Super intelligence

The terminology the U.S. government uses to describe advanced computing systems is evolving. In September 2026, the U.S. government formally adopted the term "super intelligence" (SI) to replace "artificial intelligence" (AI) across Executive Branch communications, reflecting a policy view that today's frontier systems have capabilities that extend well beyond the original meaning of the term. Whether referred to as AI or SI, these technologies are reshaping industries at scale, and the memory and storage infrastructure that supports them is more critical than ever.

What is the history of the term super intelligence?

The phrase "superintelligence" has existed in academic and philosophical discussions for decades, especially in debates about hypothetical machine intelligence that could surpass human cognitive ability across many or all domains. This academic concept should not be confused with the newer policy use of "super intelligence," which refers to a U.S. government terminology choice rather than a distinct technical category.

In other words, the current U.S. government usage carries a different and narrower meaning: it is a policy-driven rebranding of AI as defined in existing U.S. statute, not a claim that today's systems have reached the academic concept of superintelligence.

Key milestones in context:

• 1956: The field of artificial intelligence is formally named and established in the United States as an academic discipline
• 2017: The Transformer architecture is introduced, significantly advancing the capability of AI systems across language and reasoning tasks
• 2022–2024: Generative AI reaches mainstream adoption; large language models demonstrate capabilities in writing, coding, analysis and scientific reasoning
• July 2025: The White House releases America's AI Action Plan, identifying more than 90 federal actions focused on accelerating SI innovation, building American SI infrastructure, and extending U.S. SI leadership internationally
• March 2026: The White House unveils a National Super Intelligence Policy Framework with legislative recommendations
• September 29, 2026: Executive Order 14434 directs executive departments and agencies to use "Super Intelligence" and "SI" in place of "Artificial Intelligence" and "AI" in official correspondence, public communications, websites, reports, policy documents, and other non-statutory documents within the executive branch
• Within 60 days of the order: The Assistant to the President for Science and Technology (APST) is directed to submit proposed legislative language to establish a formal U.S. government definition of "Super Intelligence" and "SI" that reflects the capabilities described in the order, meaning a more technically specific definition is still forthcoming

What is super intelligence?

Super intelligence traditionally refers to a hypothetical form of intelligence that surpasses human cognitive abilities across virtually all domains. In a separate policy usage, the U.S. government states via Executive Order 14434 that the terms "super intelligence" and "SI" mean the technologies and systems encompassed by the term "artificial intelligence" as defined in section 9401(3) of title 15, United States Code. In practical terms, SI is the U.S. government's official label for the same broad category of technologies that the industry has historically called artificial intelligence, including machine learning, large language models, neural networks and related systems.

According to the executive order, the U.S. government’s rationale behind the terminology change centers on capability. Today's frontier systems do much more than imitate or automate discrete aspects of human intelligence; they increasingly amplify human ingenuity and unlock new forms of creativity, empowering people to achieve what was previously impossible across science, medicine and nearly every other domain of human endeavor. It is worth noting that the underlying technology itself has not changed as a result of this policy; only the U.S. government-mandated terminology has changed. U.S. government departments and agencies may begin using "super intelligence" or "SI" in certain official communications, guidance, websites and other non-statutory materials.

How does super intelligence work?

From a technical standpoint, systems referred to as super intelligence operate on the same foundational architecture as artificial intelligence, including transformer-based large language models, deep neural networks, and inference engines trained on large datasets.

Key components of how these systems function:

• Model training: Models are pre-trained on large datasets using distributed compute, high-bandwidth memory, and parallel processing infrastructure
• Inference and deployment: Trained models are served at scale, requiring fast memory access and low-latency storage to respond in real time
• Fine-tuning: General-purpose models are adapted for specific domains using curated datasets and additional training cycles
• Retrieval-augmented generation (RAG): Systems pull contextual information from external data stores at inference time to improve accuracy and relevance
• Agentic operation: Newer systems can plan and execute multi-step tasks autonomously, interacting with tools and external services

The memory and storage requirements for these workloads are substantial. High-bandwidth memory (HBM), high-capacity DRAM, and fast data center SSDs form the hardware foundation that determines how quickly and efficiently SI systems can process data.

What are the key types of super intelligence?

Under the U.S. government definition, super intelligence encompasses the full spectrum of technologies previously classified as AI. The categories below reflect how these systems are broadly organized by the technical community, regardless of what they are officially called.

1. Foundation models

Foundation models are large-scale systems pre-trained on broad datasets and capable of performing a wide range of tasks. They serve as the base layer for many downstream applications, including conversational assistants, coding tools, and scientific analysis systems.

2. Agentic systems

Agentic AI systems go beyond responding to individual prompts. They plan sequences of actions, use external tools, and operate with a degree of autonomy to complete multi-step tasks. This category is seeing rapid development across enterprise automation and research applications.

3. Domain-specialized systems

These are models fine-tuned or purpose-built for specific fields such as medical imaging, legal analysis, financial modeling, or materials discovery. They apply the capabilities of general foundation models to narrower, high-stakes domains where accuracy and reliability are paramount.

4. On-device systems

Edge AI brings inference capabilities directly to devices, including smartphones, vehicles, and industrial equipment, without requiring constant cloud connectivity. These systems rely on memory-efficient architectures and low-power storage to operate within tight hardware constraints.

How is super intelligence used?

The applications covered under the SI umbrella span virtually every sector of the economy. The following use cases reflect how organizations are deploying these technologies today:

• Scientific and medical research: SI systems accelerate drug discovery, protein modeling, and clinical trial analysis by processing datasets at a scale and speed that human researchers cannot match alone
• Software development: Code generation, debugging, and documentation tools reduce development time and surface errors earlier in the build cycle
• Manufacturing and operations: Predictive maintenance, quality inspection, and supply chain optimization systems improve efficiency and reduce downtime in industrial environments
• Data center infrastructure: Large-scale SI training and inference workloads drive significant demand for high-bandwidth memory, high-capacity DRAM, and NVMe SSDs across cloud and enterprise environments
• Autonomous systems: Vehicles, robotics, and drones use real-time inference to perceive and respond to their environment
• Enterprise productivity: Summarization, document analysis, and decision-support tools are being integrated into standard business workflows across industries
• Government and defense applications: The U.S. government’s Genesis Mission initiative directs more than $5 billion toward harnessing super intelligence to accelerate breakthroughs in science, medicine, energy, and manufacturing