When Will the AI Perpetual Motion Machine Grind to a Stop?
Imagine a world where artificial intelligence giants are spinning a financial whirlwind so powerful, it threatens to redefine the very rules of tech economics—until, inevitably, reality crashes in. That's the thrilling yet unsettling saga of OpenAI's ambitious maneuvers, where partnerships and valuations are skyrocketing, but questions linger about how long this high-wire act can last. Stick around, because this isn't just about tech deals; it's a story of innovation, excess, and the ticking clock of sustainability.
Author | Song Wanxin
Editor | Zhang Fan
Over the last couple of months, OpenAI has been buzzing with activity in the world of capital and resources. On one side, they've secured a massive amount of computational muscle, propelling their company valuation to an astonishing $500 billion and cementing their status as the planet's most valuable startup. On the flip side, by harnessing their cutting-edge technology and market dominance, they've crafted a tight-knit ecosystem that circulates funds within the upper echelons of the U.S. AI industry.
Building on their monumental partnerships with Oracle (valued at $300 billion) and NVIDIA (at $100 billion), OpenAI made waves on October 6th by unveiling a collaboration with AMD. This deal includes a multi-generational agreement for supplying 6 gigawatts of AMD Instinct GPUs, marking one of the biggest single AI chip deals in history. To put this in perspective, think of GPUs as the engines powering AI models—like having a fleet of supercharged sports cars dedicated to training algorithms. The initial rollout involves 1 gigawatt of AMD's Instinct MI450 series GPUs, set to deploy starting in the latter half of 2026.
But here's where it gets controversial... What really stands out is the intricate financial setup. As part of the pact, AMD has issued warrants allowing OpenAI to potentially acquire up to 160 million AMD shares, which translates to about 10% of AMD's total current shares. For newcomers to this, warrants are like options contracts that give the right (but not the obligation) to buy stock at a set price later. However, OpenAI must hit specific milestones to unlock these shares, granted in phases. These include technological rollout progress and AMD's stock performance, with a notable benchmark: AMD's shares need to hit $600, up from their current $215.
This kind of looped investment sparks intense debate. On one extreme, enthusiasts argue OpenAI is boldly betting big on AI's future, committing enormous resources upfront to computing costs because they're utterly confident in its potential. Think of it as investing in a groundbreaking factory long before the products hit shelves. On the other pole, skeptics whisper about an "artificial puffing up of the U.S. AI stock bubble"—where valuations soar without real cash flowing in, potentially inflating like a balloon ready to burst.
The AI Perpetual Motion Machine?
Right now, OpenAI is juggling multiple challenges. Their deep-rooted alliance with Microsoft is showing cracks, as OpenAI has pushed for negotiations to slash costs and diversify computing sourcing beyond Microsoft's grip. This aims to loosen Microsoft's hold on their AI offerings and infrastructure.
Yet, stepping away from Microsoft has ignited OpenAI's own version of a "perpetual motion machine" for AI computing, transforming equity into computational power in a self-sustaining cycle.
Dive into the details: NVIDIA has poured in up to $100 billion strategically into OpenAI. In return, OpenAI must exclusively use NVIDIA's GPUs for the next decade, consuming their production output and constructing a 10-gigawatt data center. For those unfamiliar, a data center is like a massive warehouse filled with servers that crunch data for AI—imagine it as the beating heart of AI training.
And this is the part most people miss... Crucially, much like AMD's staggered share grants, NVIDIA's funding ties directly to OpenAI's data center build-out. It unfolds via "option contracts," where for every gigawatt of power deployed (roughly 400,000 to 500,000 GPUs), NVIDIA injects $10 billion, building up to the full 10-gigawatt goal.
For OpenAI, this means more computing capacity boosts their valuation, which in turn unlocks even more computing power—a virtuous (or vicious?) loop of "equity ↔ computing power." This setup cleverly aligns financial flows with real progress, sidestepping the risk of OpenAI's hefty $5 billion annual losses from cash flow woes.
Now, let's shift to Oracle. In June, OpenAI inked a $300 billion computing services deal with them, spanning about five years. Starting in 2027, OpenAI will shell out roughly $60 billion yearly to Oracle.
