The New Gold Standard: How NVIDIA Turned GPUs Into Currency

Let me drop a name: Nscale. Have you heard of it? Most likely not. Yet this September, the company raised $1.1 billion – the largest Series B in European history – followed by another $433 million through a pre-Series C SAFE, reaching a $3 billion valuation in less than two years. How did they do it, I wondered. What’s their story?

Nscale sits at the junction of capital, energy, and geopolitics. Its rise is compelling on its own – but far more revealing as part of a larger pattern.

Welcome back to our AI Infrastructure Unicorns series. This one is unusual. Through Nscale, we’ll explore how the AI industry is maturing and how NVIDIA has learned to mint unicorns at will – financing its own customers, seeding sovereign compute hubs, and turning supply chains into strategy. A system where data centers become national assets and GPUs behave like currency.

It’s the story of an AI boom transforming into an industrial system.

In today’s episode:

With the Unicorn series, we’ve traveled to China and explored Silicon Valley. This episode takes us to Norway, where hydroelectric power hums through the racks of a company that barely existed two years ago and is now valued at crazy $3 billion. And if you think it’s a Norwegian company – you’re wrong. Nscale – this wunderkind startup – is a national pride of the UK. It used to be a crypto miner; now it’s all-in on AI. Let’s untangle.

In many ways, Nscale follows the path that CoreWeave pioneered. CoreWeave, a New Jersey-based company that started as an Ethereum mining collective in 2017, became the first great pivot story of the AI infrastructure age (CoreWeave: The New Gold Standard in AI Infrastructure). When in 2022 Ethereum moved to proof-of-stake, CoreWeave repurposed its GPU mining rigs to power machine learning workloads, grew into a hyperscale cloud provider, and by March 2025 went public at a valuation around $18 billion (as of market close on November 7, 2025, its market capitalization stands at about $68.32 billion.) It was the first proof that crypto’s stranded compute could be alchemized into the fuel of the AI boom.

Nscale shares that lineage. The company was spun out of Arkon Energy, an Australian Bitcoin mining operation with a simple strategy: chase cheap, renewable power wherever it exists. Founded by Josh Payne, Arkon had acquired and built facilities in Glomfjord, Norway (hydroelectric surplus), Ohio, USA (stranded energy), while keeping corporate headquarters in Sydney. The geographic sprawl wasn't elegant, but it was profitable – at least while crypto was.

When the crypto market collapsed in 2023, most miners liquidated their GPU inventories and shut down. Payne saw something different: the perfect infrastructure for AI was already built. The same requirements that made a good crypto mine – cheap electrons, cold climates for cooling, fiber connectivity – were exactly what AI compute clusters needed. The hardware was there. The power contracts were signed. The only thing that needed to change was the workload.

At the end of 2023, Arkon raised $110 million and executed the pivot with surgical precision. First, in May 2024, Payne incorporates a new entity, Nscale, in London to tap European AI funding and government support. Then in July, Nscale acquires Kontena, a Finnish firm that built modular data-center pods – prefabricated "AI rooms" that could be shipped and assembled within weeks.

From there, they moved fast. Nscale first hydropower-fed campus in Glomfjord, Norway, became the prototype for low-carbon AI infrastructure. Soon after, a 50/50 joint venture with Aker ASA and OpenAI – Stargate Norway – was launched to deploy 100,000 NVIDIA GPUs by 2026.

By September 2025, Nscale turned to the UK, announcing a partnership with Microsoft, NVIDIA, and OpenAI as part of the UK-US Technology Partnership. The plan: to deliver up to 58,640 NVIDIA GB300 GPUs nationwide, within a global build-out of 300,000 units.

At its center stands the Loughton AI Campus, a 50 MW facility (scalable to 90 MW) set to become the UK’s largest AI supercomputer in 2027, powered by 23,040 GB300 GPUs for Microsoft Azure. In parallel, Stargate UK will bring OpenAI’s compute to sovereign UK sites, starting with Cobalt Park and an initial 8,000 GPU offtake in 2026.

