Nvidia's $91.8 Billion Quarter Arrives With Memory Bottlenecks and a $500 Billion Korean Alliance in Tow

Nvidia reports Q2 earnings with $91.8B revenue forecast, but memory constraints and insider selling temper optimism ahead of August 26.

The countdown to August 26 has Nvidia investors juggling two very different narratives: a record-setting partnership pipeline that stretches into the hundreds of billions of dollars, and a nagging set of supply-chain and financing questions that refuse to go away. When the chipmaker reports fiscal second-quarter results for the period ended July 26, the consensus calls for roughly $91.8 billion in revenue and earnings per share of $2.08 — numbers that will test whether the market's enthusiasm has finally outrun the fundamentals.

The share price suggests confidence remains intact. At €190.84, the stock sits about 14.20 percent above its 200-day moving average, roughly 5.76 percent shy of the €202.50 52-week high touched on May 14. That proximity to a record heading into one of the year's most anticipated earnings reports cuts both ways: it reflects months of accumulated optimism, but it also leaves little room for disappointment.

The Memory Problem Nobody Can Ignore

The most immediate operational concern centers on Nvidia's next-generation silicon. Reports indicate the company is testing versions of its upcoming "Rubin Ultra" graphics processor with less memory capacity than originally planned, owing to tight supplies of high-bandwidth memory. The constraint is hardly unique to Nvidia — demand for AI hardware has outstripped what even specialized memory makers can deliver — but it undercuts the notion that the sector's dominant player operates free of supply-chain friction.

That memory squeeze makes the timing of the SK Group partnership all the more significant. The two companies unveiled plans this week for a collaboration valued at more than $500 billion aimed at building out AI infrastructure for global data-center demand. SK Telecom will construct a two-gigawatt AI factory equipped with Nvidia's forthcoming Vera Rubin platform, while SK hynix will co-develop the next generation of HBM memory chips. The first facility is slated to come online in 2027.

The deal follows a string of announcements that underscore how quickly the AI buildout is scaling. On July 29, Nvidia struck a long-term partnership with Safe Superintelligence to accelerate the AI startup's growth. CEO Jensen Huang has been characteristically bullish, describing Vera Rubin chips as "already in production" with "massive volumes" to follow.

Institutional Buying Meets Insider Selling

Positioning data from the first quarter shows institutional money flowing in. Lazard Asset Management increased its Nvidia stake by 22.9 percent, holding roughly 10.3 million shares valued at about $1.8 billion. Sculptor Capital also established a new position. Against that backdrop, Family Manage LLC's sale of 10,197 shares looks like noise rather than signal.

The picture from Nvidia's own executive suite is more complicated. Insider sales over the past three months total approximately $410 million, according to media reports — hardly unusual for a company trading near record highs, but noteworthy when paired with broader questions about how the AI boom is being financed.

Those questions resurfaced on July 31, when reports emerged that Nvidia was in talks with OpenAI about a $250 billion backstop to secure computing capacity for a US data-center project. The news revived concerns about circular financing — the practice of Nvidia effectively funding its own customers so they can buy Nvidia chips. The stock fell roughly five percent that day, dragging the broader chip sector down with it, and Nvidia ceded its position as the world's most valuable listed company to Apple.

Investor Michael Burry amplified those worries on Tuesday, pointing to a potential $250 billion demand risk in the same area covered by the OpenAI discussions. Yet analysts remain overwhelmingly constructive: 36 currently rate the stock a buy, with Bank of America Securities' Vivek Arya reaffirming a "Buy" recommendation on August 3. Bernstein's Varun Govindaraj and Wells Fargo both hold price targets of $315.

A Second Growth Track Takes Shape

Beyond the core chip business, Nvidia is quietly broadening its footprint. The "Build in America" initiative aims to expand AI infrastructure and manufacturing domestically with US partners, complemented by participation in the National Science Foundation's AI Hubs program. This week also brought the commercial release of "Alpamayo 2 Super," an open model for autonomous vehicles and robotaxis, ahead of October's GTC Berlin conference focused on AI infrastructure and industrial digitization.

China remains a structural drag: export controls have reduced the country's contribution to data-center revenue to essentially zero, according to a report from late June, a year after the region contributed billions. The shortfall has been absorbed by the sheer scale of US and Asian orders, but it leaves a persistent hole in the geographic mix.

The stock has recovered much of its late-July setback, closing Thursday at €189.82 — roughly nine percent above the prior week's level but still more than six percent off the 52-week high. The market, for now, appears to be weighting the growth story more heavily than the risks.

August 26 will settle the question of whether that weighting is justified. The numbers themselves are unlikely to disappoint — the more pressing issue is whether the customers financing these data centers can keep up with Nvidia's ambitions.