{
  "id": 15985782,
  "title": "To help world gorge on GPUs, Nvidia brings cash bucket to fatten wafer-thin supply chain",
  "url": "https://urgent.news/2026/10/22/to-help-world-gorge-on-gpus-nvidia-brings-cash-bucket-to-fatten-wafer-15985782",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-10-22T00:54:33.000Z",
  "source": {
    "name": "The Register",
    "slug": "the-register",
    "url": "https://www.theregister.com/ai-and-ml/2026/10/22/to-help-world-gorge-on-gpus-nvidia-brings-cash-bucket-to-fatten-wafer-thin-supply-chain/5304623"
  },
  "original_language": "en",
  "account": "Nvidia has recognized the importance of an enhanced silicon substrate and supply chain for its future success, so it has utilized its substantial cash reserves to ensure its suppliers have the necessary resources to maintain a robust flow of GPUs and AI processors. This is not an uncommon practice for manufacturers, as it helps create certainty and spread financial risk. However, Nvidia's announcement of such deals during its Berlin GTC conference is unique, indicating the company's desire to demonstrate to investors its commitment to meeting product demand and anticipating its continued growth for the foreseeable future.\n\nSeveral of Nvidia's recent announcements focused on silicon substrates, or wafers, which are round, blank silicon slabs onto which chipmakers etch circuits. These substrates are produced by companies that sell to foundries like TSMC and Intel, which then transform them into various types of chips. At GTC Berlin, Nvidia's CEO, Jensen Huang, highlighted that demand for AI hardware is immense, but capacity remains the primary constraint. Specifically, the company aims to expand substrate capacity for its products by 2.5 times over the next three years.\n\nTo achieve this goal, Nvidia has incorporated Samsung's Electro-Mechanics business unit into its substrate supply chain. Samsung reportedly possesses immediate capacity at a factory in South Korea, and Nvidia plans to utilize this resource. The deal stipulates that by 2028, approximately one-quarter of Samsung's substrate output will be allocated to Nvidia. Additionally, Nvidia has an existing agreement with Japanese substrate provider IBIDEN, with plans to triple production at an IBIDEN plant by 2028. Furthermore, Nvidia has secured a capacity reservation agreement with Taiwanese substrate player Unimicron, which includes plans for plant expansion. These agreements ultimately benefit Nvidia's silicon substrate supply chain.\n\nNvidia also aims to bolster its capacity for optical connections, a critical component in building AI systems capable of handling modern workloads. The company has signed a capacity reservation agreement with Korean silicon packaging company Amkor, which will result in a 300 percent increase in capacity. Nvidia emphasizes that this partnership will support higher production volumes for current and next-generation platforms. Optical connections are essential for AI systems, as they can efficiently manage the data deluge generated by modern workloads. Nvidia has thus expanded its partnership with optical connection provider SPIL, with the goal of boosting production capacity by 15 times by mid-2027, and again doubling that capacity by the end of 2028. These initiatives suggest that Nvidia believes the AI infrastructure boom is set to continue for many years to come.",
  "summary": "Signs Samsung to build substrates, bolsters spend with other players",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "The Register Science",
        "title": "To help world gorge on GPUs, Nvidia brings cash bucket to fatten wafer-thin supply chain",
        "url": "https://urgent.news/2026/10/22/to-help-world-gorge-on-gpus-nvidia-brings-cash-bucket-to-fatten-wafer",
        "published": "2026-10-22T00:54:33.000Z"
      }
    ]
  },
  "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."
}