Citi maintains Buy rating on NVDA with a US$300 target price
AI summary card
Citi maintains Buy rating on NVDA with a US$300 target price
Through discussions with NVIDIA's investor relations team, Citi addresses investor questions regarding the product roadmap, CPO, AI demand composition, power economics, capital returns, and valuation, while continuing to list NVDA as its preferred Buy among large data center semiconductor stocks.
- NVIDIA stated that the Kyber Rubin Ultra roadmap remains intact, and the NVLink domain showcased at Computex is also unchanged.
- CPO scale-out has entered production with Spectrum-X; starting with CY2028 Feynman, customers can choose NVLink paired with either CPO or Copper.
- The company believes AI demand is not limited to hyperscalers; AI labs, sovereign AI, on-premises enterprise deployments, and physical AI will drive the long-term expansion of the non-hyperscaler market.
- The long-term US$100B/1GW figure is explained as an energy-efficiency metric: future US$100B of computing will require only 1GW, versus approximately US$30B-US$40B per 1GW currently; Blackwell is 25x more energy efficient than Hopper.
- Citi's US$300 target price is based on approximately US$11.9 of C27E EPS and a discounted P/E of approximately 28x, implying a 47.0% expected share price return.
Report interpretation
Overview
This report is a company research update from Citi on NVIDIA Corp. Its core format is to organize questions that investors have recently focused on and provide responses from management or the investor relations team. The report covers the Kyber Rubin Ultra roadmap, CPO and Copper options, Meta's cloud plans, the demand composition of hyperscalers and neoclouds, open- and closed-source models, computing costs per GW, space data centers, the impact of long-term storage agreements, capital allocation, and external technology partnerships.
Core views
Citi maintains its Buy view on NVDA and calls it the preferred Buy among large data center semiconductor stocks. Key supports include continued strength in AI computing demand; NVIDIA's strong ability to secure supply amid constrained DRAM availability; an intact product roadmap; continued advancement of platforms including CPO, NVLink, Spectrum-X, and Feynman; expansion of AI demand sources from early hyperscalers to AI labs, sovereign AI, on-premises enterprise deployments, and physical AI; and improvements in GPU generational energy efficiency that could support higher revenue intensity per GW.
Analysis framework
The report uses an investor Q&A format, breaking down the themes of greatest recent market interest and combining management commentary, the product roadmap, customer demand composition, and the valuation framework to form an investment view. For valuation, it applies an approximately 28x P/E multiple to approximately US$11.9 of C27E earnings power and discounts the result to derive a US$300 target price.
Methodology notes
Target price based on approximately 28x P/E and C27E earnings power
Citi applies an approximately 28x P/E multiple to approximately US$11.9 of C27E earnings power and discounts the result to derive a US$300 target price for NVDA; the multiple is described as consistent with the three-year average.
The Buy rating is determined jointly by expected total return and risk
Citi discloses that its Buy, Neutral, and Sell ratings are based on expected total return over the next 12 months and risk; in this report, both NVDA's expected share price return and expected total return are 47.0%.
Summarizing key questions through investor relations communications
The report addresses investor questions on the roadmap, CPO, customer demand, the model ecosystem, power economics, and capital allocation one by one to assess whether the investment thesis remains valid.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- NVDA.OCore covered asset
- Strengths
- Strong AI infrastructure demand, strong ability to secure DRAM supply, an intact roadmap, continued advancement of the CPO and NVLink platforms, and additional support from capital returns and buyback expectations.
- Weaknesses
- The valuation depends on high growth and a high P/E multiple, making it sensitive to the pace of data center demand, platform adoption, and market volatility.
- Comparison
- Listed as the preferred Buy among the large data center semiconductor stocks covered by Citi.
- Risks
- Competition, slower-than-expected platform adoption, volatility in the automotive and data center markets, and the impact of cryptocurrency mining on gaming sales.
- META.USCustomer variable related to AI cloud demand
- Strengths
- The report notes investor interest in Meta's cloud plans, reflecting continued market focus on AI capital expenditure by large internet platforms.
- Weaknesses
- NVIDIA does not comment on customer plans, so the report cannot confirm Meta's specific procurement or cloud expansion pace.
- Comparison
- Compared with hyperscalers, the report places greater emphasis on the future expansion of non-hyperscaler demand.
- Risks
- If cloud plans or AI capital expenditure by major customers fall below expectations, this could affect the market's assessment of the sustainability of NVIDIA's demand.
Key data
- RatingBuyCiti maintains its Buy rating on NVDA.
- Target priceUS$300.00The target price is based on an approximately 28x P/E multiple and discounted C27E earnings power of approximately US$11.9.
- Reference priceUS$204.12The price timestamp is 2026-07-08 16:00.
- Expected share price return47.0%Expected share price return as disclosed in the table.
- Expected dividend yield0.0%Expected dividend yield as disclosed in the table.
- Expected total return47.0%Expected total return as disclosed in the table.
- Market capitalizationUS$4,939,704MMarket Cap as disclosed in the report table.
- Long-term computing efficiency statementUS$100B/1GWManagement explained this as the result of future efficiency improvements: US$100B of computing will require only 1GW, compared with approximately US$30B-US$40B per 1GW currently.
- Blackwell energy efficiency relative to Hopper25xThe report states that Blackwell is 25x more power efficient than Hopper.
- Capital return target50% of cash flowNVIDIA reiterated its plan to return 50% of cash flow to shareholders this year.
Impact & implications
The investment implication is that concerns regarding roadmap delays, CPO adoption, the AI customer mix, power constraints, and capital allocation have not changed Citi's positive view. If NVIDIA continues to demonstrate that AI demand is expanding from hyperscalers to a broader customer base while generational GPU energy efficiency continues to improve, data center revenue intensity and valuation support may remain intact.
Risks
- Competition in gaming could cause NVIDIA to lose market share and pressure its stock price.
- Slower-than-expected adoption of new platforms could weigh on data center and gaming sales.
- Order volatility in the automotive and data center markets could increase volatility in the stock price and valuation multiple.
- Changes in cryptocurrency-mining-related demand could affect gaming business sales.
- If the CPO, NVLink, Feynman, and other roadmap initiatives underdeliver, the long-term platform narrative could weaken.
- If AI customer capital expenditure or ROI falls below expectations, revenue intensity per GW and demand sustainability could be affected.
What to watch
- Whether the Kyber Rubin Ultra roadmap continues to advance as planned.
- Details of customer adoption of CPO scale-out within Spectrum-X.
- Customer choices between NVLink paired with CPO or Copper during the CY2028 Feynman phase.
- Whether AI demand continues expanding from hyperscalers to AI labs, sovereign AI, on-premises enterprise deployments, and physical AI.
- Changes in AI cloud and data center capital expenditure by Meta and other major customers.
- Whether NVIDIA increases buybacks, and the capital allocation flexibility provided by US$25B of debt financing.
- Whether Blackwell and subsequent GPU generations deliver the expected energy-efficiency improvements and support higher revenue intensity per GW.