The core message of the session was that AI should not be understood as software alone: it requires a physical and institutional build-out of compute, power, sustainable energy, connectivity, chips, networks, researchers, universities, institutes, regulation, and adoption. The moderator framed Abu Dhabi’s opportunity around what is being built, how those projects may be financed, and how external investors could participate. Eyad Abdulla argued that Abu Dhabi is not simply “writing a check” for AI companies to build data centers, but is constructing an integrated ecosystem. Masdar City’s role, in his account, is to host and connect critical parts of that ecosystem, including Mohamed bin Zayed University of Artificial Intelligence, established in 2019 and described as the first university in the world specialized for AI, and the Technology Innovation Institute. He also described Masdar City’s six focus clusters: AI, smart mobility, life sciences, space technology, agri-tech, and power.

Eddie Cheng explained the session title, “compute is the new concrete,” by describing how compute has shifted from general-purpose processing toward token consumption driven by large language models and AI-native applications. For Origen, this means transforming into an AI-native organization and delivering AI-native products and services. He contrasted traditional applications, which rely on rules, inputs, predefined outputs, and interfaces, with AI-native applications that are anchored around context, learn user needs, and complete tasks automatically. He emphasized the speed of adoption by comparing electricity, which he said took around a century to reach 90% of households, with the internet, which took about 30 years, and AI, which he expects to take 10 years, with large language models beginning in 2020 and the world already six years into that journey.

The discussion then turned to what “the right kind” of compute means for ordinary residents. Cheng’s answer was deliberately human-centered: the right kind of AI should not compete for attention. He compared humans to AI models with daily token limits, arguing that people have limited energy and attention. Good AI products, in his view, should consume the minimum attention required to complete useful tasks; if they constantly demand engagement, users stop consuming AI and AI starts consuming them. Abdulla connected this back to city planning, saying AI is now part of the “concrete” of a city and must be planned into urban development rather than treated as optional.

Abdulla closed the provided portion by breaking the AI ecosystem into three layers: infrastructure, capability, and adoption. Infrastructure includes servers, compute, and connectivity; capability includes researchers, engineers, universities, and institutes; adoption requires government, businesses, companies, residents, and citizens to use the technology. He described Masdar City as a global test bed, citing autonomous mobility, smart technologies, Encelco’s AI-assisted drug and prescription development, the Personal Rapid Transit system launched in 2009 that has carried more than 3 million people, level-four autonomous vehicle testing across roughly 2.5 kilometers, regulatory work with ITC, and the BioSphere lab for biotech collaboration. The provided transcript ends as Cheng begins explaining why AI adoption is moving faster than electricity adoption.