Nvidia's H20 AI chips surge in demand from Chinese tech giants

How China is challenging Nvidia's AI chip dominance | The US has dominated the global technology market for decades. But China wants to change that | BBC News

AI Chipmaker Cambricon’s Sales Soar 14-Fold With Nvidia Shut Out--- generative artificial intelligence --- Cambricon Technologies Corp. reported a 14-fold surge in quarterly revenue, one of the starkest signs yet of how China’s chipmakers are benefiting from a national drive to replace restricted Nvidia Corp. gear during a domestic AI development boom. The company, which competes with Nvidia as well as Chinese leader Huawei Technologies Co., swung to a net profit of 567 million yuan ($79.6 million) in the September quarter, compared to a net loss of 194 million yuan a year ago, according to a filing to the Shanghai Stock Exchange. That’s off revenue of 1.73 billion yuan. Dubbed “China’s Nvidia” by retail investors, Cambricon is one of a select group of Chinese tech firms that’ve risen to the fore thanks in part to US technology sanctions. From Alibaba Group Holding Ltd. to DeepSeek, the country’s artificial intelligence developers increasingly rely on local alternatives to the Nvidia accelerators essential to training AI, many of which remain off-limits. That’s while Beijing discourages the country’s firms from adopting available options such as the H20. Cambricon Set for Annual Profit Boom on AI Frenzy

Nvidia's AI chip dominance is under siege. Everyone sees competitors emerging. Few understand the strategic chess match unfolding. Here's what it really signals: The surface reading: 📰 What happened: Companies like AMD, Google, Amazon, and China's tech giants are developing alternatives to Nvidia's AI chips 📰 Immediate impact: Increased competition in the AI chip market 📰 Market reaction: Temporary dips in Nvidia's market value when competitors announce breakthroughs The deep reading: 🔍 Hidden signal 1: China's DeepSeek created a ChatGPT rival using fewer high-end chips What it means: Innovation is happening despite export restrictions Historical parallel: ARM vs Intel in mobile computing Outcome then: New architecture dominated an emerging market 🔍 Hidden signal 2: Amazon's Trainium 3 chip developed with Anthropic What it reveals: Cloud providers want to control their AI infrastructure Who benefits: Companies building custom AI solutions Who suffers: Those dependent on general-purpose hardware 🔍 Hidden signal 3: Nvidia announced next-gen Rubin chips with 7.5x performance What it predicts: Nvidia is accelerating its roadmap to maintain leadership The domino effect: Competition heats up → Chip prices eventually decrease → AI deployment costs fall → More companies adopt AI solutions → Democratized AI access across industries Why most miss this: ❌ Focused on: Quarter-to-quarter market share ✅ Should focus on: Long-term ecosystem development ❌ Believing: Any single competitor will dethrone Nvidia ✅ Reality: Specialized chips for different AI workloads will create a diverse market ❌ Time horizon: Next earnings report ✅ Real impact: Five-year industry transformation Those who see it: Building flexible AI infrastructure Those who don't: Overpaying for performance they don't need Do you see what I see?

#IndustryNews 𝗛𝗼𝘄 𝗖𝗵𝗶𝗻𝗮 𝗜𝘀 𝗖𝗹𝗼𝘀𝗶𝗻𝗴 𝘁𝗵𝗲 𝗚𝗮𝗽 𝗼𝗻 𝗡𝘃𝗶𝗱𝗶𝗮’𝘀 𝗔𝗜 𝗖𝗵𝗶𝗽 𝗟𝗲𝗮𝗱 Nvidia CEO Jensen Huang recently warned that China was only “nanoseconds behind” the US in the global AI chip race. The comment underscored how quickly China is advancing—and how seriously it’s challenging Nvidia’s leadership in this critical tech frontier. As the world’s second-largest economy, China is making aggressive strides in artificial intelligence and robotics. Central to this ambition is developing homegrown, high-end chips—the backbone of advanced AI systems. In 2024, Chinese startup DeepSeek made global headlines when it introduced an AI model rivalling OpenAI’s flagship systems... 📖 Click to read the full article: https://lnkd.in/gedXwupz 👉 Follow EmbeddedTech India Expo for more such industry news! 👉 Don’t forget to subscribe to the 𝗖𝗼𝗻𝘃𝗲𝗿𝗴𝗲𝗻𝗰𝗲 𝗡𝗼𝘄 newsletter to get the latest of tech straight to your inbox. . . . #AI #Semiconductors #Nvidia #ChinaTech #EmbeddedTechIndia #Innovation #TechNews #indiatechnews #ConvergenceNow #Convergenceindiaexpo Indrani Priyadarshini

