Nvidia reports $96,2 billion in revenue and is targeting $108 billion in the next quarter.

The chipmaker on Wednesday posted quarterly results significantly exceeding Wall Street expectations, driven by a surge in data center revenue. CEO Jensen Huang believes artificial intelligence has reached a turning point.

Nvidia posted $96,2 billion in revenue for the second quarter, compared to analysts' expectations of $92,2 billion: this time, Nvidia once again surpassed forecasts. Revenue from data centers surged 117% to $89 billion. For the following quarter, the company anticipates $108 billion in revenue, again exceeding consensus estimates.

“AI has reached its inflection point. It is doing useful work. Its tokens are productive and profitable. Now, computing power is generating revenue and demand is accelerating,” said Jensen Huang in a statement accompanying the results.

Despite these figures, the stock fell 1,8% in after-hours trading, having already lost 1,6% during the regular session. This scenario played out before after the first-quarter results, when the stock dropped nearly 5% in the following days. The dynamic is well-known: to boost the share price, Nvidia must not only beat an already optimistic consensus but also deliver stronger-than-expected guidance.

With a market capitalization of over $5 trillion, Nvidia is worth more than the GDP of Japan, the world's fourth-largest economy. This dominant position, however, rests on a concentrated customer base: Amazon, Google, and Microsoft account for the bulk of its revenue, even though these three companies each develop their own chips to reduce their dependence on the Californian manufacturer.

To further integrate itself into the ecosystem it powers, Nvidia has chosen to participate directly in infrastructure financing. This month, the company formed a $500 billion capital pool with six Wall Street management firms for data center projects. It has also committed up to $105 billion to an OpenAI data center in Pike County, Ohio, with an initial capacity of 4,25 gigawatts and an option for an additional 3,75 gigawatts.

The group's current growth is driven by its Blackwell chips, which now power the majority of AI-dedicated data centers. Their successor, named Vera Rubin after the American astronomer whose work on galaxy rotation provided some of the first solid evidence for the existence of dark matter, is scheduled to begin shipping in the second half of the year. Nvidia has a tradition of naming its architectures after scientists, following Ampere, Hopper, and Blackwell.

On the geopolitical front, Washington initially banned the sale of the H20 chip designed for China in April 2025, before reversing its decision. Nvidia has since obtained authorization to supply the higher-performing H200 chip to pre-approved Chinese customers. Reports indicate significant allocations to ByteDance and Tencent, even though Beijing encourages local companies to prioritize domestic solutions. The $108 billion forecast for the next quarter excludes China.

The group highlights an order book that it estimates at around $1 trillion for 2026 and 2027, a figure derived from its own statements and not from independently verified financial documents.

Nvidia's earnings release came during a session marked by slightly better-than-expected inflation data. The Federal Reserve's key consumer spending index rose 0,2% in July, compared to the anticipated 0,1%, maintaining the annual rate at 3,7%. Core inflation remains at 3,3% year-over-year, above the Fed's 2% target for the 65th consecutive month. The FOMC kept its interest rates in a range of 3,50% to 3,75% in July, with three regional Fed presidents voting for a quarter-point increase. Markets are pricing in a roughly 40% probability of a rate hike in September.