Nvidia Pays Its Own Customers: The $10 Billion Finale That Rewrites the Chip Playbook

Nvidia and SoftBank each sent a final $10B tranche to OpenAI, completing $30B commitments, as Morgan Stanley reiterates Overweight with a $300 target.

During California's gold rush, the merchants who sold shovels rarely joined the prospectors panning in the riverbed. They supplied the tools, collected the dollars, and left the risk to others.

Nvidia has abandoned that time-honored division of labor. The world's largest chipmaker is no longer content to arm the artificial intelligence boom from the sidelines — it now bankrolls its biggest buyers directly, and on a scale that has few precedents even by Silicon Valley's generous standards.

A circular flow of capital

Nvidia and SoftBank Group each wired a final $10 billion tranche to OpenAI on Thursday, according to media reports, completing separate $30 billion commitments tied to the ChatGPT developer's latest funding round. The money travels a remarkable loop: billions move from the hardware pioneer to the world's most prominent software lab, and a substantial share flows straight back to pay for the computing capacity that would be unthinkable without Nvidia's accelerator cards. Bold strategic partnership, or the opening chapter of a troubling mutual dependency? Observers have wrestled with that question for some time. For Nvidia, the offensive is simply consistent — whoever wants to shape the architectures of the future cannot wait at the factory gate for orders to arrive.

From boardroom to the White House

The company's reach extends well beyond its balance sheet. Reuters reported that CEO Jensen Huang joined other industry leaders at a meeting with US President Donald Trump on Tuesday, where, according to House Speaker Mike Johnson, the executives signed an agreement on shared standards for artificial intelligence. Nvidia now moves through diplomatic circles almost like a state in its own right, helping to negotiate the rules of a technology poised to reshape both economy and society. Greater exposure inevitably invites stronger headwinds — an almost natural law of the stock market.

Legal friction is mounting accordingly. Roughly a week ago, the company put its trademark dispute over its Modulus software platform with Modulus Financial Engineering on hold, pausing proceedings in a US federal court. Fresh trouble looms elsewhere: Netlist filed a complaint with the US International Trade Commission on Tuesday, seeking an import ban on memory chips from Micron as well as products from Nvidia, Google, and Broadcom that allegedly infringe Netlist patents.

Investors shrug off the noise

None of this has rattled shareholders. Confidence in the tech giant's market dominance still rules the tape. The stock closed Friday at EUR 207.85, just 1.5% below its 52-week high. Support also comes from the company's own coffers: after Nvidia boosted its share buyback program roughly a week ago, the price has climbed 3.2%. Board member Timothy S. Teter's sale of 30,460 shares on September 21 under a pre-arranged trading plan fits the routine rhythm of executive transactions and barely registers in the order book.

Morgan Stanley's $300 target

On Friday, Morgan Stanley reaffirmed Nvidia as its top pick in the semiconductor sector, repeating an "Overweight" rating with a $300 price target. The bank's optimism rests on the widening base of revenue streams. The business no longer leans solely on the familiar American technology giants — roughly half of sales now come from specialized model developers, alternative cloud providers, enterprises, and government institutions. International business is gaining weight too, with a clear majority of connected cloud partners running their infrastructure outside the United States, while leading cloud providers keep expanding spending on the company's hardware. Morgan Stanley expects actual demand to outstrip supply in the coming fiscal year, leaving growth constrained mainly by available delivery capacity.

The numbers behind the thesis

The analysts project vigorous expansion ahead. Annual revenue should climb from about $216 billion in 2026 to roughly $406 billion in 2027, with approximately $690 billion deemed achievable the year after. Gross margin is expected to hold steady at a high level, with technological progress delivering efficiency gains. The Vera Rubin platform promises a substantial increase in token output versus the current Blackwell generation, and for the later Feynman architecture the bank models further rising revenue per gigawatt of delivered data center capacity.

New ways to finance expensive hardware

Alongside the operational buildout, the structures customers use to pay for costly hardware are shifting. According to the Financial Times, Amazon is negotiating a special-purpose vehicle to move roughly $8 billion worth of Nvidia processors off its balance sheet through a leasing model. Such sale-and-leaseback arrangements let cloud operators lighten their books while running enormous computing capacity — and for Nvidia, they lock in demand for its accelerators at the heart of the global data center expansion.

Nvidia is transforming at breathtaking speed from a highly specialized semiconductor manufacturer into the dominant power center of the entire AI economy. Through multibillion-dollar capital stakes and political presence, it is cementing an ecosystem with no obvious way around it. Cluster risks are undeniable should the pace of technological development ever slacken. But as long as demand for computing power runs unabated, Jensen Huang and his team hold every decisive lever at once.