{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreiagw3ffkik3qqgbipgc3br53fzrgbl2sdqu5n4o44yktbsaod5wum",
    "uri": "at://did:plc:mg5ozsljpp6t5b4lvwys4t72/app.bsky.feed.post/3mgckvppepxs2"
  },
  "coverImage": {
    "$type": "blob",
    "ref": {
      "$link": "bafkreia7vd2uknpssbzingplmab3de24nfvgnmmanlgu3dtizzxrqbinqi"
    },
    "mimeType": "image/jpeg",
    "size": 55961
  },
  "description": "The company's CEO said open interfaces and interoperability are essential to connecting intelligence across network domains.\n",
  "path": "/nokia-called-token-certainty-the-new-network-performance-standard/",
  "publishedAt": "2026-03-05T10:15:01.000Z",
  "site": "https://broadbandbreakfast.com",
  "textContent": "BARCELONA, March 5, 2026 — Nokia, the Finnish telecommunications equipment giant, said Wednesday that artificial intelligence has generated 1.3 trillion annual network sessions and 100 trillion tokens per day. The figures are driving a structural shift in how carriers must design infrastructure, the company said, away from stream-based traffic models and toward what it called \"token certainty.\"\n\n**Justin Hotard,** Nokia's president and chief executive, said the volume of AI traffic on mobile networks has reached 77 exabytes per month, with more than half carried over mobile connections. He said the figures signal that the industry is no longer at the start of an AI adoption curve but already deep into a scaling phase.\n\nHotard said traditional network performance metrics, including \"five-nines\" and \"six-nines\" connectivity standards, industry benchmarks measuring the percentage of time a network remains operational, would not be sufficient to handle AI workloads.\n\nOperators must instead architect networks around token delivery, he said, ensuring each token, the discrete unit of data processed by AI language models, arrives with the correct latency, performance, security and quality for the device receiving it.\n\nThe shift in traffic behavior is what makes the transition structurally different from prior network generations, Hotard said. AI traffic is \"bursty and dynamic\" rather than linear and predictable, he said.\n\nSmart glasses, wearable devices that stream live video to cloud AI systems for real-time processing, generate heavy uplink demand, Hotard said. Chatbot and machine-to-machine communications, in which AI systems exchange data without human involvement, generate downlink pressure, and neither fits the capacity-planning models carriers have used for voice, data and video.\n\nHotard said Nokia is already building dense optical fiber links between data centers to support what the industry now calls “AI factories,” large-scale compute clusters whose workloads can no longer be optimized within a single facility. That transformation has flattened network topologies and is changing the fundamental unit of networking, he said.\n\nNokia is working with semiconductor giant Nvidia, along with major public cloud providers and physical AI device partners, Hotard said. He described the effort as a co-innovation model built around open interfaces and interoperability.\n\nNokia's destination is a fully AI-native architecture, Hotard said, one in which dynamic models align connectivity, compute and control across all network domains simultaneously. That approach would replace the siloed model operators have used to manage radio access, transport, core and cloud networks independently, he said.\n\n### _Honor Unveils AI Robot Phone_\n\nEarlier in the keynote, Honor, the Shenzhen-based smartphone maker that spun off from Huawei in 2020, unveiled what it called a \"robot phone.\" The device features a motorized gimbal camera, a stabilizing mount that moves independently of the handset, capable of autonomous movement, the company said.\n\nHonor chief executive **James Lee** said the product, called the Aifon, combines on-device AI with cloud-based large models to automate filmmaking tasks including camera movement, music selection and video editing.\n\nLee said Honor's broader strategy organizes AI capabilities into three layers: personal intelligence running on-device, universal intelligence drawing from cloud knowledge bases, and ambient intelligence embedded in physical devices including robots and electric vehicles. The framework reflects what Lee described as Honor's commitment to human-centric AI development.",
  "title": "Nokia Called 'Token Certainty' the New Network Performance Standard",
  "updatedAt": "2026-05-21T22:08:21.113Z"
}