Germany Tagesschau
After the Huang affair: Everything new in the AI chip industry soon?
Competitors like AMD and Qualcomm are already rubbing their hands after Huang's departure as CEO of Nvidia. What is the potential upheaval in the chip industry - and how much does it actually have to do with it?
Deutsch (original) ·Auto-translated to English
After the Huang affair, everything new in the AI chip industry? As of: November 17, 2026 • 4:16 a.m
Competitors like AMD and Qualcomm are already rubbing their hands after Huang's departure as CEO of Nvidia. What is the potential upheaval in the chip industry - and how much does it actually have to do with Huang's departure?
The time of the quasi-monopolist is over, some now say. It will be easier for competitors to bring their AI chips onto the market. Because the face of the chip manufacturer is gone: Jensen Huang, apparently thrown out by the NVIDIA board of directors. In any case, Huang did not explain his departure himself. It was announced by the Board of Directors.
There has been competition for Nvidia for a long time. But the group has so far dominated the headlines with ever new record figures. Google has been developing its own AI chips for ten years - and is now announcing a major business deal.
Oracle wants to equip cloud infrastructure with Google chips from 2027. And OpenAI - so far only a tenant in Google's data centers - is reportedly planning to soon buy Google chips for its own data centers. In total, Google reports around 100 customers for its AI chips.
With the departure of Nvidia founder Huang, the company is losing its innovator. more
AI infrastructure specialist Ben Pouladian from the tech portal The Information finds it remarkable: "It's great how Google is now offering its chips to other companies. I never expected that. I'm really surprised. It's exciting to see how a software company becomes a real chip manufacturer."
Nvidia relies on versatile graphics processors with enormous computing power. Google's so-called TPUs, on the other hand, are special chips - more efficient and cheaper for certain tasks. Microsoft and Meta are also developing their own AI chips. Some of the newer specialty chips don't even need water cooling. An important argument for data centers, whose high water consumption always causes criticism.
Especially where huge AI models do not need to be trained, customized chips may be a better choice. Broadcom boss Hock Tan also said this at CNBC: "We develop chips together with AI companies. Their developers tell us which computing tasks they need to handle. And we design a chip that is tailored precisely to this. Such a special chip is significantly more powerful for these tasks than a universally applicable graphics processor."
China is catching up, corporations are building new factories: The risks for the semiconductor boom and AI rally are growing. more
After Nvidia, Broadcom even has a larger market share than AMD among the classic chip manufacturers. But the top dog still has a decisive advantage over both, says Ben Pouladian - because Nvidia works with the CUDA programming platform. That is a significant difference. "100,000 developers worldwide are continually improving it. There are also extensive software libraries for programmers."
Programmers and AI researchers began using these interfaces inside the chips around 15 years ago. In this way, developers can - to put it simply - easily and quickly set very complex tasks for the universal chips and then modify them again later. This has resulted in a lead that has so far been difficult to catch up with.
OpenAI orders AI processors worth tens of billions from US semiconductor manufacturer AMD. more
But that could also change - thanks to AI itself, says venture capital expert Umesh Padval at The Information: "The Claude chatbot is now so good at replicating CUDA that start-ups no longer necessarily need the Nvidia platform." So Nvidia's lead in the software ecosystem could shrink. Tailor-made chips are becoming more attractive. And AMD even expects to double its data center business in the coming year.
Is it all a result of Jensen Huang's mistakes and the forced change of boss at NVIDIA? Rather not. It became known on October 27th that Google wanted to sell its special chips to other companies in the future instead of just using them itself. The alleged espionage affair surrounding Jensen Huang only two days later. And such chip deals have a long lead time - often months, sometimes years. Ultimately, technology, prices, delivery quantities and long-term contracts must be clarified.
This topic in the program: tagesschau.de | Everything new in the AI chip industry soon? | 11/17/2026 | 01:05 am