RTX PRO 6000 goes to the cloud! Google partners with NVIDIA to build a cloud platform covering AI GPU computing power to physical AI

According to the Zhitong Finance APP, Google's cloud platform, under the American tech giant Google (GOOGL.US), announced on Monday Eastern Time that Google Cloud G4 VMs (i.e., G4 VM virtual machines) are officially fully commercially available. This product is supported by the high-performance GPU of the RTX PRO 6000 Blackwell server version, developed by "AI chip leader" NVIDIA (NVDA.US), for a complete ecosystem from industrial and enterprise-level AI application software development to deployment, covering a broader range of enterprise-level physical AI workloads including visual computing and digital twins.

It is reported that the Google G4 series products also utilize the EPYC Turin server-level CPU platform developed by AMD (AMD.US). NVIDIA and Google stated that the newly launched NVIDIA RTX PRO 6000 Blackwell GPU performs exceptionally well in high-performance AI inference for multimodal, generative AI, and agent-based AI deployments, while also providing significant support for complex visual and industrial simulation workloads ranging from computer-aided engineering and complex content creation to robotic simulation.

In addition, the two digital twin and robotic simulation platforms launched by NVIDIA, Nvidia Omniverse and Nvidia Isaac Sim, are also available to users of the Google Cloud platform through the Google Cloud Marketplace in the form of virtual machine images. NVIDIA stated that these software and hardware tools are very helpful in providing significant power for physical workloads and related applications driven by breakthrough AI technologies in manufacturing, automotive, and logistics industries.

“G 4 VM has achieved a milestone leap in comprehensive performance on the cloud platform, with throughput up to 9 times that of the previously launched G2 platform instances, bringing phased improvements in a wide range of physical AI workloads, including multimodal AI inference, realistic design and visualization that closely approaches reality, as well as robotic simulation and modeling using large applications developed based on NVIDIA Omniverse,” Google stated in a blog post.

The company claims that it enables users of cloud computing services to significantly enhance the workload efficiency of generative AI applications such as multimodal and text-to-image generation models, as well as agent-based AI agents. It has also greatly improved the specific time required for phased training, fine-tuning, and AI inference of large language models.

NVIDIA stated in a blog that the core of the new G4 VM is the NVIDIA RTX PRO 6000 Blackwell server version GPU, which is the ultimate data center GPU for workloads such as digital twins, simulation, and visual computing. Its unique design combines two powerful engines: the fifth-generation Tensor Core achieves a significant leap in comprehensive AI performance, supporting new data formats such as FP4, achieving a stronger energy efficiency ratio with lower memory usage; the fourth-generation RT core provides more than twice the real-time ray tracing performance of the previous generation, achieving graphics and realistic simulations beyond movie levels As part of Google Cloud AI's supercomputer model on the Google Cloud Platform, the G4 virtual machine natively integrates services such as Google Kubernetes Engine and Vertex AI, significantly simplifying containerized deployment and the machine learning operations of physical AI workloads. This flexibility also extends to using Dataproc to accelerate large-scale data analysis on Apache Spark and Hadoop.

Google collaborates with NVIDIA to create a cloud platform covering AI large model training/inference computing resources to complete physical AI systems

NVIDIA pointed out in a blog post: "These latest announcements related to our collaboration with Google establish a complete end-to-end computing platform built on the Nvidia Blackwell supercomputing platform—covering everything from Nvidia GB200 NVL72 and Nvidia HGX B200 for large-scale AI training and inference, to RTX PRO 6000 Blackwell for AI inference/fine-tuning/evaluation/distillation and visual computing workloads on G4 VM."

Before the G4 was released, Google Cloud had already provided a large AI computing platform that directly accessed NVIDIA B200/GB200 (NVL72) computing clusters—corresponding to the A4 (B200) and A4X (GB200 NVL72) models. The positioning of the G4 VM based on RTX PRO 6000 Blackwell is to "fill the inclusive layer of the Google Cloud AI product pyramid": using a general-purpose GPU platform that can run AI inference/fine-tuning and also perform high-end graphics/simulation, quickly bringing a broader range of enterprise workloads (especially low-latency inference, digital twins/industrial simulation, and other physical AI workloads) to the cloud, with better availability and cost-effectiveness to meet the current demand for AI cloud adoption; while B200/B300 continues to play the flagship AI computing role for large model training and ultra-large-scale AI inference.

Industrial simulation, digital twins, and complex visual computing (for simulation/synthetic data/perception) all fall within the core scope or underlying support of "Physical AI"; "Physical AI" emphasizes enabling robots/autonomous systems to perceive, reason, and act in the real world, and these three capabilities are the key toolchain that advances models from "only able to converse" to "able to work in the physical world."

