Nvidia told investors that its AI infrastructure roadmap remains on track and demand across customer segments remains strong. Following discussions with Nvidia’s investor‑relations team, Citi reaffirmed its rating of the chipmaker as its top buy‑rated mega‑cap data‑center semiconductor stock, citing Nvidia’s strong access to constrained DRAM memory supplies. The company highlighted that the Rubin Ultra roadmap is fully intact and the NVLink domain architecture unveiled at Computex has not changed. It also confirmed that co‑packaged optics for scale‑out networking are already in production with Spectrum‑X and that customer adoption is high. Beginning with the Feynman platform scheduled for calendar 2028, customers will be able to choose between NVLink implementations using co‑packaged optics or copper interconnects.
Nvidia said AI spending, initially driven by hyperscale cloud providers, is now expanding to AI laboratories, sovereign customers and on‑premise enterprise deployments, which have gained momentum over the past two years and are expected to capture a larger market share as physical AI adoption accelerates. The firm noted that both open‑source and proprietary AI models are important, with frontier models pushing performance and open models enabling scale for enterprises and governments. Its own models, Nemotron and Cosmos, are intended to support enterprise and sovereign adoption rather than compete with leading frontier‑model developers.
Management clarified Chief Executive Jensen Huang’s earlier comment that $100 billion of compute could eventually require only one gigawatt of power, explaining that the figure reflects anticipated improvements in energy efficiency rather than infrastructure cost. Nvidia estimates that current systems generate roughly $30 billion to $40 billion of compute revenue per gigawatt, and that newer GPU generations such as Blackwell are significantly more power‑efficient than the previous Hopper generation.
Financially, Nvidia reiterated its target of returning 50% of cash flow to shareholders this year and indicated that share buybacks could increase over time. The company maintained its guidance for a mid‑70% gross‑margin and disclosed that a recent $25 billion debt offering was intended to provide greater financial flexibility. Nvidia also stated that it has not formally announced a collaboration with AI chip startup d‑Matrix, although it remains open to incorporating third‑party technologies into its platform. Additionally, Nvidia declined to comment on Meta Platforms’ cloud expansion plans.