Super Architecture for Super AI: How Matrix Built doubleAI’s Supercomputing Infrastructure
The AI Arms Race
The world has seen revolutions before, but nothing prepared the market for the AI revolution.
Within just a few months, artificial intelligence moved from a distant promise in research labs to a critical working tool transforming industries end to end.
And as always, behind the social media buzz and the excitement lies a complex physical reality: AI is an extremely hungry consumer. It demands computing power, energy, and bandwidth at unprecedented scale.
As a result, the biggest challenge for technology companies is no longer just writing code. We have returned to infrastructure. Advanced AI development requires data centers capable of handling enormous thermal loads, massive power consumption, and data transfer at near-light speed. Without flawless physical and architectural infrastructure, even the most brilliant algorithm will hit a hardware bottleneck.
This is where Matrix enters the picture. At a time when everyone is talking about what AI can do, Matrix has had a unique opportunity to act as an enabler, helping lay the foundations. The company’s ability to take on complex end-to-end projects, design architecture from scratch, and deploy infrastructure that allows supercomputers to perform at peak output is what turns a startup vision into working reality.
doubleAI: Cracking the Code of Superintelligence
When doubleAI was founded by Prof. Amnon Shashua, Prof. Shai Shalev-Shwartz, Prof. Yoav Levine, Prof. Or Sharir, Dr. Noam Weiss, and Dr. Gal Binyamini, it did not join the generative AI revolution. It set out to lead its next stage: reasoning-based AI.
While most current AI solutions are built on word prediction and probabilities, doubleAI was designed around the concept of Expert AI. Its goal is to create a research tool of absolute reliability, capable of solving complex scientific, mathematical and computational problems at the highest level of human expertise.
What differentiates doubleAI is its ambition for superintelligence. Beyond AI, the company is pursuing what it calls A-Class AI: a system that does not merely summarize existing knowledge, but can produce research breakthroughs, identify deep logical relationships, and serve as a strategic partner in scientific discovery. For researchers and scientists, the system is envisioned as nothing less than an engine of truth, where every answer is grounded in rigorous logic and critical data analysis.
Turning that vision into reality required computing power on an extraordinary scale. Research based on logical inference and real-time data correlation requires an architecture that does not simply process data, but effectively “thinks” in nanoseconds. The need for zero-time response is not just about user experience; it is essential to allowing the system to scan enormous solution spaces and reach reliable conclusions.
As a unicorn startup operating in relative stealth, doubleAI needed a partner capable of building this technological lab with the speed of a startup and the precision of a research laboratory. Naturally, it turned to NVIDIA, whose computing platforms power the AI revolution.
To build the infrastructure, doubleAI selected Matrix IT, recognizing that its exceptional requirements demanded an integrator with pioneering expertise, one deeply familiar with the inner workings of DGX systems and capable of architecting a full end-to-end solution.
That is how Matrix entered the picture. The mission was to build from scratch, within just weeks, the physical and logical infrastructure that would allow the brilliant minds at doubleAI to run their advanced algorithms on world-class architecture. The race had begun.
The Technological Challenge: Engineering a Supercomputer in Zero Time
Yaniv Perry, Senior Strategic AI & HPC Consultant at Matrix IT, shares a perspective from the field:
“The trend everyone talks about globally is AI capabilities. They are extraordinary, but we are here to remind people that this does not happen at the push of a button. Behind every prompt lies powerful physical infrastructure. Everything depends on it: load balancing, resource utilization, resilience, productivity and capacity. We had to deliver a solution strong enough not just for today, but for years ahead in a fast-moving technological world.”
To realize its superintelligence vision, doubleAI needed a research powerhouse. That is where the NVIDIA DGX BasePod B200 entered the picture.
This AI infrastructure was designed for the heavy computational demands of modern AI and enables complex reasoning models to run at double the speed of the previous generation, delivering the zero-time response critical to research.
Accordingly, Matrix designed the entire supporting system, from stable power allocation to airflow and cooling optimization throughout the data center, ensuring the system could operate at peak performance around the clock.
One of the key elements in achieving the zero-time performance required by doubleAI was the communications architecture. To prevent data bottlenecks between servers, Matrix implemented the NVIDIA Quantum-2 NDR communications platform — effectively the nervous system of the AI environment, enabling GPU chips to communicate in full synchronization and function as a single computational unit.
In parallel, a perimeter security layer was established with advanced firewalls and a managed development environment, including Dell Technologies servers for storage and management layers, allowing doubleAI to focus entirely on its research.
A major achievement of the project was taking all these edge components, from servers and switches to monitoring and support systems, and turning them into a fully operational environment within just weeks.
It was the work of true pioneers, building a complete architecture from the ground up while managing complex logistics and planning, ensuring that when called upon, the supercomputer would meet expectations and support go-to-market execution in practice.
The Defining Moment, Against All Odds and in Record Time
Ultimately, projects like this are not measured by the number of servers or switch speeds, but by the ability to perform under uncompromising pressure, meet strict demands, and effectively adapt and deliver in real time.
For doubleAI, our finish line was their starting line.
With product launches planned and major fundraising rounds ahead, every day of delay in infrastructure meant delay in the broader vision.
Matrix entered the project not as an outside contractor, but as a full strategic partner. Our teams understood the magnitude of the moment. Building a complex supercomputing environment from scratch in a matter of weeks required relentless focus, engineering creativity, and total commitment.
The teams did not just design and deploy. They managed the project end to end, from physical installation through monitoring, support and dedicated support, ensuring doubleAI could push the button and launch without disruption.
The success story of doubleAI and Matrix is proof that when bold vision meets infrastructure excellence, the sky is no longer the limit.
For Matrix, it is yet another demonstration of its role as a pioneer — not merely following developments, but building the infrastructure on which the future will run.