China’s progress in chipmaking is a reminder that you can’t win a marathon by tripping the runner next to you. The U.S. tried to hold the lead by limiting access to Nvidia GPUs, but Huawei’s advances show that restrictions only buy time. Eventually, determined competitors find a way to build their own tools. The lesson here is that staying ahead isn’t just about controlling who gets the shiniest chips. It’s about innovating faster, investing deeper, and recognizing that the global AI race is as much about resilience as it is about raw performance. That should be a wake-up call for U.S. policy makers and companies alike. https://lnkd.in/ghU7AHHk #AI #semiconductors #geopolitics

Nvidia’s Jensen Huang on the Global AI Race: “The U.S. isn’t far ahead of China.” In a candid interview on CNBC’s Squawk Box, Nvidia CEO Jensen Huang offered a rare and balanced view of the global AI landscape — acknowledging that while the U.S. leads in chip design, China is rapidly catching up across infrastructure, energy, and AI applications. Here are five key takeaways from Huang’s remarks “We’re not far ahead.” Huang said the U.S. must adopt a “nuanced strategy” to maintain leadership. While America dominates in advanced chip design (like Nvidia’s Blackwell processors), China’s progress in AI infrastructure and model development is closing the gap. “China is well ahead in energy.” Huang pointed out a less-discussed factor — energy capacity. In 2024, China produced over 10,000 terawatt hours of electricity, double the U.S. output. For an industry fueled by GPUs and data centers, that’s a decisive advantage. “Don’t underestimate China’s chipmakers.” Despite U.S. export restrictions, companies like Huawei and startups building Ascend chips are innovating aggressively. Alibaba and Baidu have begun training AI models on their own processors, moving toward tech independence. “The AI race will be won at the application layer.” Huang warned that under-regulation allows China to adopt AI faster across industries — from manufacturing to consumer apps. “This industrial revolution wins at the AI application layer, at the diffusion layer.” His message to the U.S.? Don’t just build better models — deploy them faster. “We can’t isolate ourselves.” Huang cautioned against limiting American technology to domestic markets: “If America’s tech stack is 80% of the world, we’re winning the AI race. If it’s 20%, we’ve lost.” He stressed that collaboration and global diffusion — not isolation — will determine who truly leads the next era of AI. It’s a nuanced, sobering reminder that AI leadership isn’t just about GPUs and algorithms, but also about energy, adoption, global reach, and ecosystem strategy. Read the full article here: 5 things Nvidia’s Jensen Huang said about the state of the AI race with China — CNBC (https://lnkd.in/eFhTAqA9) Reflection question: Do you think the AI race will be won through innovation and research, or through scale, infrastructure, and global adoption? #Nvidia #JensenHuang #AIrace #ChinaAI #USAI #Semiconductors #AIInfrastructure #ArtificialIntelligence #TechnologyLeadership #Energy #AIecosystem #LLMs #GlobalAI #Innovation #TechPolicy