To fulfill this, Oracle plans to buy 4-5 million GPUs from NVIDIA, costing around $100 billion. This weaves a fascinating capital loop: "NVIDIA → OpenAI → Oracle → NVIDIA."
The "AI Perpetual Motion Machine"
Numbers don't lie—Oracle's projected cloud revenue for fiscal 2026 stands at $18 billion. With the OpenAI deal, it's forecasted to leap to $144 billion by 2028, a jaw-dropping 700% surge. This directly competes with Microsoft's Azure and Amazon's AWS in the AI cloud space. Meanwhile, NVIDIA locks in steady orders, and OpenAI gains the funds to fuel their computing needs.
Superficially, OpenAI, NVIDIA, and Oracle are shelling out hundreds of billions, but they're leveraging "future earnings discounting" for capital efficiency. For instance, Oracle's $300 billion order is basically a long-term pay-as-you-proceed plan, where GPU purchases from NVIDIA are offset by upcoming fees from OpenAI. NVIDIA's $100 billion stake is secured by OpenAI's future buying commitments.
Beyond financial ties, they've forged a software monopoly. NVIDIA's CUDA platform and TensorRT engine integrate seamlessly with OpenAI's model training, and Oracle Cloud comes loaded with NVIDIA's AI tools for plug-and-play ease. This "lock-in" creates a barrier: Switching chips could cost OpenAI hundreds of billions in migration expenses. It's like being locked into one phone ecosystem—convenient, but switching to another could be a nightmare.
How to Wrap Your Head Around a $500 Billion Valuation
In the short term, markets have applauded the "AI perpetual motion machine." Post-OpenAI announcements, AMD's stock jumped 43% over four trading days, and NVIDIA's surged over 4% on their deal reveal day. OpenAI's valuation for its profit-making arm rocketed from $260 billion early this year to nearly $500 billion.
But here's the kicker—markets driven purely by hype, without solid cash flow, often sober up. The engine assumes growing computing power will fuel commercialization hopes, hiking OpenAI's worth.
Financially speaking, OpenAI raked in $4.3 billion in the first half of 2025, but losses ballooned from $3.1 billion last year to $13.5 billion. Their deals this year could total $1 trillion in costs, with CEO Jensen Huang estimating $50-60 billion per gigawatt for AI data centers.
On expenses, R&D at $6.7 billion dominates, but marketing leapt to $2 billion this year—double last year's—reflecting intense competition in real-world applications.
Operationally, while users flock to OpenAI, revenue trails behind spending.
Changes in ChatGPT's MAU and revenue from various segments
Tracing ChatGPT's rise: From November 2022 to December 2023, monthly active users (MAU) hit over 200 million. In 2024, they exploded from 320 million to 650 million, a 103.88% annual jump. This year, Q1 saw 35.61% growth, nearing 900 million MAU.
Monetization streams: Individual subscriptions, enterprise deals, and API services.
Individually, subscriptions lead revenue—52% in January last year, up to 68.5% in April this year, with average revenue per user climbing from $0.703 to $0.792.
Note this: Last December, subscriptions grew 251.7% year-over-year, outpacing MAU's 103.9%. But in Q1 this year, it slowed to 35%, matching MAU, signaling peaking pay-user growth.
Enterprise subscriptions dipped to 21% amid API boom. Last year, APIs slumped to $42 million annually but rebounded 58.3% in Q1 to $66 million, or 9.3% of revenue.
Though small, APIs are key for long-term resilience, embedding AI into niche markets—unlike consumer-focused business.
OpenAI's monetization worries are evident. Can this perpetual motion machine truly endure?
Disclaimer:
The views in this piece are solely the author's. Markets carry risks—invest cautiously. Nothing here offers investment advice; consult experts before decisions. We don't provide underwriting or licensed services.
Follow for more insights.
What do you think? Is this AI perpetual motion machine a genius stroke of efficiency, or a risky illusion destined to collapse under its own weight? Do you agree that these partnerships are inflating a bubble, or is OpenAI genuinely pioneering the future? Share your takes in the comments—let's debate!
