{
  "id": 19646493,
  "title": "AI's infrastructure boom is standing on porcelain legs",
  "url": "https://urgent.news/2026/11/12/ais-infrastructure-boom-is-standing-on-porcelain-legs",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-12T11:15:27.000Z",
  "source": {
    "name": "TechRadar",
    "slug": "techradar",
    "url": "https://www.techradar.com/pro/ais-infrastructure-boom-is-standing-on-porcelain-legs"
  },
  "original_language": "en",
  "account": "The AI infrastructure boom is facing challenges due to a lack of attention given to the foundational systems that enable AI functionalities. Search and analytics, although integral to AI applications, are often treated as separate entities within the infrastructure. Investors and tech companies predominantly focus on chips, models, and applications, overlooking the core systems responsible for data retrieval, analysis, and aggregation.\n\nSearch systems are adept at locating relevant information within unstructured data, while analytical databases excel at processing structured datasets, performing calculations, and identifying trends. Historically, these two systems operated independently, each catering to distinct user requirements. However, the advent of natural-language interfaces and AI agents has highlighted the limitations of this separation. A single user query can trigger multiple searches, retrievals, and analytical queries before generating an answer.\n\nThe complexity arises when an AI agent translates a user's request into several machine tasks, each requiring database operations. Tasks that initially appear as a unified action to the user often generate a series of database operations. This issue is exacerbated when an AI agent handles multiple tasks simultaneously, leading to amplified database workload. The added latency, cost, and potential data inconsistency resulting from these hand-offs between systems can significantly impact performance.\n\nMaintaining separate search and analytics systems necessitates extra data copies, synchronization pipelines, and continuous engineering efforts to ensure data consistency. These additional components contribute to increased costs and complexity, particularly as AI agents intensify query volumes. One potential solution is to store data in open formats, such as Apache Parquet files in object storage, enabling different tools to access the same underlying data. By decoupling storage from processing layers, businesses can reduce redundancy and streamline operations. However, even with this approach, search and analytics systems will continue to require specialized components, like indexes and computing layers, to cater to diverse workloads.",
  "summary": "AI agents are exposing the costly divide between search and analytics.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}