Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesAMD and NVIDIA did not announce a partnership with each other. On May 13, 2025, each chipmaker separately agreed to work with Saudi Arabia’s PIF-backed AI company, HUMAIN. AMD described a collaboration worth up to $10 billion for as much as 500 megawatts of AMD-based capacity over five years. NVIDIA described a separate Saudi AI-factory program of up to 500 megawatts, starting with 18,000 GB300 Grace Blackwell GPUs and eventually reaching several hundred thousand GPUs. Those are announced targets—not proof that the facilities, power systems or commercial services are already operating.
The two deals at a glance
| Measure | AMD–HUMAIN | NVIDIA–HUMAIN |
|---|---|---|
| Announcement date | May 13, 2025 | May 13, 2025 |
| Announced value | Up to $10 billion, as described by AMD | No comparable dollar value disclosed in NVIDIA’s announcement |
| Projected capacity | Up to 500 MW over five years | Up to 500 MW |
| Hardware | Instinct GPUs, EPYC CPUs, Pensando DPUs and Ryzen AI processors | Several hundred thousand NVIDIA GPUs over five years |
| First phase | Not quantified in the announcement | 18,000 GB300 Grace Blackwell GPUs with InfiniBand networking |
| Software and platforms | ROCm and AMD’s broader data-center software stack | NVIDIA Omniverse, alongside NVIDIA’s AI software |
| Geographic scope | Facilities extending from Saudi Arabia to the United States | Saudi AI factories, with the announcement focused on the kingdom |
| Timing stated | Multi-exaflop capacity targeted by early 2026, a forward-looking AMD statement | Five-year GPU and data-center buildout |
Do not add the two 500-MW figures together as a single 1,000-MW project. The companies did not say that their systems will share sites, customers or a common technical platform.
AMD’s announcement is the source for the $10 billion figure and its 500-MW plan. NVIDIA’s figures come from its AI-factory announcement.
What AMD and HUMAIN said they will build
AMD said HUMAIN will deploy up to 500 MW of AMD-based compute capacity over five years, with facilities planned in Saudi Arabia and the United States. AMD’s contribution spans the accelerator, server, networking and software layers:
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
#1 Best Overall
- PLEASE NOTE: Exporting an NVIDIA RTX Pro 6000 GPU outside the US requires strict adherence to the U.S. Export Administration Regulations (EAR) and issuance of an export license from the Bureau of Industry and Security (BIS). Compliance and Know Your Customer (KYC) screening may be required as a condition of order acceptance. [NVIDIA Blackwell Streaming Multiprocessor] The new SM features increased processing throughput, and new neural shaders that integrate neural networks inside of programmable shaders | DLSS 4: Multi Frame Generation ensures ultra-smooth frame pacing for lifelike simulations.
- [Double-Flow-Through Design] The RTX PRO 6000 Blackwell features a double-flow-through cooling design, optimizing efficiency and airflow to sustain peak performance under 600W power loads. | [5th Gen Tensor Cores] Deliver up to 3X the performance of the previous generation and support for FP4 precision for faster AI model processing times with reduced memory usage, enabling local fine-tuning of LLMs and generative AI | [4th Gen Ray Tracing Cores] Double the ray-triangle intersection rate of the previous generation to create photoreal, physically accurate scenes and immersive 3D designs with RTX Mega Geometry, which enables up to 100X more ray-traced triangles.
- [PCIe Gen 5] Support for PCIe Gen 5 provides double the bandwidth of PCIe Gen 4, improving data-transfer speeds from CPU memory and unlocking faster performance for data-intensive tasks like AI, data science, and 3D modeling. | [GDDR7 Memory] With 96 GB of GPU memory and 1.8 TB ps bandwidth, it can tackle massive 3D and AI projects, fine-tune AI models locally, explore large-scale VR environments, and drive larger multi-app workflows.
- [DisplayPort 2.1] Achieve unparalleled visual clarity and performance, driving high resolution displays at up to 8K at 240 Hz and 16K at 60 Hz. Increased bandwidth enables seamless multi-monitor setups while HDR and higher color depth support ensures superior color accuracy for precision work, such as video editing, 3D design, and live broadcasting.
