Citibank: Nvidia's single GW revenue may rise from $18 billion to $40 billion, as AI infrastructure value continues to increase

Nvidia is evolving from a GPU chip company into a full-stack platform company covering computing power, networking, security, and AI infrastructure. The AI wave is simultaneously expanding both its demand scale and the value per project.

Citibank recently released a report, maintaining Nvidia's "Buy" rating with a target price of $315. The report points out that Nvidia's additional $150 billion stock buyback authorization raises its buyback capacity to $235 billion through FY28, reflecting the company's confidence in AI-driven free cash flow growth.

At the same time, Nvidia is expanding into AI security, compliance, governance, and comprehensive AI factory infrastructure. The rapid proliferation of open-weight models continues to drive up token consumption and the underlying demand for computing power.

More notably, Nvidia is increasing the value of each AI infrastructure deployment. The company expects that the revenue opportunity per GW will rise from $18 billion in the Hopper era to $25 billion in the Blackwell era, and further to $40 billion in the Rubin era.

$150 Billion Buyback: Cash Flow Confidence Behind the Massive Authorization

After Nvidia's new $150 billion stock buyback authorization, the total buyback capacity through FY28 reaches $235 billion. With the company’s market cap at $5.52 trillion, buybacks of this scale have become a key part of capital allocation.

Citibank believes the increased buyback scale reflects Nvidia’s confidence in future free cash flow growth. As demand for AI infrastructure continues to rise, the company expects to maintain strong profitability and cash generation capability, enabling it to continue expanding AI investments while delivering significant returns to shareholders.

In other words, Nvidia is not making a trade-off between growth investment and capital returns. On one hand, the company continues investing in AI infrastructure, software, and emerging applications; on the other, it retains ample room for additional buybacks.

Expanding From GPU to AI Security: Broader Business Boundaries

Nvidia has launched the Open Agent Safety Platform, entering the realms of AI agent security, compliance, and governance. The platform includes the open-source OpenShell runtime and the Sentry hardware monitoring system running on BlueField-4 DPUs, which can implement policy controls for AI agents and detect or block unauthorized behaviors.

As AI agents enter enterprise-critical workflows, demand for security and control rises in tandem. Nvidia is thus extending from underlying computation into the security layer required by AI application operation, expanding its addressable market.

Meanwhile, the open model ecosystem is expanding rapidly. Open models now account for around 75% of token generation, up from about 40% a year ago; overall token consumption is up 25-fold year-over-year. Nvidia believes that greater openness in models is pushing AI adoption, and increased token demand is further driving the need for underlying compute power.

The company continues to invest in language AI, physical AI, robotics, and autonomous driving and has plans to acquire Hugging Face to further strengthen ties with the developer ecosystem.

Per GW Revenue Opportunity Rises from $18 Billion to $40 Billion

Nvidia estimates that the revenue opportunity per GW has increased from $18 billion in the Hopper era to $25 billion in the Blackwell era, and will reach $40 billion in the Rubin era.

This growth is not only due to improved GPU performance but also from including CPUs, LPUs, networking devices, and AI factory infrastructure as part of the overall solution. As a result, Nvidia can cover more stages of AI infrastructure deployment, increasing project revenue per customer.

Order data also shows demand for AI infrastructure continues to expand. Nvidia revealed that one leading AI laboratory has directly signed for 2.6 GW of Nvidia AI infrastructure, with related equipment to be delivered by 2028. In addition, multiple cloud service providers (CSPs) and new cloud vendors have reached a combined indirect agreement value of over $180 billion.

This means Nvidia’s growth stems not only from the expansion of AI infrastructure deployments but also from the company covering more products and stages within individual projects. As Hopper evolves to Blackwell and Rubin, the revenue opportunity per GW continues to rise.