Nvidia Earnings: A Geopolitical Barometer
This isn't only about revenue - it's about the new 'AI Order'
This afternoon's Nvidia earnings call isn't just a market update; it's a geopolitical event. Profits are up almost 70% YoY to $44.1 bn, much of that driven by ‘data centre growth.’
Nvidia bulls feel vindicated. But this is not just about markets. Too often, earnings discussions obsess over quarter-on-quarter growth or AI ROI, missing the bigger picture: the snowballing, all-encompassing transformation driven by commoditized, ever-improving intelligence. That is what is happening now with AI.
Compute, infrastructure, and energy form this foundation, a truth no one understands better than Jensen Huang.
So, beyond the numbers — here are five takeaways.
1. The Rise of the AI Factory
Forget the notion that Nvidia makes only GPUs. Nvidia is an infrastructure powerhouse, building the ‘AI Factories’ of the future. This term, coined by Jensen Huang, isn't metaphorical: AI Factories are not the data centers of old. They are specialized facilities purpose-built to mass-produce intelligence, making them critical to national and economic power.
While many players invest heavily in custom AI chips, NVIDIA’s true, comprehensive strength lies in its full-stack platform. This means it controls virtually every layer necessary to build and run AI Factories. It’s vital to recognize Nvidia sells not just a single component, but an entire, tightly integrated ecosystem. This positions it to capture a disproportionately larger share of the growing AI pie, regardless of new chip entrants, due to core components other chipmakers simply can’t offer, including:
Software Dominance (CUDA): This is Nvidia's secret weapon. CUDA is their proprietary software framework, the operating system for AI on their hardware. With decades of development and a massive, entrenched community of millions of developers, CUDA creates an almost insurmountable barrier to entry. It's why developers choose NVIDIA, effectively locking them into its ecosystem.
Integrated Systems: Nvidia doesn’t just ship individual chips. They design and sell complete DGX and HGX platforms—"AI-in-a-box" supercomputers. These fully engineered systems combine powerful GPUs, specialized central processors, and incredibly fast networking, ensuring seamless performance and simplified deployment for customers, from small companies to national AI initiatives.
Networking Leadership: High-speed networking is crucial for AI performance. Through its Mellanox acquisition, Nvidia provides high-performance InfiniBand and Spectrum-X Ethernet solutions. These are the ‘digital highways’ connecting thousands of GPUs, allowing them to communicate and act as a single, cohesive supercomputer. Without this lightning-fast data flow, individual GPUs would be isolated and ineffective.
This full-stack approach means NVIDIA sells entire solutions, not just components. It ensures their continued market dominance, making true replication not just difficult, but a multi-decade, multi-trillion-dollar challenge for new entrants. Crucially, this makes NVIDIA an indispensable strategic partner in the global AI race. Controlling these AI Factories means controlling the very means of intelligence production, establishing a new form of national power and digital sovereignty for those who gain access to this critical foundation.
Indeed, the bulk of Nvidia’s growth this quarter lies in that ‘AI Factory’ expansion: $39.1 bn to be exact.
2. Beyond the H20s: Geopolitical Chess in Motion
This isn't Jensen Huang's first export control rodeo. Nvidia has navigated chip restrictions since Trump 1.0 (2018), when the U.S. first recognized computational power as a battleground in its supremacy contest with China.
The latest blow: Nvidia’s China-specific H20 chip, designed to comply with earlier restrictions, was banned in April 2025. The result was a $5.5 billion inventory write-down in Q1 FY2026 and a staggering $15 billion in lost potential sales. Nvidia’s market share in China collapsed from 95% to around 50%, as Chinese firms pivoted to domestic or legacy chips.
Huang views these restrictions as self-defeating, spurring Chinese innovation and shutting out U.S. infrastructure from the world's largest developer base. As he put it:
The idea of AI diffusion limiting other countries' access to American technology is a mission expressed exactly wrong. It should be about accelerating the adoption of American technology everywhere before it's too late…The idea that we would have America not compete in the Chinese market, where 50% of the developers are, makes absolutely no sense."
-Jensen Huang, Stratechery, May 2025
But geopolitics is never static. As one hand restricts, another builds alliances. Since Trump—‘ The AI President’— took office, Big Tech has stepped into digital diplomacy. Jensen Huang is no longer just a CEO; he's a soft power emissary, signing billion-dollar AI infrastructure deals that align with U.S. strategic interests.
Just look at the past month:
Saudi Arabia: During Trump’s Gulf visit, $600 billion in deals were inked. Nvidia will supply thousands of GB300 Grace Blackwell supercomputers to build sovereign AI infrastructure, with compute costs alone reaching billions.
UAE: Altman’s $500 billion Stargate project (OpenAI + Oracle + SoftBank) will house over 2 million Nvidia GB200 chips in Abu Dhabi. At $70,000 per chip, this represents over $140 billion in silicon for a single 5-gigawatt campus.
Europe: Nvidia is building Europe’s largest AI factory near Paris with MGX (Abu Dhabi): a 1.4-gigawatt, $9.6 billion facility operational by 2028, and part of a €50 billion UAE investment in French AI infrastructure.
Others: Major partnerships are also underway in Qatar, Sweden, and with every major U.S. hyperscaler.
