Enterprise AI Agents Powered by Client-Trained SLMs Running at the Edge

The world's first independent AI company delivers what few can: a platform where clients train custom SLMs on their own data to run AI agents and cost-effective inference on-premises or at the edge - full enterprise sovereignty over AI.

SAN FRANCISCO - December 12, 2025 - Loop.AI, the pioneer in custom small language models and vertical agentic AI since 2012, today announced its recent $100M funding round at a valuation of $4.2 billion led by AI Fund I with participation from new and existing investors to support continued development. While the market is flooded with new startups reselling generic public models, Loop.AI leverages more than a decade of deployments to deliver what others are still piloting: autonomous AI agents performing real job functions within the org-chart of large organizations. Loop.AI enables clients to train entirely new Small Language Models on their own proprietary data-a technically demanding process that Loop has streamlined so any enterprise engineer can handle it without deep expertise in model architecture, training infrastructure, or data pipelines.

The "Post-Commodity" Thesis. General-purpose Large Language Models (LLMs), despite reaching commodity status, still incur high costs when applied to real-world agentic applications. Loop.AI moves beyond the "black box" hype with an AI Control Center that gives enterprises full sovereignty over AI agents integrated into their organizational structures. By decoupling the AI model from the application logic, Loop empowers organizations to navigate the current chaos and continuous incremental changes in AI model performance with confidence, avoiding vendor lock-in.

Solving the "Last Mile" of Enterprise AI. Loop.AI's dual-platform architecture solves the "last mile" problem of enterprise AI using NVIDIA GPUs either on prem or in the cloud:
• Loop Q (Cognitive Platform): Custom-trains Small Language Models (SLMs) on proprietary client data with local inference delivering maximum precision on-premises or in the client's cloud while minimizing costs and hallucinations for most agentic use cases.
• Loop AI Agents Orchestra (The Agentic AI Control Center): A no-code visual interface that manages the entire agent lifecycle. It allows teams to drag-and-drop any AI provider (Open Source or Commercial) or the Loop Q SLM, benchmark performance vs. cost in real-time, and deploy agents that integrate securely with legacy systems.

Leadership Perspectives. "We bet on specialized, client-trained models back in 2012 and have been collaborating with NVIDIA since 2014-long before their first ML-specific GPU hit the market. Today, even Jensen Huang agrees the future of agentic applications isn't just about giant LLMs; it's about efficient, small language models that run on the edge and get real agentic work done with high accuracy. While large public models are ubiquitous, they are often too slow and risky for real enterprise agents. With Loop Q and Loop AI Agents Orchestra, perfected over a decade of large-scale implementations, we empower clients to train and run multimodal Small Language Models on their own servers, hot-swap them instantly, or complement them with any LLM of their choice only when necessary. Our unified control panel gives enterprises full oversight and management of all their AI, regardless of vendor." said GM Calafiore, Founder and President.

"Loop's true differentiator isn't merely technological proof; it's a decade of relentless AI research and platform development forged exclusively by real-world production deployments," said Andrea Pitrone, COO and Chief Product Officer. "This hard-won expertise-the synthesis of SRI/Stanford rigor and 20 years of Accentur-scale enterprise delivery-is the foundational code of Loop Orchestra and Loop Q. It is an unparalleled advantage difficult to replicate both for startups and large software companies. We have demonstrated that local, client-built, Small Language Models power agentic applications with the high and consistent accuracy required for enterprise use, exemplified by an AI agent autonomously processing millions of dollars in financial transactions for a Fortune 100 bank for several years with no issues. Training custom SLMs specific to the agentic use case on client data is rare-most vendors just fine-tune or resell generic models. We've built the infrastructure so clients can train their own models without data scientists, focusing instead on their AI agent workflows. SLMs deliver the accuracy enterprises actually need. This codified experience empowers enterprises to effortlessly deploy AI agents, train proprietary models, and automate complex work-falling back to any LLM provider for specific use cases, guaranteeing zero vendor lock-in."

For more information, visit https://www.loop.ai
Full Press Relelase: https://u.loop.ai/Za5Jvh

Contacts
Media: press@loop.ai
Investors: ir@loop.ai

535 Mission St
San Francisco, CA 94105
USA

Founded in 2012 in San Francisco, Loop.AI is a private, research-driven applied lab and the world's first independent enterprise AI company. Drawing on the elite scientific heritage of the Artificial Intelligence Center at SRI International and Stanford, its platforms-Loop Q and Loop AI Agents Orchestra-enable large organizations to integrate AI agents directly into their organizational structures using low-latency, cost-efficient Small Language Models trained on client data and running inference at the edge, while remaining fully compatible with any AI provider as needed. While scaling its classical AI solutions to streamline agent deployments, LoopQuantum.AI is independently researching the next generation of AI based on quantum-first principles.

This release was published on openPR.