US President Announces Massive AI Initiative with OpenAI, SoftBank, and MGX

Joint Venture to Invest $500 Billion in AI Data Centers

US President Donald Trump announced a joint venture between OpenAI, Japanese conglomerate SoftBank Group, and Abu Dhabi’s AI-focused sovereign wealth fund MGX, to invest as much as $500 billion over multiple years in US data centers dedicated to artificial intelligence.

Major Tech Firms to Spend Heavily on AI

The funding, intended to provide "colossal" data centers of 500,000 square feet, comes on top of an enormous amount of capital spending planned by the largest tech firms. According to one Wall Street analyst, Jackson Ader of the KeyBanc Capital Markets brokerage, "Capital spending for the major hyperscalers" – that is, Amazon, Alphabet, Microsoft, Oracle, Meta, and Alibaba – "will be $286.5 billion in 2025, and $308.8 billion in 2026, after making some adjustments to consensus estimates."

Stargate Adds Fuel to AI Spending Narrative

While the initial $100 billion planned for Stargate is just a dollop on top of all that planned spending, it is still encouraging to Ader and others who have been forecasting high AI spending for years to come.

Analysts Weigh In

Numerous questions remain regarding the initiative. SoftBank’s Son is responsible for putting together the initial $100 billion, and it is not yet clear where that money will come from. In a note to clients, tech analyst Brad Zelnick of the investment bank Deutsche Bank noted that one big question is the "ownership and financial structure of the new joint venture, the sources and amount of committed funding already in place, plans for future funding, and how it will be accounted for within the financial results of each company."

Critics Weigh In

AI critic Gary Marcus weighed in on Substack, calling it an "AI catfight." "Elon thinks Sam is playing the same move that they once played together," when the two were forming OpenAI and had not yet produced money but were telling people they had. "He could be right," adds Marcus, "neither is known for absolute candor."

Conclusion

The Stargate initiative is a significant step forward in the development of artificial intelligence, with the potential to drive innovation and growth in the field. While there are many questions still to be answered, the joint venture between OpenAI, SoftBank, and MGX is a major win for the AI industry.

FAQs

Q: What is the Stargate initiative?
A: The Stargate initiative is a joint venture between OpenAI, SoftBank Group, and MGX to invest in US data centers dedicated to artificial intelligence.

Q: How much money is being invested in the Stargate initiative?
A: The initiative is expected to invest as much as $500 billion over multiple years.

Q: Who is involved in the Stargate initiative?
A: The initiative involves OpenAI, SoftBank Group, MGX, and other technology partners, including Oracle, Microsoft, ARM Holdings, and Nvidia.

Q: What is the purpose of the Stargate initiative?
A: The purpose of the Stargate initiative is to drive innovation and growth in the field of artificial intelligence.

Q: What are the potential benefits of the Stargate initiative?
A: The potential benefits of the Stargate initiative include driving innovation and growth in the field of artificial intelligence, creating new job opportunities, and increasing economic growth.

Like Dominoes Falling Over

Photovoltaic system disconnect.

For ease, they used a Flipper Zero device they had configured to send the proper telegram to the photovoltaic system. They did this after discovering that the Flipper Zero’s RFID reading mode could be used to send signals modulated with frequency-shift keying to receivers within a one-meter distance.

Credit: Positive Security

With confidence that an attacker could send unauthorized Radio Ripple Control telegrams that instructed real electrical systems connected to the grid, the researchers got to wondering: What’s the maximum amount of damage a malicious actor—most likely one working for a nation-state—could inflict?

The Maximum Amount of Damage

The researchers surveyed the grid to measure the capacity of power that small- and medium-sized renewable facilities could feed into the grid. They arrived at the estimate of 40 GW. Combined with the 20 GW of load they theoretically can add, that amounted to an unbalanced capacity of 60 GW, enough to power roughly all of Germany. They posited that a sudden change that added or ditched that amount of electricity from the grid all at once could create enough instability to take it down entirely.

Like Dominoes Falling Over

In a published summary of last month’s presentation, the researchers explained their thinking behind the estimate:

To understand, we need to look at the grid frequency. It’s 50 hertz, and it should always stay there.

• If it reaches 50.2 hertz or more, interventions are triggered to reduce the supply. For example, using the technology we’re discussing today to turn off solar parks.
• If the frequency drops below 49.8 hertz, other interventions occur, such as activating energy reserves or disconnecting industries that have contractually agreed to this happening. Also, the first hardware fails as it happened at Vienna airport.
• If the frequency reaches 49 Hz or less, automated stepwise load shedding begins, up to 50% at 48.5 Hz. That might sound a bit technical and sober, but what it means for the European grid is over 200 million people without power.
• At 47.5 Hz, power plants disconnect from the grid to protect themselves from damage. At that point, the grid needs to be rebuilt from scratch.

In theory, with a fully loaded grid at 300 GW, creating a 1 Hz change to reach this private load-shedding threshold requires an imbalance of 18 GW. However, such a large imbalance—though not even that massive compared to the 60 GW estimate—has never been seen.