The pattern is clear: China gets locked out, allies get locked in. Whether or not this is effective long term remains to be seen. Today's earnings, therefore, will reveal how effectively Nvidia executes this geopolitical mandate—becoming not just a tech company, but a vital instrument of American strategic power.
3. The Scaling Wars: When Size Becomes Everything
These AI Factories embody the brutal logic of the AI Scaling Law: more compute, data and algorithms equal exponentially better AI. This simple equation drives a relentless, multi-billion-dollar arms race for computational supremacy.
Exhibit A: Elon's Colossus. When Musk launched xAI, he bypassed traditional data center providers entirely, instead building Colossus—going from zero to 200,000 H100 GPUs in just 122 days near Memphis. Jensen Huang marveled:
"As far as I know, there's only one person in the world who could do that; Elon is singular in his understanding of engineering and construction and large systems."
The math is staggering: at $30,000-$40,000 per H100, that's a $6-8 billion compute cluster. Musk's next target? Million+ GPUs at a potential cost of $25-40 billion.
Exhibit B: Altman's Stargate. Where Elon leads, his ‘Frankenstein’ and greatest nemesis, Sam Altman follows. The $500 billion OpenAI-Oracle-SoftBank AI Infrastructure ‘Stargate’ venture is well into building its first site in Abilene, Texas. (Why West Texas? Natural gas access, favorable regulations, and endless space.) This facility will house $40 billion in NVIDIA GB200 "superchips," delivering 1.2 gigawatts—potentially scaling to 5GW. Of course, for Altman, the goal is the same as Elon: who can scale AI faster?
Underpinning this is the hyperscaler backdrop (they want to do the same!). Meta, Microsoft, and Google collectively spend tens of billions annually on NVIDIA infrastructure. Meta alone targeted 350,000 H100s by end-2024. Microsoft continuously expands Azure's AI backbone, and Google Cloud offers NVIDIA platforms to enterprise customers worldwide.
The takeaway: AI supremacy is about innovation — but it's also about who can afford NVIDIA's foundational compute at scale. The largest cost for AI Factories? The compute itself. NVIDIA's full-stack advantage—chips plus networking plus software—ensures their continued dominance in this scaling war.While efficiency improvements will matter, betting against compute is betting against the fundamental physics of AI progress.
4. The Energy Bottleneck: Power as the New Chokepoint
Here's the uncomfortable truth: AI's appetite for electricity is reshaping global energy policy. A single 5-gigawatt AI Factory consumes as much power as a small country like Singapore. We're witnessing a geopolitical "energy arms race" where access to cheap, reliable, high-density power determines AI leadership.
China, with its massive infrastructure investment, industrial base, and projected electricity generation capability, will be able to handle this transition with far more ease than any other player. For the U.S. and Europe, however, it's an Achilles' heel (especially for Europe).
The U.S. response has been swift and dramatic. As soon as Trump came to power, his administration declared a National Energy Emergency (Executive Order 14156), directing federal agencies to use emergency authorities to quickly boost domestic energy production and infrastructure, explicitly citing high demand from new data centers.
Last week, a new Executive Order launched a nuclear revival, targeting:
Regulatory reform: Streamlining Nuclear Regulatory Commission approvals
Industrial base: Rebuilding domestic nuclear fuel supply chains
Advanced reactors: Fast-tracking testing and deployment for national security infrastructure, explicitly including AI data centers.
Nuclear power—long stigmatized—is now considered essential for AI's continuous, high-density power requirements. The result has been massive VC investment flowing into nuclear startups, particularly Small Modular Reactors (SMRs) designed for on-site AI Factory power generation.
Jensen Huang captures the reality: "Every single data center in the future will be power limited. We are now a power limited industry." NVIDIA is responding by partnering on new architectures, including 800V HVDC systems crucial for AI Factory efficiency and scalability.
The geopolitical stakes are clear: China produces over 9,000 terawatt-hours annually versus America's 4,000. However, the U.S. is strategically focused on high-density, AI-optimized power sources. If compute is China's bottleneck due to chip restrictions, energy supply is becoming America's critical constraint in maintaining AI leadership.
The bottom line: Today's earnings aren't just financial metrics—they're a barometer of how rapidly AI is reshaping global power dynamics, one factory at a time.
5. New Alliances Being Forged: Redrawing the World Through AI Infrastructure
While China spent decades extending its influence through the physical sinews of the Belt and Road Initiative, the U.S. is now engineering its own, arguably more potent, digital counter-BRI through the strategic deployment of AI infrastructure by its tech titans. NVIDIA, unable to freely penetrate the Chinese market with its most advanced chips, has become a pivotal force in this new form of technological statecraft.
The multi-billion-dollar "sovereign AI" deals in the Gulf and Europe are not mere market expansions; they are explicit attempts to redraw the world's geopolitical map, consolidating technological spheres of influence. These partnerships are deeply strategic, forging new alliances and integrating nations into a U.S.-aligned AI ecosystem, a process that simultaneously boosts America's soft and hard power.
Nations seeking "sovereign AI" gain powerful capabilities that nonetheless often rely on foundational U.S. technology. This creates a shared technological destiny and a strategic reliance that reshapes geopolitical alignments. The ability to generate, train, and deploy advanced AI models will be a cornerstone of future geopolitical power, impacting everything from economic productivity and scientific discovery to military capabilities.
The earnings, therefore, are not just a financial report; they are a barometer of how rapidly AI is creating a more fractured, yet technologically aligned, global order.