- [Universal MIG] Divide a single RTX PRO 6000 Blackwell into multiple isolated instances, each with dedicated resources, allowing for concurrent execution of multiple workloads, optimized GPU utilization, and secure isolation of different applications or users. [WARRANTY] 3 YR Manufacturer's Warranty. Bulk OEM Packaging. Retail Packaging is NOT included.
- Instinct GPUs for AI training and inference.
- EPYC server CPUs.
- Pensando data-processing units for infrastructure workloads.
- Ryzen AI processors for selected edge and client applications.
- ROCm software and related development tools.
HUMAIN is expected to handle hyperscale data centers, sustainable power systems, global fiber interconnects and end-to-end delivery. AMD also said the program could activate multi-exaflop capacity by early 2026. That is a company forecast, not independent confirmation that multi-exaflop infrastructure was commissioned by that date.
What NVIDIA and HUMAIN said they will build
NVIDIA described a separate program of up to 500 MW and several hundred thousand NVIDIA GPUs over five years. Its first phase was specified as an 18,000-GPU system based on NVIDIA GB300 Grace Blackwell, connected with InfiniBand networking.
NVIDIA also positioned Omniverse as a multi-tenant environment for simulation, robotics and digital twins. The partnership includes workforce training and developer upskilling, suggesting that the intended customer base extends beyond government model training to industrial and commercial users.
NVIDIA did not publish a dollar value comparable to AMD’s “up to $10 billion” figure in the cited announcement. A GPU count or power target cannot be converted reliably into total spending without details about configuration, buildings, financing, networking, cooling, land and operating costs.
Who HUMAIN is—and why “startup” is incomplete
HUMAIN is a newly launched Saudi AI company and subsidiary of the Public Investment Fund (PIF), not simply a conventional venture-funded startup. Its intended scope covers data centers, high-performance computing, cloud platforms, large language models, sector-specific applications and talent development.
That structure gives the project a national-infrastructure dimension. HUMAIN can act as an operator and platform provider while the Saudi state supplies strategic backing and access to long-horizon capital. The announcements indicate ambitions to serve Saudi and international customers, but they do not publish a detailed pricing model, customer list or utilization commitments.
What “AI service centers” means here
“AI service centers” is useful editorial shorthand, not the companies’ principal formal name. AMD uses language such as AI computing capacity and hyperscale data centers; NVIDIA calls its facilities AI factories and GPU infrastructure.
In practical terms, these are specialized data centers intended to provide:
Rank #2
- NVIDIA Volta GV100 Architecture — 4,608 CUDA Cores, 640 1st-Gen Tensor Cores delivering 14 TFLOPS FP32 and 112 TFLOPS deep learning performance for AI training, inference, HPC, and scientific computing workloads
- 32GB HBM2 ECC Memory — 900 GB/s Bandwidth — High-bandwidth memory on a 4096-bit bus with ECC error correction provides the memory capacity and throughput required for the largest AI models, simulations, and datasets
- PCIe 3.0 x16 Interface — 250W TDP — Standard PCIe Gen3 connectivity with passive cooling designed for enterprise rack server deployment in HPE ProLiant, Dell PowerEdge, and Supermicro platforms with adequate chassis airflow
- NVLink — Scale to 96GB Unified Memory — Connect two V100 GPUs via NVLink at 300 GB/s bi-directional bandwidth to scale GPU memory from 32GB to 96GB for larger AI training and HPC workloads
- Multi-Precision Computing — Supports FP64 (7 TFLOPS), FP32 (14 TFLOPS), FP16 (112 TFLOPS) and INT8 precision modes for flexible deployment across training, inference, and scientific simulation workloads
- Large-model training and inference.
- Cloud access to GPUs and other accelerated computing.
- Sovereign government and enterprise workloads.
- Robotics, industrial simulation and digital twins.
- Developer and startup computing capacity.
Megawatts and GPU counts describe the scale of the infrastructure. They do not establish model quality, revenue, customer adoption, energy efficiency or lower prices for AI users.
Why Saudi Arabia is pursuing this scale
Saudi Arabia’s strategy aligns with Vision 2030’s effort to diversify beyond oil. AI infrastructure requires land, electricity, cooling, fiber connectivity, financing and specialized talent—areas where a state-backed vehicle can coordinate projects over many years.
