Wilfred Gomes is the co-founder, CEO, and president of Mueon Corporation, a next-generation infrastructure startup rethinking how data centers are built for the AI era. The company’s flagship innovation, Cubelets™, modular, stackable units that unite compute, memory, power delivery and thermal management, replace the traditional rack-based model that have dominated data centers since their inception, enabling up to 10x gains in density, energy efficiency and deployment speed.
He served as a Fellow in Microprocessor Design and Technologies at Intel for nearly three decades, where he drove innovations in advanced packaging, EDA for data center, AI, and client platforms, with a focus on logic, memory and implementation efficiency. He was a co-inventor of Intel’s Foveros 3D integration platform and was instrumental in bringing 3D stacking into mainstream production. With over 250 patents across high-performance CPU and GPU design, Wilfred has played a pivotal role in charting the path toward the next era of AI-scale workloads.
Tell us about your company?
Mueon is redefining how modern data centers are built for the AI era. Founded with the belief that the core infrastructure of computing requires a fundamental change, Mueon is creating a new architectural foundation that unites compute, memory, power delivery, and thermal management into a single, modular system. Our flagship innovation, Cubelets™, replaces the traditional rack-based model that has dominated data centers since its inception, removing the current limits that constrain how silicon systems are built and scaled. Mueon is focused on making data centers more efficient, reliable, powerful, and significantly scalable, while reducing their carbon footprint.
What problems are you solving?
The AI era has pushed data centers to their physical and architectural limits. Power, cost, and complexity are now critical bottlenecks holding AI innovation back. Traditional rack-based systems draw enormous amounts of electricity, generate heat that’s increasingly difficult to manage, and are expensive and slow to scale. Even with massive investment, operators are running into physical limits on how much performance they can extract from traditional architectures.
A major part of the problem lies in the memory hierarchy itself. Today’s compute systems treat memory as a fragmented stack, forcing data to constantly move between layers and adding latency, inefficiencies, and extra costs. Everyone wants memory that’s 10-100 times larger and faster, but no technology exists today to make that possible.
At Mueon, we’re rethinking this from the ground up. Our Cubelet architecture integrates compute, memory, power delivery, and thermal management into a single modular unit designed to bring data and compute closer together – creating high bandwidth and low latency domains. In practice, that means memory that behaves as a single, instantly accessible pool, a capability no traditional system can match. The result is a new class of data infrastructure that eliminates the tradeoffs between power, cost, and performance, and sets the foundation for computing systems with no architectural ceiling.
What application areas are your strongest?
Mueon’s strongest application areas center on three tightly linked domains: memory, power delivery, and thermal management. These are not independent domains; they have to be addressed together. At scale, each one affects the others, and true performance gains come through co-optimization. Mueon’s Cubelet architecture was built precisely for that intersection. By integrating and co-designing these three elements within a single system, Cubelets achieve breakthroughs that aren’t possible in legacy architectures, giving us a clear edge in scaling AI systems. Memory performance improves because power and cooling are managed at the chip and module level. Power efficiency increases because heat is dissipated intelligently. Thermal balance is maintained because compute, memory, and power delivery are treated as a unified whole.
This requires a fundamental change in architecture, one that brings the right technologies together in the right way. Our success comes from applying this co-optimization framework across every layer of the stack, enabling higher density, efficiency, and scalability, while remaining fully compatible with existing AI and cloud software environments.
What keeps your customers up at night?
Our customers are grappling with the realities of scaling infrastructure in the AI era, and their concerns can be summarized into three key areas: power, cost, and complexity.
Power – AI workloads are expanding so quickly that operators are running into challenges with electricity and cooling capacity. Without new approaches, operators risk hitting unscalable ceilings. Even with major investment, the physical and environmental limits of current data centers make it increasingly difficult to scale.
Cost – Building large-scale systems requires billions in capital for servers, power, cooling, networking, and real estate. Customers fear that these investments may not keep pace with AI’s rapidly changing demands, leaving them with stranded assets that are costly to operate and unable to deliver the performance they need.
Complexity – Coordinating compute, memory, power, and thermal management across tens of thousands of racks is a daunting engineering challenge. This slows development cycles, increases operational risk, and leaves customers feeling that AI innovation is outpacing their ability to adapt.
Mueon removes these limits; our Cubelet architecture integrates compute, memory, power, and thermal management into a single modular system, lowering costs, simplifying operations, and enabling AI infrastructure to scale efficiently.
What does the competitive landscape look like and how do you differentiate?
Much of the industry is still focused on extracting marginal gains from the same rack-based paradigm that has defined data centers for decades. Traditional OEMs and hyper-scalers push for incremental improvements in chip performance, cooling, or energy use, but they’re constrained by the physical and architectural limits of the rack. Some companies explore new cooling methods or form factors, but most solve for one layer of the problem rather than redefining the full system.
At Mueon, we see this as a moment to move beyond incremental improvements to redefine how entire systems are built. Moore’s Law has driven extraordinary advances at the silicon level, but it hasn’t been matched by equivalent innovation in how systems are built and scaled. We believe the next leap in computing performance will come from rethinking those abstractions – from chip to system to data center.
Our Cubelets architecture embodies that leap. By integrating compute, memory, power delivery, and thermal management into a single modular unit, we eliminate the artificial boundaries that slow innovation. This architecture delivers order-of-magnitude gains in density, deployment speed, and energy efficiency, while remaining compatible with today’s AI and cloud software stacks. Mueon is leading the next wave of abstraction, one of the first to deliver a fundamentally new model for building data centers in the AI age.
What new features/technology are you working on?
Our goal is to remove the physical limits that have constrained how silicon systems are built and scaled; there should be no limit to how large or how complex a chip can be. We’re developing technology that allows silicon to be built, stacked, and interconnected at unprecedented scale – whether that means going smaller or larger.
Scaling Down – Pushing toward the smallest possible dimensions for compute, memory, and interconnects to maximize efficiency and density.
Scaling Up – Enabling arbitrarily large chips and multi-layered stacking architectures that operate as a single coherent system.
Together, these dimensions unlock new possibilities for performance, efficiency, and system design. By breaking free from traditional limits on packaging and processing, Mueon is creating a foundation where compute can expand organically – without the bottlenecks that defined the last generation of data center architecture. We’re not just enabling faster chips, we’re creating the foundation for entirely new classes of computing.
How do customers normally engage with your company?
Right now, most of our engagements are with leading AI companies and hyperscalers that are actively building the next generation of AI chips and data infrastructure. These organizations are deeply aligned with our mission to remove the limits on how silicon can be designed, scaled, and deployed. We’re working closely with them to co-develop systems that push the boundaries of performance and efficiency.
We welcome conversations with anyone tackling similar challenges or exploring new chip models. Collaboration is core to how we operate, whether you’re building advanced AI systems, experimenting with chip architectures, or simply have ideas about where silicon design should go next.
Also Read:
CEO Interview with Alex Demkov of La Luce Cristallina
CEO Interview with Dr. Bernie Malouin Founder of JetCool and VP of Flex Liquid Cooling
