Wissner-Gross: Nvidia’s $96B Quarter Could Mask a ‘Financial Engineering’ Boom — BigGo Finance

Nvidia just posted a quarter so large it out-earns the GDP of most countries on Earth. $96.2 billion in revenue. Net income of $59.7 billion. A forward guidance of $108 billion. And yet, on a recent episode of the Moonshots podcast, the sharpest minds in the room were not popping champagne. They were asking a quietly terrifying question: what if some of this unprecedented demand is a mirage—an artifact of circular financing rather than genuine economic need?

“I’d be the last person to suggest that this is one big wash trade circular financing scheme,” said Alexander Wissner-Gross, a physicist and AI researcher, “but I do wonder the extent to which the financial markets have fully priced in Nvidia’s financing of their customers.” His concern cuts to the heart of the AI boom. Nvidia, the single most valuable company in the world by market heft, is not just selling chips. It is increasingly acting as a bank, deploying capital in checks ranging from $5 billion to $20 billion—dwarfing the rest of the economy, as host Peter Diamandis noted—partially to ensure its own customers can afford to buy its products.

The Circular Economy: A Bug or a Feature?

Wissner-Gross’s warning is stark. “I’d like to see no hiccups in this singularity,” he said. “I would not like to discover 2 to 3 years from now that some or a large portion of all of this demand at the infra layer was being artificially propped up through financial engineering.” It is the kind of systemic risk that keeps regulators up at night: a company whose revenue growth is so tightly coupled to its own investments that the line between vendor and creditor blurs.

But not everyone on the podcast agrees that circularity is a problem. Dave Blundin, an AI investor, offered a more provocative reframe. “Every economy is circular in nature. You know, dollars just move in a circle,” he argued. His analogy: if a civilization on Alpha Centauri achieved superintelligence, it would build a vast internal economy and “not give a rat’s ass about Earth.” The AI economy, in his view, is becoming that civilization—a closed loop that generates its own demand and cares little about the legacy economy. It is a seductive idea. It is also, according to Wissner-Gross, exactly the kind of thinking that could mask a bubble.

Even if the worst-case scenario materializes, Wissner-Gross is not predicting a crash. “If there is a correction, it will be a mini winter, not a full winter,” he said, because the underlying resource is too fundamental. “Compute is fundamentally substituting for real estate and human labor and other raw inputs as the fundamental substrate for human civilization.” In other words, even if the financing is frothy, the product is not a fad. Nvidia’s guidance supports this. The company told investors to expect roughly 70% sales growth in fiscal 2028—far above Wall Street’s 44% consensus—and management indicated that without supply constraints, growth would exceed 100%.

The TSMC Chokehold and the New Moat

For all of Nvidia’s dominance, the podcast identified a critical vulnerability that does not get enough attention: Taiwan Semiconductor Manufacturing Company (TSMC). Nvidia has no real manufacturing diversification. It is roughly one-third of TSMC’s output, an arrangement that makes vertical integration a delicate dance. Nvidia cannot afford to alienate its sole supplier, so any attempt to move up the value chain must be quiet. The company’s other supposed vulnerability—the erosion of its CUDA software moat—was debated at length.

Salim Ismail, a former Singularity University executive, was blunt about Nvidia’s strategic shift. “Nvidia is trying to move out of just being a chip company to trying to become literally an operating system for the whole of intelligence,” he said. The playbook, according to Ismail, could involve acquiring Hugging Face, the popular open-source AI model repository. He compared the potential move to Microsoft’s acquisition of GitHub—a purchase that bought developer loyalty as much as technology. By owning the layer where developers share models, Nvidia would cement its position as the platform through which AI is built, not just the company that sells the silicon.

Blundin went further, arguing the CUDA moat is already dead for inference—the process of running trained models—and has a limited lifespan for training. The new moat, he contends, is interconnect. Jensen Huang’s masterstroke was the acquisition of Mellanox, which gave Nvidia control over the high-speed networking needed to bind together enormous GPU clusters. Training next-generation models requires coherent clusters of 100,000 to several million GPUs. That scale is Nvidia-only territory. It is a hardware moat that companies like AMD and custom-silicon efforts from Google and Amazon cannot easily replicate.

OpenAI’s Jalapeno and the Coming Hyperscaler War

The competitive landscape is shifting fastest at the inference layer, where OpenAI’s new custom chip, Jalapeno, is making waves. Developed with Broadcom, the chip reportedly delivers 1.5 to 1.9 times more AI work per watt and up to 3.6 times lower latency than Nvidia’s GB200/GB300, while consuming half the power. The most startling figure cited on the podcast: a claimed 54x throughput increase compared to existing architectures when running both open-source and closed models.

