Nvidia has attached the largest number yet to corporate America’s favorite expression of confidence. On September 28, its board added $150 billion to the chipmaker’s share-repurchase authorization, lifting the remaining program to $235 billion and setting an expected completion date in fiscal 2028. The increase is the largest ever announced, according to the company. It is also a useful test of how quality investors should think about buybacks: as evidence of financial strength, but never as proof that a stock is cheap.
At Tuesday’s price of $228.86, Nvidia was worth about $5.56 trillion and traded at roughly 28.7 times trailing earnings. The entire remaining authorization therefore equals only 4.2% of the company’s market value. The headline is monumental; the prospective reduction in the share count is not. The real question is whether Nvidia can keep generating cash quickly enough—and buy patiently enough—to make the program accretive after the current AI boom matures.
Record dollars translate into modest share shrinkage
The September 28 authorization is not an obligation to spend, and the timetable gives management room to adjust purchases to market conditions. If Nvidia used all $235 billion at today’s price, it could retire about 1.03 billion shares, or slightly more than 4% of shares outstanding. A rising share price would reduce that benefit; a falling one would enlarge it.
Recent experience shows why gross spending and net shrinkage are different measures. Nvidia spent $39.0 billion repurchasing stock during the first half of fiscal 2027. Yet its diluted weighted-average share count in the second quarter was 24.285 billion, down about 1% from 24.532 billion a year earlier. Some purchases offset employee issuance and stock-based compensation, which totaled nearly $4.0 billion in the first half. Timing also matters because shares bought late in a period contribute less to the average count.
None of that makes the program cosmetic. A sustained 4% reduction would meaningfully lift each remaining shareholder’s claim on future profits. But investors should monitor diluted shares, not the press-release total. Authorization is capacity; cancellation is the result.
The cash machine earns its capital-allocation freedom
Nvidia can contemplate a program of this size because its economics have become extraordinary. In its fiscal second-quarter report, revenue doubled to $96.2 billion, Data Center sales climbed 117% to $89.0 billion, and the gross margin reached 75%. Operating income was $63.7 billion. Those figures describe more than growth: they show exceptional operating leverage and pricing power in the defining infrastructure cycle of the decade.
Cash conversion was less spectacular than accounting profit, but still formidable. Second-quarter operating cash flow was $24.1 billion. After $2.7 billion of purchases related to property, equipment and intangible assets, plus small principal payments on those assets, free cash flow was about $21.3 billion. Nvidia returned roughly $26 billion through repurchases and dividends during the quarter, drawing on a balance sheet strengthened by prior cash generation.
The crucial quality marker is that distributions have not displaced investment. Research and development expense reached $7.1 billion in the quarter, 64% more than a year earlier. On the same day as the buyback announcement, Nvidia introduced an open agent-safety platform spanning software and hardware controls, with more than 100 organizations participating. The company is funding new architectures and ecosystem expansion while returning surplus capital. That hierarchy—invest first where returns appear exceptional, then distribute excess cash—is sound.
Price discipline will decide whether the buyback creates value
The counterargument is straightforward. At 28.7 times trailing earnings, Nvidia is paying about $28.70 for each dollar of current annual profit it acquires through repurchases, equivalent to a 3.5% earnings yield before growth. That can be an excellent trade if AI demand, margins and Nvidia’s competitive advantage endure. It can be a poor one if today’s profits represent a cyclical peak.
The latest numbers themselves contain reasons for restraint. Accounts receivable and inventories expanded sharply during the first half as the company scaled. Data Center represented more than 92% of quarterly revenue, concentrating the business around a small number of enormous infrastructure programs. Customers are racing to build capacity; their future return on that spending remains the ultimate constraint on Nvidia’s growth.
Management’s broad authorization is therefore best viewed as an option, not a target. The highest-quality decision may be to accelerate purchases during a dislocation and slow them when enthusiasm outruns cash-flow value. Investors should judge the program by three outcomes: the net decline in diluted shares, the price paid relative to normalized earnings, and whether product investment continues to widen the moat.
Nvidia’s announcement confirms a rare level of financial power. It does not settle the valuation debate. The buyback will create durable value only if management applies the same discipline to buying Nvidia shares that it applies to designing Nvidia chips.
