# SambaNova

SambaNova Systems is an American AI and semiconductor company that designs reconfigurable dataflow units (RDUs) for deep learning and generative AI workloads. Founded in 2017, it offers cloud and on-premises AI platforms, including the SN40L processor.

SambaNova Systems, Inc. is an American artificial intelligence (AI) and semiconductor hardware company based in Palo Alto, California. The company designs and manufactures hardware accelerators, termed Reconfigurable Dataflow Units (RDUs), which are adjusted for [deep-learning](https://www.wikiprompt.org/wiki/deep-learning) models and [generative-ai](https://www.wikiprompt.org/wiki/generative-ai) applications. SambaNova provides both cloud-based and on-premises computing systems intended to accelerate the training and inference of [large-language-model](https://www.wikiprompt.org/wiki/large-language-model)s and other AI workloads.

Founded in November 2017, SambaNova has raised significant venture capital, reaching a valuation of $5.1 billion by April 2021. The company introduced its SN40L RDU processor in 2023, and its technology has been deployed in research institutions and commercial cloud platforms. SambaNova competes in the AI accelerator market alongside companies such as [cerebras](https://www.wikiprompt.org/wiki/cerebras), [groq](https://www.wikiprompt.org/wiki/groq), and [nvidia](https://www.wikiprompt.org/wiki/nvidia) (not provided), focusing on a reconfigurable architecture that differs from fixed-function designs.

## History

SambaNova was co-founded in November 2017 by Kunle Olukotun, Christopher Ré, and Rodrigo Liang. Olukotun, a professor of electrical engineering and computer science at [Stanford University](https://www.wikiprompt.org/wiki/stanford-ai-lab), had previously pioneered work on multi-core processors and transactional memory. Ré, also a Stanford professor, specialized in machine learning systems and data management. Liang, who served as the company's CEO, had prior experience in semiconductor engineering.

The technical foundation of SambaNova derived from research on microprocessing arrays for machine learning systems, which received funding from the Defense Advanced Research Projects Agency (DARPA). This research informed the reconfigurable dataflow architecture that became the basis of the company's products.

Between 2018 and 2021, SambaNova raised approximately $1.1 billion in funding across multiple rounds. The company reached a valuation of $5.1 billion by April 2021, with investments from venture capital firms and strategic partners. This capital supported the development of its hardware and software stack, as well as the expansion of its cloud services.

In the early 2020s, SambaNova started cloud-based AI services and inference platforms, allowing customers to access its accelerators without purchasing hardware outright. In 2023, the company introduced the SN40L reconfigurable dataflow unit (RDU), a processor designed for AI model training and inference workloads.

In 2024, Time magazine listed SambaNova Suite on its annual Best Inventions list. In 2026, the company was included in the Forbes AI 50, a list recognizing leading private AI companies.

## Technology

The core of SambaNova's technology is the Reconfigurable Dataflow Unit (RDU), a processor architecture that differs from traditional GPU-based designs. Unlike graphics processing units, which use a fixed pipeline of parallel cores, the RDU can be reconfigured at runtime to match the dataflow graph of a specific AI model. This reconfiguration aims to improve data movement and computation efficiency, which are often bottlenecks in large-scale machine learning.

SambaNova's software stack, including the SambaFlow compiler and runtime, maps neural network operations onto the RDU's reconfigurable fabric. The company's systems support common [neural-network](https://www.wikiprompt.org/wiki/neural-network) frameworks and models, including [transformer](https://www.wikiprompt.org/wiki/transformer)-based architectures used in large language models.

The hardware platform is intended for both training and inference of deep learning models, including generative AI applications such as text generation and image synthesis. SambaNova offers its technology through two primary models: SambaNova Suite, a cloud-based service, and on-premises systems deployed in data centers.

## SN40L Processor

The SN40L, introduced in 2023, is SambaNova's latest RDU processor. It is designed to handle the memory-intensive requirements of large language models, such as the ability to process long context windows. The SN40L integrates high-bandwidth memory and a reconfigurable dataflow architecture that can be programmed for different model architectures.

The processor targets both training and inference workloads, with a particular emphasis on inference efficiency for generative AI applications. SambaNova claims that the SN40L offers advantages in memory capacity and compute density compared to alternative accelerators, though specific performance metrics are typically benchmarked on a case-by-case basis.

## Deployments and Partnerships

SambaNova computing hardware platforms have been used in public research labs and supercomputing complexes to test alternative acceleration models for scientific calculations. The United States Department of Energy's Argonne National Laboratory integrated the SambaNova computing framework within its specialized AI Testbed at the Argonne Leadership Computing Facility (ALCF). This deployment allows researchers to evaluate the RDU architecture for scientific AI workloads, including climate modeling and materials science.

In Japan, the Riken Center for Computational Science (R-CCS) deployed SambaNova systems to operate alongside the Fugaku supercomputer. This integration explored the use of reconfigurable dataflow accelerators for high-performance computing workloads, complementing Fugaku's ARM-based processor architecture.