A month later, Nscale and Microsoft expanded their deal to 200,000 GB300 GPUs across Europe and the U.S., including hyperscale sites in Texas and Portugal – turning Nscale’s footprint into a distributed, energy-aware GPU network spanning continents.

The speed created an identity crisis that was almost comic. Here was a company with Australian roots, Norwegian power, and British incorporation being hailed as the UK’s “national champion” for AI – even though its largest site sat hundreds of miles above the Arctic Circle. But the confusion masked something more important: in less than five months, Nscale had turned crypto’s stranded assets into critical AI infrastructure. At least in press-releases.

That story alone would make Nscale an interesting pivot case. But its real significance appears when you zoom out. What connects companies like Nscale and CoreWeave isn’t geography or governance – it’s GPUs, and the gravitational pull of NVIDIA. Let’s see how these unicorns fit inside NVIDIA’s monetary machine →

Every NVIDIA-backed infrastructure company operates inside a small constellation of industrial partners. In Nscale's case, that constellation includes Aker ASA, Dell Technologies, Nokia, Microsoft, OpenAI, and NVIDIA itself. Each partner fills a structural role that links physical energy to digital capacity.

Each partnership is both commercial and strategic. Together, they form a reproducible model that NVIDIA can replicate in other geographies: industrial partner for power, hardware partner for integration, network partner for connectivity, and AI tenant for utilization.

That's right – NVIDIA is everywhere. This is the real connective tissue between companies like Nscale and CoreWeave: both exist because of NVIDIA's hardware, and more importantly, because of its patronage. The company that once sold GPUs now engineers entire economies around them, building a self-reinforcing system where capital, compute, and demand feed each other in a perfect loop.

Look closely, and NVIDIA's network resembles a financial order more than a supply chain. Let me explain →

After World War II, the Bretton Woods system turned the dollar into the anchor of global finance. Every major currency was pegged to it, and the United States supplied liquidity through gold and industrial output. It was more than a monetary agreement — it was an architecture of power built on convertibility and trust.

NVIDIA’s network follows a similar pattern, only its reserves are made of silicon. GPUs function as the reserve currency. Sovereign compute hubs like CoreWeave, Nscale, and Together AI operate as member states, each holding their share of NVIDIA’s chips. CUDA provides convertibility. NVIDIA’s capital deployments act as liquidity injections across the ecosystem.

As the dollar once turned capital flows into geopolitical leverage, this new order channels compute flows into industrial dominance.
CoreWeave serves as the American central bank of this system. Nscale anchors the European branch. OpenAI is the primary borrower. Aker supplies the energy reserves. And Jensen Huang, both architect and issuer, stands at the center like a chairman of the Federal Reserve of AI.

NVIDIA now performs the role of a monetary authority for machine intelligence. Every model, every inference, every sovereign AI project ultimately clears through its chips.

Is it a bubble? We will discuss it later.

We can see how NVIDIA's model is more than corporate strategy – it starts to look like industrial policy. Each investment plants a flag in a new region, backed by local energy, political alignment, and sovereign branding. NVIDIA brings the chips; partners like Aker, Microsoft, and Dell bring land, power, and customers; governments bring legitimacy. Together they build AI factories – vertically integrated systems where demand for NVIDIA hardware is designed, financed, and secured years in advance.

It’s a closed monetary circuit that turns CapEx into currency:

Capital becomes compute, and compute becomes capital – an economic loop as precise as lithography.

The structure also redefines dependency. In the old order, chipmakers relied on hyperscalers like Amazon and Google. In the new one, NVIDIA cultivates its own aligned ecosystem – CoreWeave in the U.S., Nscale in Europe, Crusoe in energy regions, Lambda for developers – each small enough to stay loyal, yet large enough to multiply reach.

While analysts debate whether this counts as circular finance or synthetic demand, NVIDIA’s intent is clear. The company no longer sells hardware; it builds markets. It designs the scaffolding for the next boom and occupies every layer: supplier, creditor, and beneficiary.

Nscale’s rise fits neatly into this pattern. As NVIDIA expands through projects like Nscale, its capital fuses with government planning, turning AI infrastructure from a private buildout into a tool of national strategy.

