Nvidia's Next Strategic Power Move is Now Visible
The Nvidia result was dominated by the impact of China's embargoes, but could not hide the new development.
The new US embargoes dominated the Nvidia call. The lucrative H20 processors, developed to circumvent the embargo, were far too visible for the Trump administration to overlook and impacted the Q2-25 numbers.
Jensen Huang spent a significant amount of his time arguing against the embargos, highlighting that they would protect the Chinese GPU manufacturers, which would generate approximately $8 billion worth of revenue per quarter.
Embargoes never work as intended, but they have an impact. For Nvidia, the revenue impact in the quarter was $2.5 billion in embargoed revenue loss and $4.5 billion in material losses that impacted profitability.
The embargo's impact changed the quarterly growth significantly, from 18.3% to 12.3% and limited an otherwise stellar quarter.
The charges drove the profits down quarter-over-quarter (QoQ) for the first time in three years, but this is temporary, as Nvidia has already cleared the H2 shelves.
While whining about the new China embargoes with one tongue, the other one praised Trump’s leadership in rescinding the AI diffusion act that Biden had enacted, dividing the world into tiers. It is a complex game to navigate, as everything Nvidia does to circumvent the new rules will be the target of the next round of embargoes.
Apart from the political quagmire, Jensen Huang highlighted some positive developments during this year.
AI diffusion: the widespread, accelerating adoption of AI as fundamental infrastructure, akin to electricity or the internet, rather than a mere technological curiosity. He suggests this is catalysed by a global awakening and policy shifts purportedly designed to ensure American technology stacks form the world's AI backbone.
Enterprise AI: Driven by agentic AI, intelligent systems capable of understanding ambiguous tasks, problem-solving, and using tools. NVIDIA's move to offer integrated compute, storage, and networking solutions for this market, like the RTX Pro Enterprise server, speaks volumes about the anticipated scale.
Industrial AI: Set to transform the physical world, driven by trends like onshoring and the rise of smart factories. Huang envisions "AI factories" within every physical plant, technologies such as leveraging the tech Omniverse and robotics.
Despite the China embargo issues, Nvidia remains very upbeat.
From a segment and divisional perspective, there were limited developments, and the Compute segment and the Datacenter division are still the engines of Nvidia’s growth.
Gaming had a record quarter, also driven by Blackwell in graphics cards, AI PC’s and gaming consoles. This is one of the first indications of the adjacent areas that Nvidia is beginning to penetrate, and a demonstration that AI GPUs are not limited to the data centre.
From a product perspective, things are about to get interesting again.
The networking products that had been flat for five quarters jumped to a new record, increasing their share of revenue from 7.7% to 11.3%. This is the first revelation of Nvidia’s next strategic move. What sounded like a product introduction at Computex is the next strategic power move by Nvidia, and it is already well underway. I am going to cover this strategic move in the advanced analysis later in this post.
The Outlook
As a result of the China embargoes, the guidance of 45B$ was weaker than what we have seen the last couple of years from Nvidia.
The impact that will drive the Nvida China revenue close to zero will be 8 B$ in Q3-25 up from 2.5B$ in Q2-25
A modest 2.1% increase in revenue that could have been a 14% quarter over quarter increase.
While an optimistic CEO is not sufficent evidence for me, it still looks like Jensen Huang has something to support his happy face.
Despite low CapEx growth in Q1, the forecast of the top 4 cloud companies remains strong and there are several other large projects underway like Stargate, Apple and the arabian petro projects that evertually will settle into something more tangible.
Adding Enterprise investments and industrial AI to the equation gives Jensen the right to be optimistic. There are no signs of AI cooling yet.
Deeper analysis
There were several other interesting developments in Nvidia’s numbers. The first is the development in CapEx. While this is not usually an interesting number to investigate for Fabless companies unless you are interested in the number of desks and computers owned, it is becoming significant for Nvidia.
The CapEx is now at close to 1.3B$ or more that the large hybrid companies that need to invest in fab capacity.
The Nvidia capex is now larger than the combined capex of all the other fabless semiconductor companies.
This is not only equipment for the added headcount. Nvidia is up to something else.
Every once in a while, I try to ask a question in the investor calls, as should be possible, but somehow Goldman Sachs, UBS, Bank of America and Wells Fargo seem to get in front of me (I feel ok again - my ego is only slightly bruised). While these analysts typically ask good questions, they are also paid to have their corporate projection models filled so they can estimate next quarter’s numbers.
A more curious (translated to waste in corporate) question would be: What is Nvidia building for 1.3 B/qtr? I have not heard a question about that in the Nvidia calls.
A wild guess is that Nvidia builds something related to AI, using their servers. Fun fact is that Nvidia can use their servers for a quarter of the price a cloud company has to pay. If it is server-based, the CapEx investment would equate to over 5 B a quarter or more than Apple, Oracle, Alibaba, and Tencent are spending.
There are a couple of other interesting things going on: