Analyse von fast hundert AI - Investitionen von NVIDIA: Was ist die Gegenwart und Zukunft der Künstlichen Intelligenz - Branche?

NVIDIA is undoubtedly the vane of AI. It also builds an ecosystem on a larger scale through investments. In 2024 and 2025, it invested in nearly 100 startups in various tracks of the AI field in total.

NVIDIA invested in a large number of AI model companies to lock in customers and foresee the trend of the next - generation computing power. Secondly, in addition to its own powerful network capabilities, NVIDIA expands the efficient collaborative computing power of its GPUs to higher speeds and broader scopes by investing in next - generation network chips. Moreover, NVIDIA has made a long - term layout for embodied intelligence in its investments. Analyzing its investment behavior is of profound significance for our investment and entrepreneurship.

Invest in AI models and cloud platforms to build an AI ecosystem centered around GPU/CUDA

NVIDIA's core advantage in the AI industry lies not only in its GPUs but also in its ability to lock in the global developer ecosystem through CUDA. With the breakthrough of large language models, NVIDIA first invested in almost all cutting - edge AI model companies, including language models and multimodal models.

Model companies

OpenAI

NVIDIA participated in a round of financing of OpenAI worth up to $6.6 billion in October 2024, investing $100 million in this financing. In September 2025, NVIDIA and OpenAI announced plans to jointly deploy up to 10 gigawatts of NVIDIA's AI computing power systems, and NVIDIA promised to invest up to $100 billion in this cooperation.

As the leader in the AI industry, OpenAI not only has state - of - the - art language models and visual models but also created ChatGPT, a super - application at the "entrance" level in the AI era.

xAI

NVIDIA had previously participated in several rounds of financing of xAI, such as two rounds of $6 billion financing in 2024 (Series B and Series C). In September this year, it participated in a financing round of xAI worth up to $20 billion, with an investment of $2 billion.

Jensen Huang said, "I want to invest in any company that Elon Musk is involved in." In addition to the investment, NVIDIA has a long - term cooperation relationship with xAI, including providing H100 GPUs for xAI's supercomputer Colossus.

As a company founded in July 2023, xAI launched Grok4, a model that can reach the state - of - the - art level, within just two years and became a leading company in the industry.

Mistral AI

NVIDIA invested in the Series B and Series C financings of Mistral AI in 2024 and 2025 respectively. After obtaining €1.7 billion in Series C financing, Mistral AI's valuation reached €11.7 billion.

In June 2025, Jensen Huang announced a cooperation with Mistral AI to build AI computing infrastructure in France and deploy the latest generation of NVIDIA Grace Blackwell systems. The initial plan is to deploy 18,000 Blackwell GPUs.

Mistral AI is the most powerful AI model developer in Europe. Models such as Mistral Large are close to state - of - the - art performance. Its open - source models have been downloaded more than one million times, and its ToC applications are also quite popular.

Runway

In April 2025, NVIDIA participated in Runway's Series D financing of $308 million. Runway's post - investment valuation exceeded $3 billion. Runway is one of the pioneers of visual multimodal models. Currently, it has high - performance visual models and visual design products for professional fields such as Runway Studios.

In addition to the above investments, NVIDIA's investments in the core field of AI models also include Cohere (mainly targeting enterprises), Sakana AI (mainly researching small models), and more exploratory and innovative companies such as Reflection AI, Thinking Machines Lab, and Imbue.

The core logic behind its investment in these cutting - edge model companies is that the greater the number and the stronger the performance of cutting - edge models, the more choices consumers and customers will have, and the more stable the foundation of the entire AI application ecosystem will be. NVIDIA has locked in two key roles: model companies and partners building AI computing power platforms.

AI cloud platforms

In addition to cutting - edge model companies, another important area in NVIDIA's investment portfolio is AI cloud platforms with advanced technologies and data center operators, which provide computing power for companies in the more downstream AI application layer and further expand the scale and scope of NVIDIA's ecosystem.

The most typical companies in this regard are CoreWeave, Together AI, and Nscale.

CoreWeave

NVIDIA participated in the Series B financing of the AI cloud platform CoreWeave in 2023, injecting $100 million. It also participated in the subscription when CoreWeave went public in 2025. Currently, NVIDIA holds about 7% of CoreWeave's shares. After going public, CoreWeave's market value has reached about $70 billion, and its estimated revenue in 2025 is $5.008 billion. It has 250,000 NVIDIA GPUs. It is both an investee company of NVIDIA and one of its major customers.

Together AI

Together AI is also an AI cloud platform. It received a Series B financing of $305 million with NVIDIA's participation in 2025, and its valuation was $3.3 billion.

Different from CoreWeave, it is more technology - oriented and has its own core technologies, FlashAttention - 3 and Flash - Decoding, corresponding to fast inference and long - sequence inference scenarios respectively, bringing differentiated experiences to enterprise customers and developers and expanding the technical boundaries of NVIDIA GPUs in inference.

Nscale

Nscale's main business is the construction of high - performance data centers and GPU clusters. In September 2025, it received a Series B financing of $1.1 billion with NVIDIA's participation (NVIDIA invested $683 million).

Its data centers are equipped with large - scale NVIDIA GPU clusters and integrate high - speed networks and self - developed software stacks, supporting the efficient processing of training tasks for large language models (LLMs) with hundreds of billions of parameters. Its main customers include OpenAI (co - building the Stargate project) and Microsoft.

