TLDR
- NVIDIA’s new RTX PRO 5000 GPU now features 72GB GDDR7 memory.
- xAI signs $20B deal with NVIDIA for Memphis Colossus 2 data center.
- Both RTX PRO 5000 models maintain a 300W TDP and same CUDA cores.
- NVIDIA’s Q2 data center revenue reached $41.1 billion in 2025.
NVIDIA made headlines twice this week. While Elon Musk’s xAI confirmed a $20 billion GPU deal with the company, NVIDIA also released a new 72GB version of its RTX PRO 5000 “Blackwell” graphics card. The move caught the tech world by surprise, as the GPU was launched quietly without a major press release or event.
NVIDIA Quietly Rolls Out 72GB RTX PRO 5000 Blackwell GPU
NVIDIA has released a new version of its RTX PRO 5000 Blackwell graphics card featuring 72GB of GDDR7 memory. This model includes 50% more memory than the existing 48GB version. The new GPU maintains the same 512-bit memory bus and memory bandwidth as the earlier model.
The unusual part of this configuration is how the memory was expanded. A standard 512-bit bus supports 16 memory channels, each at 32-bit. The 48GB model uses 16 x 3GB GDDR7 chips. To achieve 72GB, NVIDIA is likely using double-stacked 3GB modules on half of the memory interfaces, bringing some of the channels to 6GB. This gives the GPU an uneven memory layout but allows it to retain the same memory interface and power requirements.
Both models continue to offer a thermal design power (TDP) of 300W. They also share the same CUDA core count and memory bandwidth, meaning performance differences will come mostly from memory capacity. The upgrade targets users running large datasets, 3D design projects, or high-resolution workloads.
Elon Musk’s xAI Signs $20 Billion Deal With NVIDIA
Alongside the hardware news, NVIDIA also made headlines due to a major partnership with Elon Musk’s xAI. According to reports, xAI has agreed to purchase up to $20 billion worth of GPUs for its Memphis Colossus 2 data center.
The purchase is not expected to happen all at once. Instead, the transaction will likely take place over several quarters. While the $20 billion figure may seem large, NVIDIA’s Q2 data center revenue was $41.1 billion, suggesting the deal will account for only a small portion of its overall growth. However, it is seen as a strong signal of continued demand in the AI sector.
The partnership comes at a time when other AI companies like OpenAI are also forming supplier agreements with tech firms including AMD and Broadcom. While AMD competes directly with NVIDIA in the GPU market, Broadcom is working with clients to create custom chips, optimized for specific AI workloads.
NVIDIA’s Role in the Expanding AI Market
NVIDIA continues to lead in AI hardware, supplying GPUs to major players in the space. The company has already announced a collaboration with OpenAI to deliver 10 gigawatts of computing power in return for investment over time.
NVIDIA co-founder and CEO Jensen Huang said during the company’s Q2 earnings call that data center investments could reach $600 billion by the end of 2025. He added that this figure may grow to over $3 trillion by 2030. These estimates, if accurate, show large-scale opportunities for NVIDIA’s long-term growth in the enterprise AI and data infrastructure sector.
Analysts expect 33% revenue growth for NVIDIA by FY 2027, although the company’s past results often exceeded such forecasts. If current trends continue, the stock’s current valuation may not reflect its long-term earnings potential.
Expanded GPU Options Signal Wider Market Strategy
The RTX PRO 5000 Blackwell now comes in two memory configurations—48GB and 72GB—providing more flexibility for creative professionals and technical users. Despite the change in memory capacity, the rest of the specifications remain unchanged, keeping the focus on high-memory workloads without increasing power draw or core counts.
At the same time, NVIDIA’s broader GPU lineup includes more accessible options like the RTX 5080, which supports next-generation features including DLSS 4 and improved ray tracing. The availability of both high-end and mid-range products points to a balanced strategy that meets demand across different user groups.
As NVIDIA expands its product line and signs long-term deals with major AI firms, it continues to position itself as a key supplier for both personal and enterprise computing needs.
