Nvidia's Growth Engine Faces Its Toughest Audience Yet: The Skeptics

Nvidia guides to 70% fiscal 2028 revenue growth, beating analyst estimates, but supply constraints, China absence, and depreciation debates loom.

When Jensen Huang steps onto the Goldman Sachs Communacopia + Technology Conference stage on Thursday, he will carry a rare commodity for a tech CEO: a forward-looking growth forecast that Wall Street still considers too conservative. The company has guided to roughly 70 percent revenue growth for fiscal 2028 — a figure that overshoots analyst expectations of 44 percent by a wide margin, yet one that CFO Colette Kress framed in strikingly understated terms during the latest earnings call. The ceiling, she said, is supply constraints, not demand.

That message lands at a moment when Nvidia's stock sits roughly 4 percent below its 52-week high of €202.50, reached in May. Shares closed Wednesday at €193.90, down 2.5 percent on the day — a pause that looks more like consolidation than correction after the post-earnings rally in late August. The stock remains up 21 percent year-to-date and 33 percent over twelve months.

The Numbers Behind the Ambition

The second quarter, which ended July 26, delivered revenue of $96.2 billion, up 106 percent year over year, with gross margins holding at 75 percent on both GAAP and adjusted bases. The company has guided to third-quarter revenue of $108 billion, plus or minus 2 percent. But the headline figure that will likely dominate Thursday's conversation is the fiscal 2028 projection — and notably, that forecast assumes zero data-center revenue from China, meaning the growth story is expected to unfold entirely without access to the world's second-largest economy.

What gives that projection credibility is the pricing power visible in the secondary market. Hourly rental rates for Nvidia's H100 graphics processors have climbed 22 percent to $3.28, suggesting even previous-generation chips remain in short supply. The company has announced price increases exceeding 15 percent on its chips effective early 2026, with Intel planning similar moves on its CPU lineup. The Rubin platform, whose production shipments have already commenced, is expected to contribute roughly one-fifth of data-center revenue in the third quarter.

An Ecosystem Takes Shape

The expansion strategy extends well beyond silicon. On September 3, Nvidia announced the acquisition of Hugging Face, the AI model and developer-tooling platform, for $12.93 billion — structured as approximately $11.9 billion in cash plus up to $1 billion in stock tied to employee retention. It marks the second-largest acquisition in company history, trailing only the $20 billion Groq deal from December.

A day earlier, Nvidia unveiled an expanded partnership with MediaTek, investing $3.5 billion in convertible bonds of the Taiwanese chip designer. MediaTek, in turn, will adapt Nvidia's NVLink-Fusion platform for its own processors. The pattern is consistent: rather than simply selling chips, Nvidia is embedding its technology across the semiconductor ecosystem's software and hardware layers alike.

Insider Selling Meets Institutional Buying

The corporate activity has coincided with notable insider transactions. Director Mark Stevens sold 1,848,501 shares between August 31 and September 2 through a trust, at prices ranging from $220 to $226 — a sale worth roughly $400 million. A separate SEC filing shows Stevens disposed of just over one million shares on September 3 for approximately $235.6 million, retaining about 28.9 million shares, of which roughly 11.5 million are held directly.

Such moves are routine for executives at companies of this scale and typically reflect long-term estate planning rather than bearish conviction. Stevens still holds more than 26.5 million shares directly and indirectly after the transactions. Institutional investors, meanwhile, have been adding: CV Advisors increased its Nvidia position by nearly 30 percent in the second quarter, with State Street and Geode Capital also raising their stakes. The picture among prominent investors is more mixed — David Tepper reportedly bought, while Dan Loeb is said to have exited entirely and George Soros has taken bearish positions.

The Skeptics' Case

Not everyone is persuaded by the growth narrative. Jim Chanos has questioned whether the six-year depreciation schedules that data-center operators apply to Nvidia hardware are economically justified, arguing that older chips like the A100 could lose value faster than assumed. Michael Burry has similarly suggested that industry-wide depreciation assumptions are set too low, pointing to the financing program Nvidia has built with Blackstone, BlackRock, and Goldman Sachs — a multi-billion-dollar credit facility that lets customers finance chip purchases. Huang's defense is that the processors themselves are collateralizable assets; his critics see circularity.

Customer concentration adds another layer of concern, with three clients accounting for 44 percent of revenue. The unresolved question of when Nvidia might resume regular shipments to China continues to occupy analysts.

What Thursday May Settle

Wall Street remains largely constructive despite these concerns. Cantor Fitzgerald calls Nvidia the "cheapest" compute stock in the market with a $350 price target. BMO's Hash Kumar maintains a buy rating and a $340 target, implying more than 50 percent upside, citing persistent shortages in memory chips and foundry capacity. The consensus target sits near $325.

Thursday also marks the dividend record date for the next quarterly payout of $0.25 per share. Whether Huang can deploy fresh arguments to address the depreciation and financing questions — or whether the debate over the durability of the AI investment cycle simply gains more momentum — will determine if the market continues to underwrite the company's unusually ambitious forecast.