<h1 style="font-size: 22px; font-weight: normal;">Why Is Everyone Making Their Own? :AI Boom Brings Hardware Back to the Spotlight</h1>In the early 2020s, the tech world was dominated by software.But since 2023, with generative AI going mainstream, hardware, especially AI chips has become central again.Running models like ChatGPT, Claude, or Gemini requires massive computational power, primarily powered by GPUs. And the market leader? NVIDIA.Yet, in recent years, companies are challenging NVIDIA’s dominance. AMD, Intel, ARM, and even AI firms like OpenAI and Anthropic are developing their own chips.<h3 style="font-size: 22px; font-weight: normal;">NVIDIA’s Dominance</h3>Originally a graphics card manufacturer, NVIDIA captured the AI computing market with parallel-optimized GPU architectures.<ul style="list-style: inside none; padding: 0px; margin: 0px;"><li style="list-style: inside none; padding: 0px; margin: 0px;">AI-focused GPUs: A100, H100, B100</li><li style="list-style: inside none; padding: 0px; margin: 0px;">Global AI chip market share (2023): 80%+</li><li style="list-style: inside none; padding: 0px; margin: 0px;">Market cap briefly reached $3 trillion, making NVIDIA the world’s most valuable company for a moment</li></ul>But NVIDIA GPUs are expensive and supply-constrained. Both startups and big tech companies see the need to reduce dependency.<h3 style="font-size: 22px; font-weight: normal;">New Challengers in the Market </h3><img src="https://letspl.s3.ap-northeast-2.amazonaws.com/images_temp/1/preview-1757139987028.png" width="90%" draggable="true" style="max-width: 100%; width: 627px; height: auto; cursor: pointer;"><table style="width: 100%; border-collapse: collapse; min-width: 125px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"><col style="min-width: 25px;"><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr style="background-color: rgb(255, 255, 255); height: 40px; border-bottom: 1px solid rgb(249, 249, 249);"><th colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Company</th><th colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Key Products</th><th colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Performance</th><th colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Price</th><th colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Notes</th></tr><tr style="background-color: rgb(255, 255, 255); height: 40px; border-bottom: 1px solid rgb(249, 249, 249);"><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">NVIDIA</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">A100, H100, B100</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Best AI training &amp; inference, optimized for parallel computing</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Tens of thousands USD</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Market share 80%+, top-tier performance</td></tr><tr style="background-color: rgb(255, 255, 255); height: 40px; border-bottom: 1px solid rgb(249, 249, 249);"><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">AMD</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">MI300 series</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Competitive, slightly below H100</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Several thousand USD</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Price-competitive, energy-efficient</td></tr><tr style="background-color: rgb(255, 255, 255); height: 40px; border-bottom: 1px solid rgb(249, 249, 249);"><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Intel</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Gaudi 3</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Optimized for AI training &amp; inference, strong x86 integration</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Mid-range (thousands USD)</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Transition from CPU to AI chip</td></tr><tr style="background-color: rgb(255, 255, 255); height: 40px; border-bottom: 1px solid rgb(249, 249, 249);"><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">ARM (Apple, Qualcomm)</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Apple M series, Snapdragon AI Engine</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Low-power, edge &amp; mobile AI optimized</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Hundreds USD</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Great for mobile &amp; edge AI</td></tr><tr style="background-color: rgb(255, 255, 255); height: 40px; border-bottom: 1px solid rgb(249, 249, 249);"><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">OpenAI</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Custom (undisclosed)</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">In development</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">TBD</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Aims to reduce NVIDIA dependency, optimize for AI workloads</td></tr><tr style="background-color: rgb(255, 255, 255); height: 40px; border-bottom: 1px solid rgb(249, 249, 249);"><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Anthropic, Cohere</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Custom (undisclosed)</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">In development</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">TBD</td><td colspan="1" rowspan="1" style="border: 1px solid rgb(204, 204, 204); padding: 8px;">Build chips for cost savings &amp; competitive advantage</td></tr></tbody></table><h3 style="font-size: 22px; font-weight: normal;">The Essence of the AI Chip War</h3>It’s not just about raw performance. The battle revolves around:<ol style="list-style: inside none; padding: 0px; margin: 0px;"><li style="list-style: inside none; padding: 0px; margin: 0px;"></li><li style="list-style: inside none; padding: 0px; margin: 0px;">- Performance </li><li style="list-style: inside none; padding: 0px; margin: 0px;">→ Faster training, more efficient inference</li><li style="list-style: inside none; padding: 0px; margin: 0px;">- Energy efficiency </li><li style="list-style: inside none; padding: 0px; margin: 0px;">→ Large models consume enormous power; cost &amp; carbon footprint matter</li><li style="list-style: inside none; padding: 0px; margin: 0px;">- Supply &amp; pricing </li><li style="list-style: inside none; padding: 0px; margin: 0px;">→ High-end GPUs cost tens of thousands USD; affordable, accessible chips are needed</li></ol><h3 style="font-size: 22px; font-weight: normal;">Geopolitics and Chips</h3>The AI chip war also has geopolitical dimensions.<ul style="list-style: inside none; padding: 0px; margin: 0px;"><li style="list-style: inside none; padding: 0px; margin: 0px;"></li><li style="list-style: inside none; padding: 0px; margin: 0px;">USA restricts GPU exports to China → AI chips become a national security concern</li><li style="list-style: inside none; padding: 0px; margin: 0px;">China accelerates domestic chip development (Huawei Ascend, Alibaba Pingtouge, Biren)</li><li style="list-style: inside none; padding: 0px; margin: 0px;">South Korea, Taiwan, Japan play key roles in the semiconductor supply chain, influencing global power balance</li></ul><h3 style="font-size: 22px; font-weight: normal;">Future Scenarios</h3><ul style="list-style: inside none; padding: 0px; margin: 0px;"><li style="list-style: inside none; padding: 0px; margin: 0px;"></li><li style="list-style: inside none; padding: 0px; margin: 0px;">Multipolar Market → NVIDIA’s dominance will diversify; AMD, Intel, and specialized startups gain share</li><li style="list-style: inside none; padding: 0px; margin: 0px;">Internal AI Chips → AI companies like OpenAI, Google, and Amazon will deploy custom chips for in-house demand</li><li style="list-style: inside none; padding: 0px; margin: 0px;">Edge AI Chips → Explosion in AI chips for smartphones, IoT, and autonomous vehicles</li></ul><h3 style="font-size: 22px; font-weight: normal;">Why Every AI Company Wants Their Own Chip</h3>AI chips aren’t just components—they are core strategic assets.Companies want to:<ul style="list-style: inside none; padding: 0px; margin: 0px;"><li style="list-style: inside none; padding: 0px; margin: 0px;">- Reduce NVIDIA dependency</li><li style="list-style: inside none; padding: 0px; margin: 0px;">- Cut costs</li><li style="list-style: inside none; padding: 0px; margin: 0px;">- Maximize performance</li></ul>In the next decade, AI innovation will hinge not only on software but also on hardware dominance.The 21st-century “new oil war” is really a battle over data and chips.

The Great AI Chip War

#AI #GPT #Chip #Custome #Inference #Cost #NVIDIA #AMD

My idea works in the market?