{
  "id": 1184135,
  "title": "What’s stopping Asia’s chipmakers from taking on Nvidia?",
  "url": "https://urgent.news/2026/08/16/whats-stopping-asias-chipmakers-from-taking-on-nvidia",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-08-16T03:25:17.000Z",
  "source": {
    "name": "The Business Times - Companies & Markets",
    "slug": "the-business-times-companies-markets",
    "url": "https://www.businesstimes.com.sg/international/global/whats-stopping-asias-chipmakers-taking-nvidia"
  },
  "original_language": "en",
  "account": "The rise of Asia's AI chip startups, many founded by former Intel and AMD employees, has surged alongside the meteoric growth of Nvidia, the world's most valuable public company. However, despite thousands of startups in the region, only a handful of major fabrication facilities, such as TSMC and Samsung Foundry, can produce the advanced chips these companies need. Establishing a relationship with one of these foundries is as crucial as the chips themselves, and these startups are vying to secure access.\n\nTSMC, the market leader in advanced nodes – or generations of chipmaking technology – is highly selective in granting access to its fabrication facilities. A 3 nm node, for instance, is more advanced and efficient compared to a 15 nm node, but it is also much harder and more expensive to manufacture. Despite the limited number of foundries, investors continue to pour money into AI chip startups. VC firm Vertex Ventures has invested in three such startups in South-east Asia.\n\nOne success story is South Korea's FuriosaAI, which has secured production for its advanced AI chips and shipped products to major companies like Samsung and LG. Furiosa's flagship AI inference chip, RNGD, entered mass production in January 2026 on TSMC's 5 nm process and is estimated to cost around $10,000 per unit. However, the process of securing TSMC capacity and high-bandwidth memory, along with packaging partners and customer validation, has been described as \"very complicated.\"\n\nAnother challenge for these startups is the limited supply of high-bandwidth memory (HBM), which is essential for moving data quickly through the chip to run large AI models. HBM chips are also facing a memory constraint due to a shortage of manufacturing facilities, causing prices for memory chips to spike. SK Hynix, Samsung, and Micron dominate the market for HBM, but access to their products is not guaranteed, making it another significant barrier for AI chip startups.\n\nAdvanced packaging solutions, like TSMC's CoWoS, which combines AI chips and HBM into one chip package, also pose challenges. TSMC can only produce a limited number of CoWoS packaging, further exacerbating the bottleneck for companies trying to build high-end AI systems. Startups like Singapore-based Silicon Box are trying to address this issue by using advanced panel-level packaging to connect AI processors, high-bandwidth memory, and other components into a single chip.",
  "summary": "Foundry access is just one constraint that startups are facing",
  "key_points": [
    "Only TSMC and Samsung Foundry can produce advanced AI chips.",
    "TSMC highly selective in granting access to fabrication facilities.",
    "Limited supply of high-bandwidth memory is another significant barrier."
  ],
  "editors_take": "Asia's AI chip startups face significant barriers to challenging Nvidia's dominance, including limited access to top fabrication facilities, high-bandwidth memory constraints, and advanced packaging shortages, stifling their growth and production capabilities.",
  "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."
}