Building capacity locally could give Saudi agencies and businesses more control over where sensitive data and models are processed. It could also let the kingdom sell regional or global compute rather than remain solely a customer of foreign cloud providers. The announcements support those strategic aims, but they do not independently demonstrate future economic returns or a guaranteed role as a major global cloud provider.
Workloads the facilities are intended to support
Sovereign and commercial AI
The stated use cases include training and inference for sovereign models, government workloads, enterprise applications and GPU-cloud services. “Sovereign” generally means that data, models and operational control can be kept within a chosen national jurisdiction or subject to its rules.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Industrial simulation and physical AI
NVIDIA specifically highlighted Omniverse for robotics, manufacturing, logistics, energy and digital-twin workloads. These applications use 3D representations and simulations to test equipment, factories or processes before changing physical operations.
Developers and startups
Both announcements describe a broader ecosystem role, including developer access, training and upskilling. Whether independent developers will receive open access, what regions will be served and what prices will apply remain undisclosed.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What remains unproven
Announcement versus operating capacity
The public announcements establish partnerships and targets. They do not, by themselves, establish financing close, construction completion, hardware delivery, system commissioning, commercial launch or achieved utilization.
Power, cooling and connectivity
A 500-MW AI buildout is a major industrial project. The announcements do not specify whether the figure refers to IT load or total facility power, nor do they detail generation, transmission, cooling technology, water management, substations, fiber routes or site-by-site schedules.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Rank #3
- Professional GPU with Blackwell Architecture
- Blackwell Architecture
- 24GB GDDR7 with PCIe 5.0 & Ray Tracing
- AI Workstation
Export controls and supply
Advanced accelerators can require government licenses and remain subject to export rules, tariffs and supply constraints. AMD’s forward-looking-risk language specifically cites government action, export regulations, licensing, manufacturing and customer-demand risks. A published GPU target is therefore not the same as an immediately shippable quantity.
Software and customer economics
AMD’s ROCm-centered approach offers an alternative to NVIDIA’s mature CUDA ecosystem, but buyers must evaluate framework support, optimized kernels, multi-GPU scaling, orchestration, monitoring, porting work and developer familiarity. Neither announcement provides independent benchmarks, workload results, customer pricing or utilization data.
The commercial test is whether HUMAIN can attract paying users and operate reliably. Key unanswered questions include who will lease capacity, whether government workloads receive priority, how data residency will work, whether customers can choose AMD or NVIDIA systems and what minimum commitments will apply.
The project expanded after the May announcements
On November 19, 2025, AMD, Cisco and HUMAIN announced a planned joint venture targeting up to 1 gigawatt of AI infrastructure by 2030, beginning with a planned 100-MW Saudi deployment. This is a later development involving Cisco; it does not turn the original AMD and NVIDIA announcements into one joint AMD-NVIDIA project. Details such as completed construction, customer availability and achieved utilization still require separate confirmation. The announcement is documented by AMD.
How to interpret the business opportunity
For technology companies, the deals could add accelerator capacity outside established US, European and East Asian hubs. AMD gains a high-profile reference deployment for its ROCm-based stack; NVIDIA gains another large market for Blackwell systems, networking and Omniverse. Saudi Arabia gains a potential foundation for sovereign computing and regional cloud services.
For customers, the practical choice will depend on workload and access rather than headline capacity. Organizations considering AMD should test ROCm compatibility and porting costs. Those seeking managed NVIDIA infrastructure may compare providers such as DGX Cloud and NVIDIA AI Enterprise. Industrial teams evaluating simulation can review Omniverse. These products are separate from HUMAIN’s announced facilities, and no public signup or pricing for the planned Saudi capacity was established in the announcements.
The Bottom Line
AMD and NVIDIA made separate deals with PIF-backed HUMAIN, each describing an eventual 500-MW-scale AI buildout. AMD disclosed an “up to $10 billion” collaboration; NVIDIA disclosed a first phase of 18,000 GB300 GPUs and a projection of several hundred thousand GPUs. The plans mark Saudi Arabia’s push to become an AI-infrastructure provider, but announced megawatts, GPU counts and investment ceilings are targets—not evidence of completed, available or profitable capacity.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
