Despite hyperscaler pledges to break free from proprietary silicon ecosystems, enterprise AI capital expenditure remains firmly locked onto Nvidia's hardware stack. Speaking at the Goldman Sachs Communacopia + Technology Conference, founder and CEO Jensen Huang reiterated an aggressive expansion trajectory that shrugs off ongoing FinOps scrutiny.
According to Huang, Nvidia is confident in delivering 70% year-over-year revenue growth in the upcoming fiscal year. With market analysts projecting the company will close its current fiscal year at approximately $400 billion, a 70% increase pushes forward-looking revenue toward an unprecedented $680 billion. This trajectory is underpinned by massive enterprise appetite for dense compute clusters.
"I think we could grow 70% year over year. We’re confident about that," Jensen Huang stated regarding the company's financial trajectory.
Specifically, production orders for full-rack configurations pairing 36 Grace CPUs with 72 Blackwell GPUs are expanding at 27% month-over-month. For technical leaders and CFOs planning infrastructure over the next 12 to 18 months, this pipeline confirms that training frontier models and running high-throughput inference still demand Nvidia's integrated silicon.
Ecosystem Reach and Infrastructure Visibility
Custom ASIC initiatives from Amazon, Google, and Microsoft—alongside in-house architectural explorations from OpenAI and Anthropic—have yet to undermine CUDA’s entrenched developer moat. As Huang outlined, Nvidia infrastructure continues to power core workloads across frontier labs and open-weight ecosystems alike.
To safeguard delivery schedules against physical bottlenecks, the company maintains direct operational tracking over data center expansion worldwide. Huang noted that Nvidia monitors gigawatts of land, grid power allocations, and physical data center shells globally across its partner networks, securing unmatched supply-chain visibility.
"We’re tracking every single gigawatt of land, power, shell around the world. Literally everything on the planet," noted Jensen Huang when describing the operational scale required to anticipate compute demand.
For enterprise decision-makers, the signal is clear: while custom chips offer long-term promise for specialized workloads, market-wide capacity shortages and software stickiness ensure Nvidia will continue extracting its silicon rent well into the next investment cycle.