Nvidia Signs White House AI Pact as Buyback Reaches $235 Billion and Morgan Stanley Backs $300 Target

Nvidia signs a White House AI safety agreement, authorizes $150B more in buybacks, and gets a $300 Morgan Stanley target as AI demand outruns supply.

Jensen Huang put his signature to a voluntary artificial intelligence development agreement at the White House on September 29, according to the Associated Press, aligning Nvidia with an initiative joined by other leading technology firms. President Donald Trump indicated the framework includes both internal and external reviews — a commitment that lets the chipmaker shape its own safety architecture rather than wait for regulation to catch up with advanced AI models.

That posture dovetails with a safety package Nvidia unveiled roughly a week earlier. The Open Agent Safety Platform ships as an open software bundle with a reference design dubbed Sentry, built to monitor autonomous agents from testing through live operation. Should a system breach its defined operating limits, a quarantine mechanism steps in to shut down uncontrolled actions. The underlying aim: keep Nvidia's graphics processors central not just to model training but to the regulated deployment of future AI applications.

Capital Returns and Analyst Conviction

Shareholders are being looked after as well. Nvidia's board authorized an additional $150 billion in stock repurchases about a week ago, lifting the total remaining buyback capacity to $235 billion. Management intends to work through the program by fiscal 2028.

Morgan Stanley's Joseph Moore reaffirmed an "Overweight" rating on Friday with a $300 price target. The bank pointed to Nvidia's broadening customer base and the potential for autonomous agents to sustain GPU demand, while flagging hurdles in power supply, construction and financing of new data centers. Revenue streams have diversified well beyond the well-known American tech giants — roughly half of sales now come from specialized model developers, alternative cloud providers, enterprises and government institutions. A clear majority of connected cloud partners run their infrastructure outside the United States, and the leading cloud service providers keep raising their hardware budgets.

Morgan Stanley expects real demand to outstrip supply in the coming fiscal year, leaving growth capped mainly by available delivery capacity. The bank's projections call for annual revenue to climb from about $216 billion in 2026 to roughly $406 billion in 2027, with around $690 billion within reach the year after. Gross margin is seen holding at a high level, with technological advances driving efficiency gains — the Vera Rubin platform promises a substantial jump in token output over the current Blackwell generation, and the later Feynman architecture is expected to generate still-growing revenue per gigawatt of data center capacity.

New Financing Structures Reshape the Demand Picture

How customers pay for all that expensive hardware is shifting too. 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 arrangement. Such sale-and-leaseback structures let cloud operators lighten their books while still running enormous computing capacity — and for Nvidia, they lock in sales of its accelerators as the global data center buildout continues.

In European trading, the stock ended Friday at EUR 207.85, up 1.1%, leaving it just 1.5% below its 52-week high.