Nvidia's 'Visibility' Calms AI Bubble Fears: Buy?

During Nvidia's second-quarter earnings call, CEO Jensen Huang used the word 'visibility' exactly once, but that single word carried significant weight. He described how Nvidia can now see further upstream and downstream than ever before—into wafers, memory, optics, land, power, and the facilities that will house the next wave of AI systems. This comes against a backdrop of record results and a supply-constrained outlook.

The AI bubble narrative has persisted for over a year, fueled by concerns about circular financing, stretched balance sheets, and the fear that demand is manufactured by the same companies selling the picks and shovels. Critics point to hyperscalers and AI labs spending enormous sums on accelerators that generate tokens, with revenues justifying more chip purchases—a loop reminiscent of the fiber optic boom of the late 1990s.

Skeptics also argue that custom silicon from Amazon, Alphabet, Microsoft, and Meta could erode Nvidia's pricing power. They note negative free cash flow at some of the largest AI spenders as capital expenditures surge ahead of operating cash flows. This paints Nvidia as a fashion that will fade after the first generation of data centers is depreciated.

However, Huang's comments suggest a different reality. By working with 'land, power, and shell companies all around the world,' Nvidia can better prepare the computing infrastructure that will be built. This planning has given Nvidia unprecedented visibility, allowing it to guide for 70% revenue growth in fiscal 2028—a forecast it has historically avoided making that far ahead.

Management indicated that demand could support even higher growth, but supply constraints make 70% a conservative floor. This level of visibility is not abstract: top hyperscalers are expected to spend nearly $800 billion on capex in 2026 and $1.3 trillion in 2027, with cloud backlogs around $2 trillion. These figures are published to support build plans, not marketing hype.

Nvidia is evolving from a chip vendor into the architect of AI factories—full-stack systems where CPUs, GPUs, networking, software, and physical sites are designed together. This shift increases the revenue opportunity per gigawatt of power, with costs rising from $18 billion per gigawatt in the Hopper era to $25 billion with Blackwell and $40 billion with Vera Rubin.

This strategy means Nvidia's competitive position strengthens as it captures more of the entire AI infrastructure stack, not just individual chips. The company's equity stakes in Marvell Technology, Nokia, and Coherent extend its reach into custom silicon, radio-access networks, and optical interconnects, giving it line-of-sight into networking and telecommunications demand.

For investors, this visibility changes the risk profile. AI spending may be cyclical at the margin, with memory inflation pressuring near-term profitability, but the demand is secular and already booked across land, megawatts, and critical components. This is not a bubble waiting to pop; it's a multiyear industrial build-out.

Nvidia's ability to guide for fiscal 2028 revenue growth is unprecedented and signals confidence in its supply chain and customer commitments. If the company can deliver on this guidance, it will validate the AI infrastructure thesis and potentially drive further upside in the stock.

However, risks remain: rising memory prices could compress margins, and competition from custom silicon could intensify. But for long-term investors, Nvidia's positioning as the organizer of AI factories suggests it will capture a rising share of each new data center, making it a compelling hold.