In practice, one of the most recent incidents was in 2021, when approximately 3 GW of power were unexpectedly lost in Poland, causing the grid frequency to drop by 0.16 hertz. What this demonstrates is that the grid hasn’t yet faced such a significant imbalance.

Power Transfer

But if we start talking about imbalances of 18 GW, or 60 GW, or even more when considering other countries, there’s an additional issue besides the theoretical effect on grid frequency. That issue is power transfer.

If a significant amount of power is missing in one region, it must be transferred there over power lines that could become overloaded. These lines might then shut off to prevent damage, which could overload other lines, causing them to shut off too.

Such a domino effect—or cascade—happened in 2006, when a power line was shut off to accommodate a cruise ship transport. The planning wasn’t thorough, and a cascade of failures followed. So, the theoretical limits of the grid don’t fully capture the potential for much larger disruptions.

Send Malicious Telegrams to Select FREs

There are enough obstacles to make triggering such a catastrophic disruption challenging at best (Bräunlein’s and Melette’s assessment) or doubtful to unlikely (the assessment of an outside grid expert). The researchers noted three key requirements for such an attack.

Conclusion

Taking all of that into account, it’s clear there is enough power under radio control to cause serious trouble.

FAQs

Q: What is Radio Ripple Control?
A: Radio Ripple Control is a technology used to control and manage renewable energy sources connected to the grid.

Q: How can malicious actors use Radio Ripple Control?
A: Malicious actors can use Radio Ripple Control to send unauthorized telegrams to select FREs (Frequency Response Elements) that control real electric systems in their lab.

Q: What is the maximum amount of damage a malicious actor could inflict?
A: The researchers estimate that a malicious actor could create an imbalance of 60 GW, enough to power roughly all of Germany, and potentially cause widespread disruptions to the grid.

Q: Is it possible to trigger such a catastrophic disruption?
A: The researchers note that there are enough obstacles to make triggering such a disruption challenging at best, or doubtful to unlikely.

Cost-Effective User Throughput

Businesses are often challenged with balancing the performance and costs of inference workloads. While some customers or use cases may work with an out-of-the-box or hosted model, others may require customization. NVIDIA technologies simplify model deployment while optimizing cost and performance for AI inference workloads. In addition, customers can experience flexibility and customizability with the models they choose to deploy.

• NVIDIA NIM inference microservices are prepackaged and performance-optimized for rapidly deploying AI foundation models on any infrastructure — cloud, data centers, edge or workstations.
• NVIDIA Triton Inference Server, one of the company’s most popular open-source projects, allows users to package and serve any model regardless of the AI framework it was trained on.
• NVIDIA TensorRT is a high-performance deep learning inference library that includes runtime and model optimizations to deliver low-latency and high-throughput inference for production applications.

Available in all major cloud marketplaces, the NVIDIA AI Enterprise software platform includes all these solutions and provides enterprise-grade support, stability, manageability and security.

Cloud-Based LLM Inference

To ease LLM deployment, NVIDIA has collaborated closely with every major cloud service provider to ensure that the NVIDIA inference platform can be seamlessly deployed in the cloud with minimal or no code required.

• Amazon SageMaker AI, Amazon Bedrock Marketplace, Amazon Elastic Kubernetes Service
• Google Cloud’s Vertex AI, Google Kubernetes Engine
• Microsoft Azure AI Foundry coming soon, Azure Kubernetes Service
• Oracle Cloud Infrastructure’s data science tools, Oracle Cloud Infrastructure Kubernetes Engine

Plus, for customized inference deployments, NVIDIA Triton Inference Server is deeply integrated into all major cloud service providers.

Transforming Agreement Management With Docusign

Docusign, a leader in digital agreement management, turned to NVIDIA to supercharge its Intelligent Agreement Management platform. With over 1.5 million customers globally, Docusign needed to optimize throughput and manage infrastructure expenses while delivering AI-driven insights.

NVIDIA Triton provided a unified inference platform for all frameworks, accelerating time to market and boosting productivity by transforming agreement data into actionable insights. Docusign’s adoption of the NVIDIA inference platform underscores the positive impact of scalable AI infrastructure on customer experiences and ROI. This transformation highlights how robust AI infrastructure can revolutionize financial services.

Serving 400 Million Search Queries Monthly With Perplexity AI

Perplexity AI, an AI-powered search engine, handles over 435 million monthly queries. Each query represents multiple AI inference requests. To meet this demand, the Perplexity AI team turned to NVIDIA H100 GPUs, Triton Inference Server and TensorRT-LLM.

Supporting over 20 AI models, including Llama 3 variations like 8B and 70B, Perplexity processes diverse tasks such as search, summarization and question-answering. By using smaller classifier models to route tasks to GPU pods, managed by NVIDIA Triton, the company delivers cost-efficient, responsive service under strict service level agreements.

Elevating Creative Workflows With Let’s Enhance