At Edgescale, we’re engineering the infrastructure that brings artificial intelligence into the real world. Our work powers AI in the places that keep society running — manufacturing floors, hospitals, utilities, transportation networks, and more. By bridging the gap between the cloud and the physical edge, we enable real-time intelligence where humans and machines work together.
The United States is at a critical disadvantage in physical AI. The shortfall is not in models or research, where American ingenuity still leads, but in the unglamorous work of deploying AI into the physical world.
The reason is infrastructure: the networks that would carry intelligence into our factories, utilities, and grid have not been built at the scale the moment demands. Few have experienced this more than Marc Rouanne, one of our co-founders here at Edgescale AI, who has spent two decades on the front lines against Huawei’s dominance in networking.
Marc is, in many respects, America’s French soldier in this war – a Lafayette for the AI age, a Frenchman who has taken up America’s fight because he knows, better than most, exactly what the other side can build. He holds a PhD in information theory – the precursor to modern AI – from the University of Notre Dame, and served as President of Mobile Networks at Nokia and as Chairman of the Board of Alcatel-Lucent. As Nokia’s chief operating officer, he made it the first major vendor to join the Open RAN alliances – the movement to pry open the closed radio systems Huawei thrives in – and at DISH he went on to architect America’s first nationwide cloud-native Open RAN 5G network. He entered the United States on a visa reserved for individuals of “extraordinary ability.”
While leading the charge in cellular infrastructure and the mobile revolution a decade ago, Marc sponsored a quieter Nokia initiative called NDAC – the Nokia Digital Automation Cloud – to sell advanced private networks for manufacturing and utilities across the United States. In turn, this catalyzed the rise of other companies, such as Mavenir, set on overcoming a widening gap in industrial infrastructure. Over ten years, Nokia succeeded in deploying about 500 NDAC networks – enough to make Nokia the leading Western supplier of private industrial networks. China, meanwhile, now counts on the order of 64,000 private 5G networks, the great majority stood up in just the last few years, overwhelmingly on Chinese gear. As a deployment rate, that is not a thirty-fold advantage; it is a gap of two orders of magnitude – and it lays bare the difference between American and Chinese infrastructure foresight.
The pattern is the same one we see across our bridges, our roads, our manufacturing base, and our power grid: chronic underinvestment and slow decay layered atop infrastructure we once led the world in building. Communications and industrial equipment is outsourced to our European, Japanese and Korean allies, who struggle to maintain parity, while China’s vertically integrated, top-down mandates provide a distinct advantage. It is systemic digital rust and supply-chain frailty, spreading quietly across the very systems our economy runs on.
China’s advantage is not accidental; it is the product of deliberate, state-sponsored, systemic thinking. Where American companies treat networking as a cost center, China embeds modern infrastructure in every new facility, regardless of immediate profitability, weaving security, adaptability, and AI capability into the fabric of its critical infrastructure. We have not.
The consequence is decisive: where networking is virtualized — a modern, software-defined technique – AI becomes effectively “drop in.” China can install whatever software and AI it wishes across a plant, a utility, or an industrial corridor, and keep doing so for the next three decades. Modern networks beget advanced software and automated controls, which generate the returns that justify the investment – a virtuous cycle.
American operators, lacking that foundational layer, face enormous friction and often find it technically impossible to deploy the AI and automated controls that will define competitiveness. And the physics is unforgiving: a running plant is one of the densest data sources on earth – far too much, too fast, and too continuous to ship to a distant cloud for answers. The intelligence has to come to the plant, and only a modern network can carry it there.
History is unambiguous about why this matters: communications infrastructure precedes economic development. Fiber, cellular, broadband – each wave of connectivity was the precondition for the growth that followed, not its reward. For years, the absence of modern deployment infrastructure was a nuisance, a cost easy to defer. With AI, we have passed the tipping point: what was a nuisance is now an urgent gap and a structural weakness – the dividing line between economies that can absorb AI and those that cannot.
We cannot undo the past or rip out the infrastructure already in place. We have to be pragmatic. Edgescale’s answer is to dropship its Cubes – self-contained AI appliances that plug into existing environments and augment them in place, carrying intelligence into operations already running. The Cubes are sovereign by design: the plant’s data – and the intelligence learned from it – never leaves the plant. That matters, because operators are right to distrust software trained on someone else’s factory; what they trust is their own instruments. Intelligence that lives on the floor and reads the plant’s own gauges earns trust the way every instrument on the line always has. This lets America catch up without waiting a decade to rebuild, and take back control of its software and silicon supply chain in the process.
Yet catching up on deployment is only half the race – and the smaller half. Once AI is running a power plant or a factory, the question stops being whether the machine is clever enough and becomes whether you can trust it to act on its own – to change a setting, learn from new conditions, take on a new task – without anyone losing track of what it did or why. It is the same worry the world now has about advanced AI everywhere: a system too capable to keep a firm hand on.
China’s answer is the authoritarian one: command it from the top, and accept how brittle that makes it. The free world can do better, and the promise is simple to state – an AI that asks before it acts, keeps a person in the loop at every step, and always shows its work. Nothing it decides takes effect until a human signs off; everything it does can be traced back to a plain reason; and it can keep getting smarter without ever slipping the leash. That is how you keep industry advancing without anyone losing control of it – and it is exactly the edge a democracy is built to hold: not a machine that answers to the state, but one that answers to the people running it.
But private effort alone will not close a gap of this scale. That is why an accelerated, government-backed push to deploy advanced AI infrastructure across American manufacturing, utilities, and energy is not merely advantageous but essential. The vehicle already exists: the Department of Energy’s Energy Dominance Financing program has closed roughly seven billion dollars in thirty-year financing for Southern Company this year – the largest loan commitment in the Department’s history – alongside more than $700 million for DTE and $685 million for AEP Texas.
The lane now rebuilding the grid should carry the intelligence that will run on it: dedicate a defined share of that financing to AI-ready industrial infrastructure – the networks, the edge compute, the controls that keep people in charge. America does not need to out-build China’s last decade; it needs an industrial base that can change without losing coherence – augmenting what runs today, and keeping a human hand on everything it adds tomorrow. Without it, America cedes the industrial foundation on which all future AI will be built.