{
  "id": 7816844,
  "title": "Alexis Crowell, CMO and GM of the Americas, Axelera AI",
  "url": "https://urgent.news/2026/09/16/alexis-crowell-cmo-and-gm-of-the-americas-axelera-ai",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-16T15:29:11.000Z",
  "source": {
    "name": "Newsweek",
    "slug": "newsweek",
    "url": "https://www.newsweek.com/alexis-crowell-cmo-and-gm-of-the-americas-axelera-ai-12449516"
  },
  "original_language": "en",
  "account": "Axelera AI is a semiconductor company focused on developing hardware and software for artificial intelligence to run anywhere data is created. Their solutions promise high performance and power efficiency. The company has shipped to over 600 customers across industrial manufacturing, security, retail, and enterprise sectors.\n\nCustomers in these sectors seek AI sovereignty, wanting to know exactly where their data is processed and who can access it. This concern is present in Europe, the Middle East, and increasingly in the United States.\n\nThe evolution of AI architecture is a key focus, with some workloads suited for the cloud and others for edge computing. Training large frontier models is best done centrally due to capital intensity and scale. However, inference, or the workload that runs continuously, is more suited to edge computing due to economic realities. For applications with high regulation, privacy, and latency concerns, inference should be located close to the data input.\n\nEnergy efficiency is crucial for various applications, such as robotics, drones, and enterprise data centers. For robotics and drones, every watt spent on inference translates to runtime between charges. For enterprises with on-prem data centers, adding AI inference to existing chassis and power envelope becomes a procurement decision rather than a capital construction project. Retrofits of installed equipment, like cameras and gateways, can also benefit from AI modules that fit within the existing compute slot and power budget.\n\nTo enable AI everywhere, hardware must prioritize memory and memory bandwidth, while software should allow developers to compile, quantize, deploy, monitor, and update models without rebuilding the system.",
  "summary": null,
  "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."
}