Nvidia is down roughly 14 percent from its recent peak after Michael Burry sharpened his attack on the chipmaker’s buyback strategy and Nvidia fired back with a point-by-point memo to Wall Street. Shares of Nvidia closed at 180.26, up 1.34 percent, but volatility is elevated as investors weigh whether soaring AI profits are translating into per-share gains. At stake is a simple question with big implications for the market’s most important stock: did $100 billion-plus in repurchases meaningfully lift owner value, or did stock-based pay hand it back out the side door.
Burry’s analysis, posted this week, says Nvidia generated about 205 billion dollars in net income and 188 billion dollars in free cash flow from 2018 through mid-2025, while buying back 112.5 billion dollars of stock. He argues those buybacks added zero shareholder value because Nvidia simultaneously issued 20.5 billion dollars of stock-based compensation at grant prices, leaving diluted shares up by 47 million despite the repurchases. By his estimate, per-share earnings power is roughly 50 percent lower than it would have been without dilution. Nvidia disputes the framing and the totals. In a memo to analysts, the company said it repurchased 91 billion dollars since 2018, not 112.5 billion, and urged investors not to conflate equity compensation with the performance of the buyback program. Management says its pay structure tracks peers and supports growth.
The market’s reflex is to focus on absolute profit. Nvidia’s numbers are staggering and still growing. But what shareholders actually own is a slice, and the slice matters. If the share count is expanding through equity grants when the stock is at record highs, the math gets punitive. Repurchases executed at lofty valuations have to run just to stand still against expensive stock-based pay. And because many investors and some models focus on non-GAAP metrics that back out SBC as a non-cash item, dilution risk can get underappreciated until it hits per-share results. That is the heart of Burry’s critique: headline earnings can surge while per-share wealth creation lags. The recent 14 percent pullback suggests the market is re-rating not just growth durability, but the quality and distribution of those gains.
The Cisco analogy is doing the rounds for a reason. At the peak of the dot-com buildout, Cisco looked untouchable as orders exploded and stock-based pay kept top engineers loyal. Then growth normalized, competition intensified, and dilution compounded into a drag that the buyback could not offset. The AI era is not 2000, and Nvidia’s positioning is stronger than any infrastructure vendor of that time. Still, the parallel is less about collapse and more about compounding. If you pay with stock during the boom and buy back at peak multiples, you are using a very expensive currency on both sides. Nvidia’s memo rightly notes SBC is standard in Silicon Valley and aligned with long-term value creation. The question is cost. At trillion-dollar-plus market caps, standard practices can carry nonstandard consequences.
Repurchases are sold as a tax efficient way to boost per-share metrics and return surplus cash. They work best when two conditions hold: the stock trades below intrinsic value and the share count is not expanding elsewhere. Nvidia’s program accelerated post-2023 as profits surged, exactly when the stock also went parabolic. If repurchases lean into peak valuation and compete against high-velocity SBC, they risk becoming round trips. Burry estimates that happened here, asserting the true cost of dilution effectively consumed the buyback. Nvidia rejects that conclusion and says grants should be judged on retention and productivity, not offset against repurchases. For shareholders, the reconciliation is simple: track diluted shares outstanding and per-share free cash flow over time. If both do not improve meaningfully, the buyback did not do its job.
There is a second front that makes the buyback math more sensitive: competition. Google’s Tensor Processing Units are pressing on cost and energy efficiency for certain workloads. If hyperscalers can tilt more training or inference to in-house silicon for budget or power reasons, it can slow Nvidia’s pricing power and stretch replacement cycles. That does not end Nvidia’s dominance, but it shifts the slope of the cash flow curve just as dilution scrutiny intensifies. A small change in unit economics multiplied across massive deployments changes the present value of future repurchases. Investors betting on buybacks to keep per-share metrics on an escalator must watch whether margin and utilization hold as alternatives scale.
Three disclosures will frame the next leg for NVDA. First, the trajectory of diluted shares outstanding. If the share count does not start to compress, expect the Burry chorus to grow. Second, the cash flow allocation mix among capex, buybacks, and any incremental SBC. Upsized foundry and networking spend can be bullish for growth but leaves less room to repurchase. Third, hyperscaler commentary on unit cost per model trained and inference workloads by silicon type. If Google, Microsoft, or Amazon signal larger allocations toward internal accelerators for select use cases, the market will track what that means for Nvidia’s gross margin and price realization. Those datapoints will settle whether this pullback is a reset or the start of a longer multiple compression.
At current scale, one or two points of annual dilution materially changes EPS compounding, especially if revenue growth moderates from triple digits to something closer to market-beating but rational rates. If operating leverage flattens and the company continues to grant sizable equity at elevated prices, the repurchase math requires a bigger check to achieve the same per-share outcome. Nvidia says it has returned vast sums to investors and that compensation fuels innovation, which in turn funds the returns. Both can be true. But investors are paying for per-share growth, not gross growth. The spread between the two is what Burry is dragging into the light, and the market is listening after the recent slide.
Short-term, the stock is a tug-of-war between belief in Nvidia’s AI moat and concern that dilution, competition, and rising capex will slow per-share gains even if absolute profits rise. The company’s memo addresses the headline figures and contextualizes SBC, but it does not erase the arithmetic investors care about: shares outstanding, per-share free cash flow, and the price you pay for each. If Nvidia can show shrinking diluted shares, steady margins despite TPU pressure, and disciplined repurchases in the next two quarters, the buyback debate will fade. If not, expect a new factor in the NVDA playbook: a quality-of-earnings discount layered onto an AI leader that still delivers. That is why a single line in the cash flow statement now moves a trillion dollars of market cap.
