How Block is Leading Open-Source AI with NVIDIA DGX SuperPOD
The AI infrastructure race continues to accelerate as companies seek computational advantages to develop increasingly sophisticated models.
This race has intensified with each new generation of specialised hardware, as firms compete for the capability to train and deploy frontier AI models – those at the cutting edge of capabilities that push beyond current limitations.
In this competition, access to the most advanced computing systems has become a crucial differentiator, particularly for companies aiming to develop open source alternatives to proprietary systems.
Whilst large technology corporations have dominated headlines with their massive AI investments, financial technology (fintech) firms are also emerging as significant players in the AI infrastructure space.
Now, Block, the fintech company that provides payment solutions and digital banking services started by Jack Dorsey (who created Twitter), announced it will be the first company in North America to deploy the NvidiaDGX SuperPOD with DGX GB200 systems.
What are Nvidia DGX SuperPOD and DGX GB200 systems?
The Grace Blackwell-powered Nvidia DGX SuperPOD is an AI supercomputer designed to tackle complex AI tasks, such as training large language models (LLMs) and Gen AI.
It integrates multiple DGX systems – which are purpose-built for AI and deep learning, featuring advanced GPUs, optimised software, high-speed networking, efficient liquid cooling for scalability and performance.
The DGX GB200, part of this system, uses Nvidia's latest Blackwell GPUs to handle massive datasets and multi-trillion-parameter AI models.
Together, they provide enterprises with a powerful, energy-efficient infrastructure for accelerating AI innovation in industries like healthcare, robotics and autonomous systems.
This combination ensures faster insights, reduced downtime and optimised costs for high-performance AI workloads.
How Nvidia’s systems are enhancing Block
Block's previous AI research efforts have focused on deepfake detection – the identification of synthetically generated media that could be used for deception – and gained recognition for developing early examples of hyper-realistic generated audio.
“Our past work on generative speech models means Block is well-positioned for innovation surrounding Gen AI models and all possible applications that come with it,” says Dhanji R. Prasanna, Chief Technology Officer of Block.
“The industry and the world is undergoing a seismic shift with adopting AI tools. At Block, we think it's essential not only to apply AI to existing problems, but also to explore, learn and build in the open so that we can advance the frontier of AI in a way that truly levels the playing field for our customers and community.
“We're excited to deploy an NvidiaGrace Blackwell DGX SuperPOD and start exploring novel solutions for our customers,” he adds.
Charlie Boyle, Vice President of DGX platforms, at Nvidia says: “With Nvidia DGX GB200 systems, Block engineering and research teams can develop frontier open-source AI models that can tackle complex, real-world challenges with state-of-the-art AI supercomputing.”
The power of Nvidia systems combined with Lambda’s
Lambda is a serverless computing service by Amazon Web Services (AWS), running code in response to events without server management.
Block also selected Lambda's 1-Click Clusters – which are collections of GPUs that work together to handle intensive computing tasks for AI training, featuring Nvidia H100 GPUs and InfiniBand networking – as their AI cloud partner to test hypotheses before full-scale deployment.
These GPU clusters also offer access to hundreds of interconnected Nvidia GPUs for short-term use, supported by machine learning (ML) engineers.
Lambda has now updated these clusters to include Nvidia's new Blackwell architecture.
“As AI models grow in complexity and scale, businesses need powerful infrastructure that can match the pace of innovation,” says Charlie.
Equinix data centres provide crucial infrastructure for Block's AI computing needs
The new infrastructure will be housed at an Equinix data centre and dedicated to researching and training open source Gen AI models.
These facilities offer global connectivity while maintaining data privacy compliance, high performance computing capabilities and low-latency connections to thousands of partner networks and cloud services – latency referring to the delay between sending a request and receiving a response, which is critical for time-sensitive AI applications.
“Frontier models represent the cutting edge of AI technology, pushing the boundaries of what AI can achieve, and they require the latest in AI chips – like Nvidia's new DGX SuperPOD,” says Jon Lin, Chief Business Officer at Equinix.
“By deploying at Equinix's neutral, cloud-adjacent platform, companies like Block can unlock expanded compute scale and flexibility. This enables the customisation of AI solutions with a choice of infrastructure, cloud, models and cooling at our neutral exchange.”
Block’s further development of open source AI
Furthermore, the adoption of Nvidia's latest AI infrastructure follows Block's recent launch of an open source project called “goose” – an interoperable AI agent framework that allows users to connect LLMs to real-world applications.
While initial applications focus on software engineering, developers from both Block and the open source community are investigating additional use cases beyond engineering domains.
“We're committed to an open source approach, sharing our learnings and results along the way,” says Dhanji.
“This infrastructure investment represents our belief that AI should be accessible and beneficial to everyone in the financial ecosystem we serve.”
Explore the latest edition of Technology Magazine and be part of the conversation at our global conference series, Tech & AI LIVE.
Discover all our upcoming events and secure your tickets today.
Technology Magazine is a BizClik brand
