NVIDIA's latest earnings report signals a fundamental shift in how the AI industry views computing power: no longer as equipment that loses value over time, but as infrastructure that generates revenue indefinitely. The company reported $96.2 billion in revenue for the second quarter of fiscal 2027, up 106 percent from the same period last year, with data center revenue reaching $89 billion, up 117 percent year-over-year.

The numbers are staggering, but the real story lies in what CEO Jensen Huang said about what those numbers mean. "AI has reached its inflection point. It's doing useful work. Its tokens are productive and profitable. Now, compute is revenue," Huang stated. This framing represents a seismic shift in how the industry thinks about graphics processing units (GPUs), the specialized chips that power artificial intelligence systems.

Jensen Huang

Why Is NVIDIA Reframing GPUs as Infrastructure?

For decades, computer hardware was treated like any other piece of equipment: you bought it, used it, and eventually it became obsolete. Huang's new characterization suggests something different. When GPUs are installed in a powered data center facility with proper software and network connectivity, they can generate ongoing, usage-based income for years, functioning more like a toll road or utility plant than a laptop destined for disposal.

This reframing matters because it attracts a completely different class of investor. Pension funds, insurance companies, and infrastructure-focused investment firms have traditionally backed long-term assets like bridges, power plants, and toll roads. If AI compute can be positioned as a similar durable, revenue-generating asset, it opens the door to hundreds of billions of dollars in capital that would never touch a traditional hardware purchase.

To that end, NVIDIA announced a partnership with six major asset management firms, including Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR, to launch independent compute financing platforms aimed at mobilizing over $500 billion in third-party capital for AI infrastructure development. The message is clear: compute is no longer a one-time purchase; it's a long-term investment vehicle.

What Infrastructure Beyond Chips Is Actually Needed?

While GPUs dominate headlines about artificial intelligence, they cannot operate in isolation. A functioning AI data center requires reliable power, backup generation systems, fast networking, sophisticated memory, and cooling systems capable of managing extremely high computational loads. According to the International Energy Agency, servers account for roughly 60 percent of the electricity used within modern data centers, while cooling systems can consume anywhere from about 7 percent in highly efficient facilities to more than 30 percent in less efficient ones.

This reality is reshaping how companies approach AI infrastructure development. Instead of simply selling GPUs, companies are now combining hardware sales with energy-supported hosting infrastructure and their own computing operations to generate value from the entire ecosystem. The worldwide spending on AI infrastructure is expected to hit approximately $487 billion in 2026 and climb past $1 trillion by 2029, with a large share of that money flowing toward securing land, power, and network connectivity rather than semiconductors alone.

How to Evaluate AI Infrastructure Investments

• Power Availability: Assess whether a facility has access to reliable, long-term power sources, including backup generation systems and the ability to handle extreme computational loads without interruption.
• Cooling Capacity: Evaluate the efficiency of cooling systems, as they can consume anywhere from 7 to 30 percent of total data center electricity depending on design, making this a critical cost factor.
• Network Connectivity: Verify that the facility has fast, dedicated fiber connections and the ability to move data quickly between systems and to end users.
• Software and Optimization: Confirm that the infrastructure includes NVIDIA's CUDA software platform and other tools that allow compute resources to be reused across different clients and applications over time.
• Long-Term Revenue Potential: Assess whether the compute infrastructure can generate ongoing, usage-based income rather than serving as a single point-of-sale transaction.

NVIDIA's latest announcements reflect this broader infrastructure-first mindset. The company revealed that its Vera Rubin platform is ramping into full production with racks running at partners including CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure, and Nebius. The company also announced the NVIDIA DSX platform, providing infrastructure builders a complete playbook to design, build, and operate AI factories at scale.

Beyond the headline earnings, NVIDIA also unveiled several new products and partnerships aimed at supporting this infrastructure buildout. The company announced NVIDIA Vera, the first CPU (central processing unit) built specifically for AI agents, with broad adoption planned across leading technology providers. It also introduced NVIDIA Groq 3 LPX, an interactive AI inference accelerator now in full production, and new security innovations for the Vera BlueField-4 STX, delivering agentic AI storage processing with built-in security for AI factories.

"Compute is a scarce, mission-critical asset class with compelling investment characteristics," said Jim Zelter, president of Apollo, one of the firms backing the new compute financing initiative.

Jim Zelter, President at Apollo

The scale of capital flowing into AI infrastructure is unprecedented. Beyond the $500 billion financing initiative involving six major asset management firms, NVIDIA and SK Group unveiled a separate partnership described as "a $500-billion-plus initiative spanning AI factories and next-generation memory". SK Telecom will construct a two-gigawatt NVIDIA Vera Rubin DSX AI Factory to help meet worldwide compute needs, and the collaboration establishes a long-term supply and joint-development agreement between NVIDIA and SK hynix focused on advanced memory chips.

For smaller, regionally based developers, this wave of capital is also opening new opportunities. Companies that can demonstrate they have secured land, power, and genuine customer interest are finding financing available for projects that can progress more quickly and expand step-by-step, especially in areas with accessible land and energy resources. This democratization of AI infrastructure investment suggests that the buildout will not be limited to hyperscale companies and massive sovereign-backed transactions.

Looking ahead, NVIDIA expects revenue of $108 billion for the third quarter of fiscal 2027, plus or minus 2 percent, with gross margins expected to be 74 percent, plus or minus 50 basis points. The company is not assuming any data center compute revenue from China in its outlook, reflecting ongoing geopolitical considerations. The trajectory is clear: as AI moves from experimental to productive, the infrastructure required to support it is becoming the defining investment opportunity of the decade.