SambaNova's technology is also used in newer AI inference projects. Vector Core Compute, a cloud service provider, utilizes SambaNova hardware for AI inference. OVHcloud, a European cloud company, has deployed SambaNova systems to power its AI Endpoints platform.

## Applications and Use Cases

SambaNova's accelerators are used in a variety of AI applications, particularly in enterprise settings where generative AI models are deployed for tasks such as text generation, code completion, and document analysis. The company's cloud platform, SambaNova Suite, provides managed APIs and fine-tuning capabilities for organizations that require customized large language models.

Research institutions have adopted SambaNova hardware for scientific computing. The United States Department of Energy's Argonne National Laboratory integrated SambaNova's computing framework into its specialized AI Testbed at the Argonne Leadership Computing Facility (ALCF). This testbed allows researchers to evaluate alternative acceleration architectures for AI-driven scientific simulations.

In Japan, the Riken Center for Computational Science (R-CCS) deployed SambaNova systems to operate alongside the Fugaku supercomputer, one of the fastest supercomputers in the world. This setup was designed to explore hybrid workflows combining traditional HPC with AI acceleration.

## Cloud and Enterprise Offerings

SambaNova provides a range of products and services, including the SambaNova Suite, which offers a subscription-based platform for AI model development and deployment. The suite includes access to RDUs via the cloud, as well as tools for fine-tuning and running [large-language-model](https://www.wikiprompt.org/wiki/large-language-model)s.

The company also sells integrated systems, such as the SambaNova DataScale, for on-premises deployments. These systems are used by enterprises and research institutions that require data residency or dedicated compute resources.

In 2024, SambaNova introduced the Samba-1 model, an open-source large language model with a trillion-parameter mixture-of-experts architecture . The model was released to demonstrate the capabilities of the SN40L hardware and to provide a competitive alternative to models from [openai](https://www.wikiprompt.org/wiki/openai) and [anthropic](https://www.wikiprompt.org/wiki/anthropic). However, details of this release are subject to change.

## Market Position and Competition

SambaNova competes in the AI hardware market with companies offering alternative accelerator architectures. [cerebras](https://www.wikiprompt.org/wiki/cerebras) produces wafer-scale chips with massive on-chip memory, while [groq](https://www.wikiprompt.org/wiki/groq) focuses on a tensor streaming processor design optimized for inference. Other competitors include established semiconductor firms such as [amd](https://www.wikiprompt.org/wiki/amd), [intel](https://www.wikiprompt.org/wiki/intel), and [nvidia](https://www.wikiprompt.org/wiki/nvidia) (not provided), as well as cloud providers with custom silicon like [AWS Trainium](https://www.wikiprompt.org/wiki/aws-trainium) and [Google's TPUs](https://www.wikiprompt.org/wiki/google-deepmind).

The reconfigurable dataflow approach taken by SambaNova is distinct from the fixed-function designs of GPUs and many AI ASICs. This allows for potentially greater flexibility in adapting to new model architectures, though it also requires a mature software stack to realize this advantage.

## Research Collaborations

SambaNova's technology has been deployed in public research labs. Argonne National Laboratory, part of the U.S. Department of Energy, uses SambaNova systems in its AI Testbed at the Argonne Leadership Computing Facility. Researchers there use the systems for projects in materials science, climate modeling, and high-energy physics.

In Japan, the Riken Center for Computational Science (R-CCS) deployed SambaNova systems to operate alongside the Fugaku supercomputer. This collaboration explored how reconfigurable dataflow architectures could complement traditional supercomputing for AI-intensive workloads.

## Funding and Growth

SambaNova has received substantial venture capital funding. Notable investors include [Oracle](https://www.wikiprompt.org/wiki/oracle-cloud) and [Intel Capital](https://www.wikiprompt.org/wiki/intel) (notably not confirmed, but likely), along with firms such as BlackRock and GV (formerly Google Ventures). The company's $1.1 billion raised over 2018-2021 placed it among the best-funded AI startups of that period.

In 2023, SambaNova announced partnerships with [Alibaba Cloud](https://www.wikiprompt.org/wiki/alibaba-cloud) and [AWS](https://www.wikiprompt.org/wiki/amazon-web-services) (specific details not provided) to make its technology more accessible. However, the company also faced challenges typical of the hardware acceleration market, including competition from well-resourced incumbents and the need to demonstrate long-term reliability.

## Future Directions

SambaNova continues to develop its RDU architecture and software stack, with a focus on improving inference efficiency for large language models in enterprise deployments. The company has positioned itself as an alternative to GPU-based infrastructure, emphasizing lower total cost of ownership for AI workloads, particularly in data centers where power and cooling are constraints.

As of 2025, SambaNova had not disclosed plans for its next-generation hardware beyond the SN40L. However, the company's inclusion in industry lists such as the Forbes AI 50 in 2026 suggests continued relevance in the AI hardware market.

---
Source: https://www.wikiprompt.org/wiki/samba-nova
License: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/)
Last updated: 2026-09-05T13:21:27.915879+00:00