𝐀𝐈 𝐢𝐧 𝟐𝟎𝟐𝟔: 𝐂𝐥𝐨𝐮𝐝, 𝐂𝐡𝐢𝐩 𝐖𝐚𝐫𝐬, 𝐚𝐧𝐝 𝐭𝐡𝐞 𝐑𝐢𝐬𝐞 𝐨𝐟 𝐒𝐦𝐚𝐥𝐥 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞 𝐌𝐨𝐝𝐞𝐥𝐬⁣ ⁣ As 2026 nears, AI competition is shifting from models to infrastructure, with small language models (SLMs) emerging as game-changers. Here’s the latest as of October 2025:⁣ ⁣ 𝐂𝐥𝐨𝐮𝐝 𝐁𝐚𝐭𝐭𝐥𝐞: AWS (30% market share), Azure (22%), and Google Cloud (13%) are bundling AI models, compute, and storage to lock in customers. Anthropic’s AWS partnership, using Trainium chips, shows the strategy: invest, integrate, dominate. Who will lead this race?⁣ ⁣ 𝐂𝐡𝐢𝐩 𝐑𝐚𝐜𝐞: Nvidia’s Blackwell chips rule high-end AI, but hyperscalers like Amazon (Trainium) and Microsoft (Maia, delayed to 2026) are building custom silicon to cut costs. Intel and AMD are also gaining ground. Can Nvidia stay ahead?⁣ ⁣ 𝐑𝐢𝐬𝐞 𝐨𝐟 𝐒𝐦𝐚𝐥𝐥 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞 𝐌𝐨𝐝𝐞𝐥𝐬 (𝐒𝐋𝐌𝐬): SLMs (e.g., Phi-3, SmolLM) are efficient, 10-30x cheaper, and ideal for edge devices and agentic tasks. They match LLMs in specialized areas, reducing reliance on massive compute and enabling on-device AI for better privacy and speed.⁣ ⁣ 𝐆𝐞𝐨𝐩𝐨𝐥𝐢𝐭𝐢𝐜𝐚𝐥 𝐒𝐭𝐚𝐤𝐞𝐬: China’s 2025 rare earth export controls and U.S. chip restrictions are disrupting supply chains, impacting AI hardware access globally.⁣ ⁣ 𝟐𝟎𝟐𝟔 𝐎𝐮𝐭𝐥𝐨𝐨𝐤: Success hinges on controlling the full stack models, chips, cloud while SLMs drive edge AI and cost efficiencies. Open-source AI and energy costs could shake things up.⁣ ⁣ What’s your take? Will hyperscalers dominate, or will SLMs and new players disrupt? Share below.⁣ ⁣ #AI #CloudComputing #Semiconductors #TechTrends #SmallLanguageModels

Tech giant Nvidia is the world’s leading artificial-intelligence chipmaker, but the company’s success has also put it in the crossfire of trade tensions. The Santa Clara, California-based company, which is approaching a market capitalization of $5 trillion, has seen rapid growth due to its chips, which are predominantly used to power massive data centers used by other tech firms, like OpenAI, the creator of popular AI chatbot ChatGPT.

𝗪𝗵𝘆 𝗔𝗽𝗽𝗹𝗲 𝗜𝘀 𝘁𝗵𝗲 𝗼𝗻𝗹𝘆 𝘁𝗲𝗰𝗵 𝗴𝗶𝗮𝗻𝘁 𝗽𝗹𝗮𝘆𝗶𝗻𝗴 𝗮 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝗔𝗜 𝗴𝗮𝗺𝗲 Last week, Bloomberg shared a great visual on the web of interdependence between the big AI players: NVIDIA, Microsoft, OpenAI, AMD, Oracle, Intel, xAI and others. Noticeably absent: Apple. Not because Apple lacks an AI story, but because it’s outside the AI infrastructure bubble. While the rest of the ecosystem is building billion-dollar data centres and cooling inefficient GPUs with billions of litres of water, Apple is quietly putting extraordinary AI power into your pocket and onto your desk. The new M5 chip delivers a 3.5x AI performance boost and you can already run most open-source models locally on Apple Silicon. No cloud costs, no data sharing, no latency. Just raw, efficient power. It’s a very different philosophy:  • Not proprietary mega-models, but smaller, efficient LLMs linked in agentic networks.  • Not billions in infrastructure, but personal and private computation.  • Not dependency, but ownership of your data and compute. If (or when) the AI bubble bursts, it may simply be the world realising that most AI doesn’t need trillion-dollar infrastructure. Spend 10% of that budget on education or R&D, and the ROI will be far higher. Apple isn’t alone in this shift, but it’s leading it. The new M5 Max/Ultra might just redefine what “personal AI” really means. Perhaps it’s time to start weaning off the cloud and even ChatGPT and see what happens when AI truly becomes yours. #ArtificialIntelligence #EdgeAI #TechTrends #GenAI