Nvidia (NASDAQ:NVDA) is set to release its latest quarterly earnings and guidance on February 25. Citi Group has given an optimistic outlook for the chip giant, led by Jensen Huang, and expects the company to release strong performance guidance.
In a report to clients, Citi analyst Atif Malik stated that his model predicts Nvidia’s revenue for the fiscal quarter ending in January will be approximately $67 billion, exceeding Wall Street’s consensus estimate of $65.6 billion. Furthermore, he expects the company’s guidance for the fiscal quarter ending in April to be around $73 billion, notably higher than the market’s expectation of $71.6 billion.
Malik pointed out that the continued ramp-up of B300 products, coupled with the launch of the Rubin architecture, will drive Nvidia’s sales to accelerate by 34% year-over-year in the second half of 2026, significantly outperforming the 27% growth expected for the first half of 2026. He believes that investors’ focus has shifted away from the current earnings report and towards the annual GTC conference scheduled for mid-March, where Nvidia is expected to focus on its inference roadmap. This will include details on how the company plans to utilize Groq’s low-latency SRAM intellectual property and provide its first preliminary outlook on AI-related sales from 2026 to 2027.
Based on this outlook, Malik maintains a “Buy” rating on Nvidia and sets a target price of $270.
Looking at the company from a longer-term perspective, Malik further stated that Nvidia’s current valuation “appears attractive.” As market visibility on its 2026 performance improves, the stock is expected to outperform the broader market in the second half of 2026.
He also mentioned that the inference market is evolving toward being “more diversified,” which will offer more choices for model scale and application scenario customization. This also means that the use of AI accelerators will take on more diverse forms. However, from a system-level perspective, he expects Nvidia to continue to lead in workloads focused on training as well as inference and logical deduction, and he believes that MLPerf remains the most valuable benchmark for comparing different AI accelerators.