The "full launch" of Google's new G4 model focuses on low-latency AI, simulation, digital twins, and visual computing, expanding the availability zones to the "most concentrated" large user area in history; this fills the gap between Google Cloud Platform's A series (supercomputer training/large-scale inference) and G2 (cost-effectiveness), bringing the capabilities of NVIDIA's Blackwell architecture down to a broader range of enterprise AI inference workloads and a wider range of enterprise-level physical AI workloads The RTX PRO 6000 (server version) features both Tensor Core (including FP4) and RT Core on Blackwell, naturally fitting enterprise-level inference + real-time rendering/digital twin/simulation composite scenarios, capable of handling AI loads such as 30B–100B level AI inference/fine-tuning; Google has also listed NVIDIA's Omniverse / Isaac Sim on Google Cloud, clearly accelerating the implementation of industrial digitalization and the demand for "physical AI" in robotics. Compared to the B200/B300 high-card computing clusters that require HBM storage systems + NVLink, the G4 is aimed at more general AI inference workloads, medium-scale fine-tuning, and a wide range of physical AI workloads, lowering the budget and supply threshold for enterprises.

For the global demand for server-side inference primarily from small and medium-sized enterprises + medium-scale AI fine-tuning/distillation + digital twin/simulation and parallel cloud computing platforms (such as industrial digital twins, VFX, and robotic simulation testing), the Google Cloud G4 built on RTX PRO 6000 is undoubtedly the first choice, enjoying exclusive high-performance acceleration and the ready-to-use high-efficiency industrial AI and physical AI ecosystem of Omniverse/Isaac.

The tide of AI computing power is unstoppable, and leaders in computing power continue to celebrate stock price surges

There is no doubt that, with its powerful GPU processor system and exclusive CUDA software ecosystem, NVIDIA still holds an absolute leading position in the global AI computing power competition. In terms of performance scale, the arrival of the RTX PRO 6000 Blackwell on the Google Cloud platform will undoubtedly be a new performance growth point and AI ecosystem amplifier for both NVIDIA and Google.

In addition to partnering with Google to create a new cloud computing power platform and the newly launched desktop AI system, the company has recently continued to reach huge deals with AI leaders, including a massive investment of $100 billion in OpenAI, which will purchase up to 10 gigawatts of NVIDIA AI server clusters.

According to top Wall Street institutions such as Cantor Fitzgerald, HSBC, and Morgan Stanley, NVIDIA will continue to be the core beneficiary of the trillion-dollar wave of AI spending. These institutions believe that NVIDIA's stock price's repeated historical highs are far from over. Wall Street analysts have recently been continuously raising their 12-month target price for NVIDIA, with the latest average target price from Wall Street indicating that NVIDIA's total market value will break the $5 trillion milestone within a year. More significantly, HSBC has raised its target price for NVIDIA from $200 to $320, the highest level on Wall Street.

As the "company with the highest market value in the world," NVIDIA is regarded as the "overall leader" of the global AI computing power industry chain. Therefore, in the eyes of institutional and retail investors, NVIDIA's strong upward momentum, which continues to set new highs, indicates that this round of "super bull market" in the global AI computing power industry chain is far from over, and this industry chain will remain the most favored investment sector for global capital in the near future. It is under the epic stock price surge of large tech giants such as NVIDIA, Meta, Google, Oracle, TSMC, and Broadcom, as well as the leaders in the AI computing power industry chain, along with their continuously strong performance this year, that an unprecedented AI investment boom has swept through the U.S. stock market and global stock markets. This has driven the S&P 500 index and the global benchmark index—MSCI World Index—to significantly rise since April, recently setting new historical highs.

According to financial giants Morgan Stanley, Citigroup, Loop Capital, and Wedbush, the global investment wave in artificial intelligence infrastructure, centered around AI computing power hardware, is far from over and is only at the beginning. Driven by an unprecedented "AI computing power demand storm," this round of AI investment is expected to reach a scale of $2 trillion to $3 trillion.

Recently, the prices of high-performance storage products in the global DRAM and NAND series have surged, coupled with the fact that the world's highest-valued AI startup, OpenAI, has secured over $1 trillion in AI computing power infrastructure deals. Additionally, TSMC, the "king of chip foundry," has reported exceptionally strong earnings that exceeded expectations and raised its revenue growth forecast for 2025 to the mid-30% range. Together, these factors have significantly reinforced the "long-term bull market narrative logic" for AI computing power infrastructure sectors such as AI GPUs, ASICs, HBM, data center SSD storage systems, liquid cooling systems, and core power equipment