NVIDIA also participated in the investment of AI cloud platforms and data center construction companies such as Crusoe, Lambda, and Firmus Technologies. The valuations of these companies have reached the level of billions of dollars.

Innovative chip companies

NVIDIA builds a powerful GPU - based computing power through scale - up (stacking the number of chips on a single board) and scale - out (high - speed interconnection between different servers). Relying on its powerful network communication capabilities, it provides high - efficiency cluster computing power and has completely occupied the market of traditional server manufacturers.

Moreover, NVIDIA has also invested in network chip companies to further strengthen the layout of its intelligent computing centers, including Ayar Labs and Enfabrica. The main function of such network chips is to expand beyond NVIDIA's existing network "range" and enable high - speed and high - bandwidth communication between computing power chips to improve the network throughput capacity of computing power clusters in a larger scope.

Ayar Labs

The core technology of Ayar Labs is an optical interconnection solution that combines silicon photonics with chip packaging. Its optical I/O Chiplet can achieve a two - way bandwidth of 8.192 Tbps per chip (16 modules × 512 Gbps/module), with a bandwidth density of 896 Gbps/mm and an end - to - end latency < 25 ns (including fiber - optic transmission). NVIDIA participated in its Series C financing in 2023 and Series D financing in 2024.

Enfabrica

The core technology of Enfabrica is the Accelerated Computing Fabric (ACF) and the Elastic Memory Architecture. Its ACF - S super network card chip integrates PCIe/CXL switching and Ethernet RDMA functions, supporting the interconnection of hundreds of thousands of GPUs. The CXL memory pool solution of the ACF - S chip alleviates the bottleneck of high - bandwidth memory (HBM) of GPUs by sharing DDR5 memory.

NVIDIA has a great preference for this company. It not only participated in its Series B and Series C financings in 2023 and 2024 but also acquired Enfabrica's core team and technology license for $900 million in 2025. Rochan Sankar, the founder of Enfabrica, also joined NVIDIA.

In summary, NVIDIA has built and consolidated its own ecosystem in the AI field by investing in AI model companies, AI cloud platforms, and innovative chips around GPUs. Through these investment portfolios, it has actually become the largest and most solid foundation in the AI industry.

Invest in Physical AI to prepare for the third wave of AI

Judging from the fields where it has made real - money investments, the most promising development direction of AI in NVIDIA's eyes is Physical AI.

Of course, NVIDIA has also invested in many vertical applications in the AI field, such as Perplexity (AI search engine), Hippocratic AI (AI healthcare), Poolside (AI programming), and Kore.ai (enterprise AI). However, for Physical AI, NVIDIA not only made investments but also got involved directly.

It has its own Cosmos world foundation model, an open physical AI dataset, the Omniverse physical AI operating system, as well as the DRIVE Thor autonomous driving chip and the NVIDIA Jetson edge computing platform.

Jensen Huang believes that Generative AI is the second wave of AI, and the next wave will be Physical AI because it goes beyond the digital world and enables AI to interact with the real physical world, understanding physical laws, friction, inertia, and causal relationships.

Of course, NVIDIA has mainly invested in a number of companies related to Physical AI, such as Figure AI (humanoid robots), Dyna Robotics (commercial robots), Serve Robotics (delivery robots), Wayve (autonomous driving), Nuro (autonomous driving), Bright Machines (smart factories), etc.

Figure AI

Figure, a humanoid robot company, received a Series C financing of $1 billion with NVIDIA's participation. It is a full - stack embodied intelligence company, having both humanoid robot bodies (it recently released Figure 03), its own robot models, and manufacturing capabilities. It is a super - unicorn with the most comprehensive capabilities among robot startups.

Wayve

Wayve received a Series C financing of $1.05 billion with NVIDIA's participation in 2024. In the second half of 2025, there were reports that NVIDIA would make a strategic investment of $500 million in Wayve.

Wayve's core product is an end - to - end neural network - based autonomous driving system. It does not rely on high - precision maps and only uses data to teach vehicles how to drive. This data - driven learning method is applicable to the full range of functions from "visual assisted driving" to "hands - off fully autonomous driving" (Level 4).

Bright Machines

Bright Machines received a Series C financing of $126 million with NVIDIA's participation in June 2024. Bright Machines provides full - stack automation solutions for manufacturing companies, completely changing the existing production model. This solution can provide flexible automation support for the entire life cycle of products from design, manufacturing to disassembly.

AI models are the present, Physical AI is the future

Looking at NVIDIA's nearly 100 AI investment projects, we can see that they still bear a strong mark of corporate strategic investment, especially the investments in models and AI cloud platforms. These investee companies are generally also buyers of NVIDIA GPUs. NVIDIA's investment in these companies is to build an ecosystem centered around GPUs, and this ecosystem is to consolidate its position in the AI industry.

Investing in companies in the Physical AI field is also to build an ecosystem, but this ecosystem belongs to the future and is to establish a new growth pole for NVIDIA.

Why is Physical AI or AI - native hardware the next wave of AI? As Jensen Huang said, it goes beyond the digital world and enables AI to interact with the real physical world, laying the foundation for AI to reshape the physical world in the future.

This article is from the WeChat official account "Alpha Commune". The author is Xu Siqing, the founding partner of Alpha Commune. It is published by 36Kr with authorization.