Nvidia reports record $46.7B revenue fueled by Blackwell

Nvidia just dropped another masterclass in riding the generative AI wave, proving that the tech industry’s appetite for silicon is still far from satisfied.

The world's most valuable company reported a staggering $46.7 billion in revenue for the second quarter, up 56% year-over-year. Net income was even more eye-popping at $26.4 billion, surging 59% compared to the same quarter last year. Unsurprisingly, the bulk of that cash came from data centers, which pulled in $41.1 billion.

Blackwell takes center stage

The real star of the show was Nvidia’s newest architecture, Blackwell. The advanced chip lineup accounted for $27 billion in sales, quickly establishing itself as the default engine for frontier AI.

CEO Jensen Huang declared that Blackwell is the platform the world has been waiting for, adding that the AI race is officially on with Blackwell at its core. Huang also doubled down on his long-term optimism, telling analysts that spending $3 to 4 trillion on global AI infrastructure over the next five years is fairly sensible.

Nvidia also highlighted its recent role in powering OpenAI's newly released open source gpt-oss models. According to the company, a single Nvidia Blackwell GB200 NVL72 rack-scale system processed a blistering 1.5 million tokens per second during the rollout.

The China conundrum and a 15% tax

Despite the record numbers, Nvidia's earnings laid bare its messy geopolitical entanglement with China. In the second quarter, Nvidia shipped zero H20 chips—its tailored processor for the Chinese market—to customers in China. It did, however, offload $650 million worth of H20 units to a buyer outside the country.

While strict U.S. export controls have long curtailed advanced GPU sales, things took an unusual turn under President Trump. Nvidia was granted permission to sell chips to China under the condition that it pays a 15% export tax directly to the U.S. Treasury, an unconventional arrangement legal scholars have questioned.

Nvidia CFO Colette Kress explained that the zero shipments were due to the lack of official federal codification of that deal. Even though a few Chinese customers obtained licenses recently, Nvidia has held off on shipping. On top of that, Chinese authorities have actively discouraged local firms from buying Nvidia hardware, reportedly prompting Nvidia to halt H20 production earlier this month.

Still, Nvidia expects third-quarter revenue to hit roughly $54 billion, a forecast that notably assumes zero H20 shipments to China.

Why it matters

Nvidia’s blowout quarter proves that frontier AI scaling is still accelerating, regardless of macroeconomic skepticism.

For enterprise builders and AI developers, Blackwell's blistering 1.5 million tokens-per-second rack performance makes deploying heavyweight open source models, like OpenAI's gpt-oss, far more viable at scale. The bottleneck is no longer software architecture, but sheer access to Nvidia's hardware allocation.

We think Nvidia's ability to issue a record $54 billion forecast while writing off its China H20 shipments is an incredible flex of monopolistic pricing power. Western hyperscalers are absorbing every single die Nvidia can print, rendering export restrictions financially painless for now.

However, the long-term risk is brewing beneath the surface. Forcing Nvidia into a bizarre 15% export tax deal while Beijing tells domestic tech giants to look elsewhere is accelerating China's domestic chip ecosystem. Huang might still be selling every chip he makes today, but Nvidia looks poised to permanently lose one of the world's largest computing markets tomorrow.