Foundry

Company Overview

Foundry is a technology company that provides elastic GPU compute services for AI developers. Founded by alumni from Google DeepMind’s core Deep Learning Team and Stanford’s computer science PhD program, Foundry aims to make cloud compute more accessible, efficient, and powerful for AI development.

The company is headquartered in Palo Alto in the San Francisco Bay Area. Foundry positions itself as a “new breed of public cloud” with an orchestration platform that simplifies access to AI compute resources.

Products Overview

Foundry’s main product offering is its cloud platform that provides on-demand access to NVIDIA GPUs for AI workloads. Key features include:

1. Reserved Instances: Allows users to reserve GPU compute capacity for as little as 3 hours up to several weeks. Users can resell unused capacity.

2. Spot Instances: Provides burst compute capacity at lower prices when available. Users can set maximum price thresholds.

3. NVIDIA Tensor Core GPUs: Access to high-end GPUs like H100s, A100s, A40s, and A5000s without long-term contracts.

4. High-Performance Networking: 3.2 Tbps InfiniBand networking for distributed training.

5. Native Kubernetes Support: For simplified workload orchestration and scaling.

6. Developer Tools: Custom startup scripts, CLI/API management, co-located storage.

The platform is designed for AI engineers, researchers, and scientists working on training, fine-tuning, and inference workloads.

Founding Team

While specific names are not provided, the company states it was founded by alumni from Google DeepMind’s core Deep Learning Team and Stanford’s computer science PhD program. The broader team is described as a group of researchers, engineers, economic experts, and go-to-market leaders with expertise spanning ML theory, distributed systems, and financial modeling.

Problem and Market Fit

Foundry aims to address the challenges AI developers face in accessing and managing high-performance compute resources. The company believes that AI teams shouldn’t need to become hardware experts or spend excessive time on capacity planning.

Their solution targets the growing demand for flexible, scalable GPU compute as AI development accelerates across industries. By offering on-demand access without long-term commitments, Foundry aims to lower the barriers to entry for AI innovation.

Business Model

Foundry operates on a cloud services model, likely charging customers based on GPU usage time and resources consumed. They offer both reserved instances for guaranteed capacity and spot instances for more flexible, cost-sensitive workloads. The ability for users to resell unused reserved capacity suggests a marketplace component to their model as well.

Funding and Runway

The available information does not provide specific details on Foundry’s funding rounds or current runway. However, the company lists several high-profile investors and advisors, suggesting they have secured significant backing. Notable investors/advisors include Eric Schmidt (former Google CEO), Jeff Dean (Google Chief Scientist), and founders/leaders from companies like Databricks and OpenAI.

Competitive Landscape

While not explicitly stated, Foundry likely competes with major cloud providers offering GPU instances for AI workloads, such as Amazon Web Services, Google Cloud Platform, and Microsoft Azure. Their focus on flexibility, high-performance networking, and simplified management aims to differentiate them in this space.

Customers

The website mentions several customer logos, including Anthropic, Arc Institute, and Cohere. Customer testimonials highlight Foundry’s cost-effectiveness, flexibility, and ability to provide guaranteed compute capacity for critical AI research and development work.

Relevant News

The most recent news from Foundry is the announcement of their cloud platform, described in a blog post titled “Restoring the promise of the public cloud for AI” dated August 13, 2024. This suggests the company has recently launched or significantly expanded its core product offering.

Prior to this, Foundry announced its official introduction to the market in a blog post dated March 21, 2024, titled “Introducing Foundry.”