Wissner-Gross sees a wild possibility emerging from this. “We’re in such a crazy future where I could imagine a world where say an OpenAI compute cloud hosts an Anthropic model and OpenAI and Anthropic win at the same time,” he said. It is a scenario that would have been unthinkable two years ago: OpenAI, the model lab, becoming a hyperscaler that serves its rivals. If Jalapeno’s performance claims hold, OpenAI would have the economic incentive to rent out its cheap, efficient inference capacity. The chip moved from idea to production in months—a testament to AI-assisted chip design—and it puts pressure on Nvidia’s 90%-of-inference market share.

Meanwhile, Blundin was scathing about Apple’s response to the AI moment. Commenting on the new Mac Studio with M5 Ultra and 512GB of unified memory, he said, “They don’t even deserve to be a Mag 7 company anymore… the best they can do in the age of AI is add a bunch more RAM to a machine that they already had.” Wissner-Gross traced Apple’s neural engine hardware lineage back to the cancelled Apple car project and argued the company “fumbled the ball” on AI software despite having superior hardware.

The American and Chinese Paths to Superintelligence

The most analytically rich portion of the discussion concerned the divergence between American and Chinese AI strategies. Chinese firms like Alibaba are giving away model weights for free and pouring resources into video generation and world models. Alibaba’s Wan 3.0 can generate 30-second videos from documents and spreadsheets for a few cents per second. According to one report cited on the podcast, video generation now represents roughly 70% of all AI token consumption in China—driven by short-form content and robotics—growing faster than code generation grew in the US.

Wissner-Gross offered a unifying theory: American labs are revenue-maximizing, while Chinese labs are not. OpenAI abandoned aggressive video research partly because Anthropic “ran away with their lunch” on code generation, which commands far higher revenue per token. Chinese labs, facing enterprise trust barriers that limit sales of models like Kimi or Qwen, pivoted to video—a global product with no security concerns. The strategic question is which path reaches artificial general intelligence first. America optimizes for better code, which could accelerate recursive self-improvement. China optimizes for modeling the physical world, which could accelerate embodied AI and robotics.

The US-China competition has already turned ugly. The podcast detailed a reported bot farm of approximately 200,000 inauthentic accounts—200 of them active—pushing narratives that American AI data centers are driving up electricity prices and hurting communities. “Shocked. Shocked I say that there are foreign influence operations attempting to suppress American AI data center deployment,” Blundin said, his sarcasm underscoring how predictable the tactic was. The counter-narrative, the hosts argued, needs to be told more aggressively. The story of Quincy, Washington, where data centers helped cut the poverty rate from 29.4% to 6.2% and funded a new high school, hospital, and library, is not anecdote—it is a template for the abundance argument.

Space: The Ultimate Diversification

For Peter Diamandis, the conversation kept circling back to one name that serves as a bet on every layer of the stack. “SpaceX is everything,” he said. “It’s ultimately chips and data centers and models and comms and infrastructure and launch and you get a chance to bet on all of them.” It is, he made clear, his largest financial bet. The numbers are staggering. Elon Musk has projected SpaceX revenue of $3.5 trillion by 2033, a figure that would dwarf any company in history.

Wissner-Gross sees the most obvious path to that number through a SpaceX-Tesla merger, which would bring Tesla’s Optimus robot revenue into the SpaceX fold. He predicts such a merger could happen within the next year or two. Beyond that, the Dyson swarm—a network of orbital compute platforms—is the long-term prize. “The Dyson swarm alone could generate trillions of dollars in revenue,” Wissner-Gross said, “while Starlink and cars become rounding errors in Elon’s ecosystem.” For investors, the takeaway is not to pick a single layer of the AI stack, but to find the companies—like SpaceX or Nvidia—that have positioned themselves to capture value across multiple layers simultaneously.

The deepest tension in the entire discussion remains unresolved. Wissner-Gross’s warning about financial engineering sits uneasily against the collective optimism of the group. But even he concedes that a correction would be brief and shallow. The reason is that AI is not a speculative bet on a single product. It is, as he put it, the “fundamental substrate for human civilization.” The investment implication is not to avoid AI infrastructure. It is to understand which companies are genuinely diversified across the stack—and which are, as Blundin suggested of Apple, merely coasting on past glory. The next two years will determine whether Nvidia’s circular economy is a sign of structural strength or a house of cards. For now, the numbers speak loudly enough to drown out most doubts.