Across Europe, regulators and energy ministries are realizing that compute capacity is the new critical infrastructure – as strategic as oil reserves once were. The UK government has embraced Nscale as a domestic alternative to US hyperscalers. Prime Minister Keir Starmer's tech plan calls for the UK to be "an AI maker, not a taker."

The definition of 'sovereign compute' used by European customers and ministries is practical: run workloads under local jurisdiction, keep data handling auditable, source energy transparently, and ensure predictable access. Nscale's configuration maps to that checklist without being state-owned.

For NVIDIA, this shift is perfect. Every sovereign cloud that runs on NVIDIA chips deepens the company's moat while satisfying national demands for autonomy. It's corporate expansion camouflaged as technological independence – turning AI infrastructure from a private industrial effort into a component of national planning.

We wouldn’t be us if we didn’t look into history. NVIDIA’s model borrows from several industrial archetypes – each familiar on its own, but unprecedented in combination.

From these patterns, NVIDIA built something qualitatively new: No previous company has operated as vendor, financier, software platform, and geopolitical actor at once. It controls the bottleneck input (advanced GPUs), finances the buyers, embeds proprietary libraries across the open-source stack, and aligns with governments through sovereign AI and industrial programs.

Nscale is one of the most precise bolts in that machinery – a regional node that reveals how NVIDIA’s influence moves through capital, code, and policy in unison.

NVIDIA's system is a remarkable feat of industrial coordination, but like all tightly coupled networks, it has points of fragility that no amount of capital can eliminate.

For now, the system holds. The financial flywheel turns, the chips ship, and the power bills get paid. But the coherence of the network depends on a perfect alignment of supply, energy, and policy that may not last forever.

I’ve read so many articles comparing the AI boom to the dot-com bubble that I want to use the discussion around the NVIDIA flywheel to explain why the analogy doesn’t hold – at least not in its usual form.

The resemblance is superficial. The internet of the late 1990s was new, barely used, and largely speculative. Most companies building websites back then were constructing possibilities, not infrastructure. AI, by contrast, has been in commercial use for years. It is already embedded in logistics, healthcare, finance, and manufacturing. More important, its heaviest applications are in fields that consume vast amounts of compute and energy – drug discovery, automation, and robotics.

That difference matters. The Internet bubble built speculative demand for a technology people had yet to use. The AI boom is scaling an existing one. The risk today is not that demand evaporates overnight, but that infrastructure struggles to meet it.

Even if the current boom proves unsustainable in its present form, the overbuild may still leave behind lasting assets. The railways of the nineteenth century, the fiber networks of the early 2000s, and the cloud campuses of the 2010s all began as overextensions.

If this is a bubble, it is at least a functional one. The money flooding into AI infrastructure is not only buying chips; it is underwriting power generation, grid reinforcement, cooling technology, and high-speed networking. These are long-lived assets. GPUs may depreciate within five years, but substations and renewable installations last decades.

Nscale may look like a small startup, but inside NVIDIA's architecture it functions like a mint — a place where capital multiplies through hardware. Every rack of GPUs it installs turns into a new cycle of demand, another reason for investors and governments to pour in more money.

That's what makes Nscale more than a unicorn. It's a yield node in NVIDIA's global network — an instrument that converts venture capital, energy, and national ambition into predictable revenue for the chipmaker at the center.

This is the pattern we've been tracing: the quiet rise of unicorn factories, each one a local growth story, each feeding the same central engine. What looks like diversification is, in reality, vertical integration — a self-reinforcing system that grows both compute and capital at once.

Whether the market calls it a bubble or a boom doesn't change the mechanism. The machine works. Every time a new player enters the game, the same circuit closes — and the mint keeps printing.

In the previous century, power was measured in oil barrels and kilowatt-hours. In this one, it is measured in compute. And at the center of it all, NVIDIA sits like a quiet reserve authority – not just selling chips, but issuing the currency in which all artificial intelligence must be denominated.

The mint is running. The currency is flowing. And somewhere in Norway, in a hydroelectric facility that once mined Bitcoin, Nscale is turning electrons into intelligence, proving that GPUs aren't just products anymore.

They're the reserve currency of the future