Nutanix has invested $20 million in its own artificial intelligence infrastructure and anticipates recovering the full cost within a year, CEO Rajiv Ramaswami told The Register during a media briefing on August 27, 2026. The company built the on-premises cluster to escape escalating token costs from third-party AI services and now relies on open-weight models instead of commercial tools like Copilot and Claude. The move reflects a broader shift among enterprises matching specific models and infrastructure to individual workloads rather than applying large language models universally.
Nutanix reported $757 million in revenue for its fourth quarter, marking a 16 percent year-over-year increase. Full-year revenue climbed 12 percent to $2.85 billion, while net income reached $1.5 billion for the fiscal year. The company added 3,000 new customers during the period, many of whom chose Nutanix as a replacement for VMware. Ramaswami said hardware costs and availability are currently acute concerns for customers, issues that have slowed some software purchases.
The CEO explained that Nutanix's software teams initially relied on commercial AI coding assistants throughout the development cycle. "Usage exploded and so did costs," Ramaswami acknowledged, prompting the switch to open-weight models running on the company's own hardware. According to the CEO, Nutanix no longer pays per-token fees, though some users occasionally access frontier models or external clusters when specific workloads require them. CFO Rukmini Sivaraman said the fiscal 2026 results showed "a good balance of top and bottom line performance," with the company remaining focused on sustainable growth and improved profitability.
Nutanix is moving closer to supporting the Arm architecture, a shift driven by high hardware prices and a goal of ensuring customers can deploy its software wherever they choose. Ramaswami told The Register that adopting Arm will allow the platform to run on lower-cost hardware that customers won't hesitate to purchase in the current economic climate, advancing beyond the company's 2024 position when Arm support was considered a future possibility but not a development priority. The company is also expanding hardware compatibility and supporting external storage devices to help users migrate from VMware without replacing existing equipment, while continuing to shrink its software footprint through measures like allowing bare-metal installations of its Kubernetes Platform. Ramaswami predicted Nutanix will continue winning customers from VMware for another five years. The shift to internal AI infrastructure and open models suggests enterprises may increasingly bring compute-intensive workloads in-house when usage patterns justify the capital outlay, potentially reshaping vendor economics across the sector.