AI chips are booming with $148 billion, but this sector has stalled: Nvidia's four-year bet on manufacturing yields little return.

```

Although Nvidia raked in nearly $148 billion in revenue from its AI chip business in the nine months ending in October—far surpassing the $27.5 billion in the same period of 2023—the company’s critical attempt to transform into an integrated hardware-software platform has suffered a major setback.

As CEO Jensen Huang’s core bet to enter the $50 trillion manufacturing and logistics markets, the Omniverse software business has seen minimal returns after four years of heavy investment, with commercialization severely lagging behind.

According to sources quoted by media, due to “virtually nonexistent” demand since its launch in 2022, Nvidia shut down the Omniverse Cloud service in August 2025. Previously, the company had spent hundreds of millions of dollars renting thousands of GPUs from Oracle, Google, and Microsoft to support the project, but for most of the time, lacking external customers, the company had to look internally for ways to utilize the chips to avoid idling.

Although Jensen Huang is still heavily promoting the business at this week’s CES consumer electronics show, portraying it as a trillion-dollar “physical AI” opportunity, he is reportedly frustrated internally at the department’s slow progress. Media cited sources stating that Huang frequently criticized the team for wasting engineering resources on “demos” rather than “products,” and that the business has not seen the widespread adoption by large enterprises as expected.

Thanks to the explosive growth of the AI chip business, the market has so far not reacted strongly to the revenue shortfall of Omniverse, but this exposes the tough challenge faced by this hardware giant in establishing a second growth curve. Unless it solves problems with software usability and industry adaptation, Nvidia’s ambitions in robotics and industrial digitization could face long-term shelving.

Demand Collapse and Cloud Service Shutdown

Nvidia formally launched Omniverse in 2021, positioning it as a platform for designers to collaborate on 3D designs in virtual spaces, and aimed to expand it as a core tool for building “digital twins” of cars, factories, and robots. Jensen Huang envisioned companies using simulation before physical manufacturing, driving billions of dollars in chip purchases and up to $4,500 annual software license fees per chip.

However, reality falls far short of the vision. According to former Nvidia employees quoted by media, although the company listed clients including BMW, Siemens, Foxconn, and Boston Dynamics, very few signed up to use Omniverse Cloud servers for large-scale simulation. Software developers widely reported the platform as “hard to use,” incomplete in functionality, and prone to crashing. Developer Valentin Forager said that the system crashed directly when attempting to simulate human activity in virtual environments: “If you try anything beyond its preset scope, it just breaks.”

Additionally, the platform’s scene creation tools are complex to operate and documented poorly, making many technical issues hard to fix. At an event last November, a Nvidia representative even admitted the software was not ready to meet specific needs, and suggested customers use competitor Unity’s software instead. This lack of product maturity directly led to the termination of the cloud service project.

Internal Anxiety and Management Pressure

Although Omniverse’s revenue is negligible within Nvidia’s massive financials, Jensen Huang has shown obvious anxiety about its performance. According to people familiar with his thinking quoted by media, he has long worried that competitors would seize the initiative, thus vigorously pushing the company to find new revenue sources.

This frustration has flared up multiple times in internal meetings. Attendees quoted by media said that at an all-hands meeting in 2023, Huang openly questioned Omniverse VP Rev Lebaredian about whether the team had turned a profit. In another call, facing requests to expand the team for new product development, Huang angrily reprimanded them for nearly an hour, accusing the team of wasting engineers’ time on demos and stressing that the current headcount was already sufficient.

The actual results of collaborative projects have also triggered management dissatisfaction. Media cited insiders saying Huang lashed out when he discovered that the scale of partnership announced with BMW was much smaller than he had initially expected. Although Nvidia executives claimed BMW had over 20,000 planners using the software globally, they did not clarify how much that translated into actual sales.

Long-term Bets Face Industry Barriers

Facing skepticism, Nvidia executives compared Omniverse to CUDA—which required over a decade of investment to fundamentally change the deep learning field.

Rev Lebaredian said Omniverse is the foundational software for unlocking the huge “physical AI” market, and its long-term vision is gradually paying off through adoption by robot and automotive companies. To promote ecosystem development, Nvidia not only brought Huang’s children into the team, but also invested in companies such as Synopsys (which recently acquired Ansys) and MetAI, aiming to integrate semiconductor design and industrial simulation tools.

However, the sector faces fierce competition and structural barriers. In robot simulation, Unity Technologies and open-source simulator Gazebo are strong rivals. More crucially, many large firms such as Tesla prefer developing in-house simulation software instead of relying on Nvidia’s general-use platform.

Industry-specific technical thresholds are another major challenge. Two co-founders of robotics startups noted that Nvidia’s Isaac Sim tool is not practical for handling complex objects like clothing, which constantly change shape. Cost-effectiveness is also a hurdle: Dan Cole, COO of industrial automation equipment design firm Loupe, bluntly stated that renting cloud servers is not cost-effective.

As MetAI CEO Daniel Yu put it, Omniverse is currently just a horizontally open platform for developers to build on, not a complete application. This means Nvidia’s attempt to create a market from scratch still needs a long incubation period.

Risk Warning and DisclaimerThe market has risks; investment requires caution. This article does not constitute personal investment advice and does not consider the specific investment goals, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their particular situation. Investing based on this article is at your own risk. ```