Block Becomes First Company in North America to Deploy the Latest NVIDIA GB200 Systems for Frontier Models
Published on Wednesday, March 12, 2025
Block is proud to announce that we’re the first company in North America to deploy NVIDIA’s latest GB200 systems for training frontier AI models. This infrastructure investment reflects our commitment to advancing open source AI research that empowers economic participation.
Why This Matters
AI is transforming how people interact with technology, and we believe the most impactful AI systems will be:
- Open source - Transparent, auditable, and improvable by the community
- Purpose-driven - Focused on solving real problems for real people
- Accessible - Available to individuals and small businesses, not just large enterprises
The GB200 systems give us the computational power to train large-scale models while maintaining our commitment to open source development.
Our AI Research Focus
Block’s AI research centers on economic empowerment. We’re exploring how AI can:
Democratize Financial Services
Making sophisticated financial tools accessible to small businesses and individuals who traditionally lack access to advanced technology.
Improve Developer Productivity
Building tools like goose that help developers work more efficiently without sacrificing control or understanding.
Enhance Customer Experience
Creating AI systems that make our products easier to use while respecting user privacy and autonomy.
Support Creative Work
Developing tools that augment human creativity rather than replace it, particularly for artists and content creators.
Partnership with NVIDIA, Equinix, and Lambda
This deployment is the result of collaboration with three key partners:
NVIDIA
Providing the cutting-edge GB200 systems that enable frontier model training at scale.
Equinix
Offering the data center infrastructure and connectivity that supports high-performance computing workloads.
Lambda
Delivering the orchestration and management tools that make these systems accessible to our research teams.
Together, these partnerships give us world-class AI infrastructure while allowing us to focus on research and development.
Open Source AI at Scale
Having access to frontier-scale compute doesn’t mean we’re abandoning open source. In fact, it’s the opposite:
Training Open Models
We’re using these systems to train models that we’ll release as open source, enabling other researchers and developers to build on our work.
Publishing Research
Our findings will be published openly, contributing to the broader AI research community’s understanding of large-scale model training.
Sharing Tools and Techniques
We’ll open source the tools and techniques we develop for training and deploying large models efficiently.
Supporting the Ecosystem
Through programs like the Goose Grant Program, we’re supporting others building open source AI tools.
Responsible AI Development
With great computational power comes great responsibility. We’re committed to:
Safety and Alignment
Ensuring our models behave in ways that align with human values and intentions.
Bias Mitigation
Actively working to identify and reduce biases in our training data and model outputs.
Privacy Protection
Developing techniques that enable powerful AI while protecting individual privacy.
Environmental Sustainability
Optimizing our training processes to minimize energy consumption and environmental impact.
What We’re Building
While we can’t share all the details yet, our research teams are working on:
Next-Generation AI Agents
Building on goose to create more capable, reliable, and controllable AI agents.
Financial AI Models
Developing models specifically trained on financial data and use cases.
Multimodal Systems
Creating AI that can work across text, code, images, and other modalities.
Efficient Training Techniques
Researching methods to train powerful models more efficiently, making frontier AI more accessible.
The Open Source Advantage
Some might question why we’re investing in frontier AI infrastructure while maintaining our open source commitment. We believe open source is actually an advantage:
Faster Innovation: Open development allows us to incorporate community contributions and feedback.
Better Alignment: Public scrutiny helps ensure our models align with user needs and values.
Broader Impact: Open source models benefit far more people than proprietary alternatives.
Trust and Transparency: Users can inspect and understand how our models work.
Join Our Research
We’re hiring researchers, engineers, and scientists to work on these challenges. If you’re passionate about:
- Open source AI development
- Economic empowerment through technology
- Responsible AI practices
- Large-scale model training
We’d love to talk to you. Visit block.xyz/careers to learn more.
Looking Ahead
This infrastructure deployment is just the beginning. Over the coming months, we’ll be:
- Publishing research papers on our findings
- Releasing open source models and tools
- Sharing insights from large-scale training
- Building partnerships with other open source AI projects
Stay Connected
Follow our progress:
The future of AI should be open, accessible, and focused on empowering people. With this infrastructure investment, we’re taking concrete steps toward that vision.
Interested in collaborating on open source AI research? Contact us at ai-research@block.xyz