Kevin Pathrath an other warn of an AI Bubble and AI Ponzi with a potential 80% collapse. The core of the argument is the circular financing structure of the recent OpenAI-Oracle-Nvidia deal and the OpenAI-AMD deal.
Nvidia commits up to $100B in investments ($10B per gigawatt deployed, up to 10GW), starting with the first GW in H2 2026 on its Vera Rubin platform. This cash funds OpenAI’s infrastructure. Each Gigawatt purchased by OpenAI would be about $30 billion worth of Nvidia chips. Nvidia would be selling at 60-75% margin. $10 billion would be only part of $15-20 billion of profits from each sale. The $10 billion would be purchased at the mid-2026 valuation for OpenAI.
Is buying shares in a customer company a ponzi scheme? Profits are still made by Nvidia. IF OpenAI failed there would be no domino effect on Nvidia.
The AMD deal is selling 6GW of chips to OpenAI. OpenAI pledges to deploy 6GW of AMD Instinct GPUs (starting 1GW in H2 2026) which should be about $15-20 billion per gigawatt in AMD revenue. AMD. has about 38% margin. AMD issues warrants for up to 160M shares (~10% stake) at $0.01 each, vesting on AMD stock price targets (up to $600/share, ~3x current levels) and OpenAI’s technical/commercial milestones. This would be about $38 billion of value today. AMD is refunding the profits from the OpenAI chip purchases in AMD stock.
Is a 30% discount a ponzi scheme?
If OpenAI fails with its data center that might impact future AMD sales. AMD gets a volume increase.
AMD and Nvidia are all still making money on these deals.
This creates a self-reinforcing loop: Nvidia’s investment → OpenAI cash → AMD chip buys and deployments → AMD revenue/stock rise → OpenAI’s stake value increases → Nvidia (via its OpenAI ownership) benefits → More Nvidia investments → Further AMD expansions (lagging ~6 months behind Nvidia). Both deals sync on H2 2026 timelines, with AMD’s weaker financials (39.8% gross margin, 11.3% net margin vs. Nvidia’s 72% gross).
Kevin projects an $800B revenue shortfall by 2030 ($1.2T expected vs. $2T needed for AI buildout). If the revenue was “only” $1.2 trillion over the next 4-5 years then it would still reach $2 trillion about 18-24 months later.
Kevin fears Fed bailouts, rate cuts, and AI-induced unemployment mirroring 1929’s slow bleed.
Analysis of AI Ponzi Scheme Claims
Kevin’s narrative paints the AI sector—exemplified by the Nvidia-OpenAI-AMD deals—as a giant circle that is similar to a Ponzi scheme, where investments recycle among insiders without productive output, propped by FOMO, debt, and stock-tied incentives. However, a true Ponzi requires no genuine positive returns. Ponzi is only sustained via new inflows to pay early participants. Here the shares involved are a fraction of profits in cash involved in the related transactions.
These investments fund tangible infrastructure (data centers, chips) for AI deployment, with returns hinging on AI inference success. There is real-time AI usage from things like ChatGPT queries and they are generating scalable revenue.
OpenAI-AMD – This coopetition reduces single-supplier risk but ties payouts to speculative milestones (AMD tripling). Yet, unlike a Ponzi this generates real economic activity: Chip sales and data center builds and AI services. The risk is overbuild if demand softens—Kevin’s $800B shortfall aligns with broader forecasts needing $2T total AI spend by 2030 vs. ~$1.2T revenue but positive returns from inference could break the cycle.
If inference scales, this is venture-like bootstrapping.
The global AI inference market is projected at $106B in 2025 and could go to $255B by 2030 (CAGR 19%) or even more.
For OpenAI, inference powers 80%+ of ChatGPT revenue with H1 2025 sales at $4.3B (16% YoY growth)
Inference is ~10x cheaper than training per query, enabling 70-80% gross margins (Nvidia’s current AI revenue benchmark). Google Cloud reports AI inference yields=ing 20-50% ROI via personalized recommendations, cutting customer churn 15-30%.
IBM reports average 5.9% ROI across AI pilots, but 20-40% in inference-heavy uses like predictive maintenance (reducing downtime 25%).
Marketing/Sales: GrowthLoop cites 15-25% uplift in conversion rates from AI personalization, with ROI over 200% in 6-12 months
Kevin’s fear of a revenue miss ignores inference’s scalability: As models commoditize, take rates (2-5% on agentic commerce) could yield $200B+ for OpenAI alone by 2030.
AMD’s lower margins highlight dependency on volume deals.
Nvidia’s fortress balance sheet funds the circle without distress.
Actual Creation of Value Using AI. Especially AI inference, demonstrably creates value beyond hype, countering Ponzi fears
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
The main similarity of the AI boom to a Ponzi scheme or Tesla to a meme stock is to analysts who aren’t capable of thinking about the underlying technologies, looking at what companies are doing with them and coming to their own conclusions. To them, it’s all just a word salad explanation and bulls are in a cult. They can’t tell the difference between Elon Musk and Trevor Milton and they don’t believe anyone else can either.
Someone always profits from a bubble bursting.