Nvidia’s AI boom just ran into a new $96B question

Faizan Farooque

Sat, August 29, 2026 at 9:03 PM GMT+3 7 min read

NVDA -4.57%

Nvidia (NVDA) posted one of the best quarters in the chip giant's history.

Revenue reached $96.2 billion in the fiscal second quarter, up 106% from a year earlier and 18% from the previous quarter. Data center revenue alone climbed to $89 billion, a 117% year-over-year increase.

Nvidia maintained a 75% gross margin while producing GAAP earnings of $2.46 per diluted share.

Those figures help to partly explain the jump in Nvidia shares after the news and why investors are again more upbeat about the AI infrastructure cycle having legs.

But another development came nearly immediately after profits.

The Trump administration is mulling a wider set of semiconductor tariffs that could affect not only imported chips but also items based on those processors, including laptops, gaming consoles, and data-center servers, Reuters reported. The concept is still under discussion and could be drastically altered before anything is officially released.

That's an important distinction.

Until now, there has been no revised tariff system.

What investors do have is a semiconductor corporation generating roughly $100 billion in quarterly revenue, a booming data-center sector, and possibly changing trade costs that might impact the gear needed to support that growth.

Nvidia's numbers show why semiconductor costs suddenly matter more

The size is essential for Nvidia since the tariffs on sophisticated processors are no longer a narrow semiconductor industry concern.

In the latest quarter, NVIDIA's Data Center division made up 92.5% of overall company sales. That means the company's fortunes are heavily related to the servers, networking systems, and computer infrastructure being built by cloud providers, AI labs, and huge companies.

The company's growth rates remain extraordinary:

Total quarterly revenue: $96.2 billion, up 106%
Data Center revenue: $89.0 billion, up 117%
Sequential revenue growth: 18%
GAAP gross margin: 75%
GAAP earnings per share: $2.46
Non-GAAP earnings per share: $2.22
- Source: Nvidia press release

CEO Jensen Huang said demand is accelerating as more AI labs, startups, and companies build AI systems.

That is the commercial backdrop for any wider semiconductor tariff to try and work.

The policy is so much more narrow now.

The White House levied a 25% levy on select advanced computing chips in January, explicitly mentioning Nvidia's H200 and AMD's MI325X. But the declaration made important exceptions, including chips imported for use in U.S. data centers, research and development, startups, and some consumer and industrial uses.

Thus, the present structure was partly meant to avoid adding costs to the very data-center building that is helping to fuel Nvidia's expansion.

Nvidia, AMD, and Intel face very different exposure

The tariff talk could affect semiconductor companies differently depending on where they manufacture their products.

The company relies largely on overseas foundries to produce its most powerful processors. That style of outsourcing has allowed the corporation to put capital to the chip design and software, rather than constructing its own leading-edge fabrication operations.

So the potential tariff concern is simple: where a chip is made, where it is imported, and how the client uses it could increasingly affect its economics.

The January White House move identified powerful computing chips from Nvidia and Advanced Micro Devices (AMD) as examples of chips that would face a 25% duty if they don't qualify for an exemption.

Intel (INTC), in contrast, maintains its own production network and has been pouring money into expanding U.S. fabrication capacity.

That doesn't automatically make Intel a winner from the tariffs. Semiconductor supply chains are very complicated, with chips, production tools, packaging materials, and other components traversing borders many times.

But a system that provides preference to corporations investing in U.S. manufacturing could change relative costs in the sector.

The White House has signaled that path already.

The January declaration outlined a potential second phase that would link larger semiconductor duties with a program to offset the tariffs for companies expanding in U.S. chip manufacture.

And that makes this week's study less of a brand new policy concept and more a potential extension of a framework laid forth months before.

Data-center servers are where the numbers become much larger

But the most important piece of the idea may not be the tariff on any single processor.

Perhaps it is the risk of tariffs hitting the systems built around such CPUs.

Modern AI servers mix together GPUs or accelerators with high-bandwidth memory, CPUs, networking devices, storage, power systems, and cooling infrastructure.

One server can be assembled with components from several nations.

This is important because now Nvidia no longer just sells stand-alone GPUs. Its expansion is increasingly coming from larger computing platforms and networking systems meant to serve as integrated AI infrastructure.

More Nvidia:

And the money coming into those systems has had to be significant.

Nvidia's $89 billion in quarterly Data Center revenue is already nearly double the company's entire quarterly revenue of $46.7 billion one year earlier. That scale of increase is what makes changes in the cost of data-center equipment relevant to investors.

Even small percentage increases become relevant when cloud enterprises are ordering on this scale.

The existing 25% tariff contains an important clue

The January policy is a great yardstick to measure what might be next.

The White House slapped a 25% levy on the chips following the Commerce Department's probe of the U.S. reliance on imported semiconductors, which it determined was a national-security threat. The administration said U.S. chip output isn't enough to meet domestic demand and bigger levies could follow.

But the exemptions were sweeping.

Among the excluded uses were:

U.S. data centers
U.S. research and development
Startups
Certain consumer applications
Certain industrial applications
Public-sector uses

The exemptions greatly reduce the direct impact on U.S. AI infrastructure.

That is why the headline term "tariff" matters less than the extent of any new structure.

If data-center exemptions continue to be widespread, Nvidia's core U.S. demand could stay relatively shielded.

Taking away or limiting such exemptions could create a totally different pricing structure.

MarketWatch noted the plan under consideration could remove the previous data-center exception, which would be particularly pertinent to Nvidia, as most of its main chips are built overseas.

But nothing has been set in stone, according to Reuters, and the framework might yet be drastically changed.

Nvidia's AI surge could be heading for a costly new test.Bloomberg / Getty Images

Nvidia investors now have 2 numbers to watch

For stockholders, the tale comes down to two figures.

First, there's the expansion of Nvidia.

The company just posted $96.2 billion in revenue, $89 billion in Data Center sales, and 106% overall year-over-year growth. Those figures suggest AI infrastructure demand remains exceptionally strong.

The second is the eventual tariff rate and, more importantly, what products and uses are exempt.

A 25% tariff that exempts U.S. data-center installations is fundamentally different from a broad 25% charge that hits servers containing imported chips.

The first can live with the existing AI buildout rather happily.

Device makers, cloud businesses, and their suppliers would have to decide how much of the additional cost they absorb and how much gets passed through.

That's the business problem to monitor.

Nvidia has already shown that customers are willing to spend tens of billions of dollars every quarter on AI infrastructure.

What is not clear is how elastic that demand would be if the cost of the hardware were to increase.

This story was originally published by TheStreet on Aug 29, 2026, where it first appeared in the Investing section. Add TheStreet as a Preferred Source by clicking here.