Nvidia's older H100 graphics processing units (GPUs) are proving far more durable moneymakers than expected, with rental rates climbing 22% last month to $3.28 per hour. This development challenges assumptions about how quickly AI hardware becomes obsolete and reveals a more nuanced economics of the artificial intelligence boom than many observers realized.

Nvidia CEO Jensen Huang highlighted this trend as the company reported extraordinary quarterly results. The chip maker generated $96 billion in revenue for the quarter, up 106% year-over-year, with data center sales reaching $89 billion. Yet amid this explosive growth, Huang's focus on the staying power of three-year-old H100 chips signals something important about the AI infrastructure market: not every piece of hardware needs to be replaced immediately.

Why Are Older AI Chips Still Valuable?

The rising rental rates for H100 GPUs suggest that cloud providers and AI companies still find value in older hardware, even as Nvidia releases newer processors. When rental prices increase, it typically means demand is outpacing supply, or that the hardware remains competitive enough to justify its cost. This extends the useful life of AI infrastructure and delays the need for wholesale replacement cycles.

The economics matter because they affect how much companies spend on AI infrastructure. If older chips remain productive and rentable at higher rates, businesses can spread their capital investments over longer periods. This also creates a secondary market for used AI hardware, which could reshape how companies approach their AI infrastructure budgets.

How to Understand Nvidia's Expanding AI Ecosystem Strategy

• Strategic Investments: Nvidia invested in Firmus, a company that signed a multiyear deal to supply OpenAI from two Malaysian data centers, while deploying Nvidia's newer Vera Rubin processors across Firmus's Asia-Pacific facilities.
• Funding AI Research: Nvidia funded Mistral AI to support model training and frontier AI research, expanding potential customers for Nvidia's chips and growing its AI compute ecosystem beyond OpenAI and Anthropic.
• Analyst Confidence: Analysts maintain Buy ratings on Nvidia stock, with price targets ranging from $300 to $390, and consensus near $349.15, reflecting confidence in the company's long-term positioning.

These moves reveal Nvidia's strategy to embed itself deeper into the AI infrastructure stack. Rather than simply selling chips, the company is investing in the companies and ecosystems that will drive demand for its processors. By funding Mistral AI and partnering with Firmus, Nvidia ensures it remains central to AI development regardless of which specific models or platforms gain dominance.

What Questions Remain About Nvidia's Business Model?

Short-seller Jim Chanos has raised pointed questions about Nvidia's economics that deserve attention. Chanos questioned Huang about why Nvidia doesn't rent chips directly, why it hasn't raised prices more aggressively, and whether cloud GPU providers capture most of the value in the AI infrastructure market. These questions hint at a potential tension: if rental rates are rising, why isn't Nvidia capturing more of that upside directly?

The answer likely involves Nvidia's business model. The company primarily sells chips to cloud providers and data center operators, who then rent access to customers. Nvidia's margins come from the initial sale, not from ongoing rental revenue. This means Nvidia benefits from rising demand for chips, but doesn't directly participate in rental rate increases. Whether this model remains optimal as the AI market matures remains an open question.

Nvidia's next-quarter guidance of approximately $108 billion in revenue suggests the company expects continued explosive growth. However, the focus on H100 chip economics and the questions raised by skeptics indicate that investors and analysts are beginning to scrutinize whether Nvidia's valuation reflects realistic assumptions about hardware replacement cycles, pricing power, and the distribution of value across the AI infrastructure supply chain.

The story of the H100's extended profitability is ultimately a story about the maturation of AI infrastructure. As the market moves beyond the initial gold-rush phase, the economics become more complex. Older hardware remains useful, rental markets develop, and the question of who captures value becomes more contested. For Nvidia, maintaining dominance means not just building the fastest chips, but ensuring its ecosystem strategy keeps the company central to AI development for years to come.