Exclusive: Nvidia's H20 chip orders jump as Chinese firms adopt DeepSeek's AI models, sources say
SINGAPORE/BEIJING, Feb 25 (Reuters) - Chinese companies are ramping up orders for Nvidia's (NVDA.O), opens new tab H20 artificial intelligence chip due to booming demand for DeepSeek's low-cost AI models, six people familiar with the matter said.
The surge in orders, which is being reported for the first time by Reuters, underlines Nvidia's dominance of the market and could help alleviate concerns that DeepSeek might cause a slide in AI chip demand.
Tencent (0700.HK), opens new tab, Alibaba (9988.HK), opens new tab and ByteDance have "significantly increased" orders of the H20 - a chip specific to China due to U.S. export controls - since the Chinese AI startup burst into the global public consciousness last month, two of the people said.
In addition to their internal needs for advanced AI chips, the three tech giants provide cloud computing services through which other firms can access and use AI tools.
Smaller companies in sectors like healthcare and education are also purchasing AI servers equipped with DeepSeek models and Nvidia H20 chips, said a source at one of China's largest server makers. Previously only deep-pocketed financial and telecoms firms bought servers with AI computing systems, the source added.
U.S. President Donald Trump's administration is looking at imposing restrictions on the sale of the H20 chip to China, Reuters has reported. While the threat of further controls could be a factor in the jump in orders, the sources cited DeepSeek as the reason.
The sources did not provide details on the size of the orders. They were not authorised to speak to media and declined to be identified.
Nvidia did not respond to queries on how much demand for the H20 it was seeing from China but said its products won "on merit in a competitive field". The company is set to report quarterly earnings on Wednesday.
Tencent, ByteDance and Alibaba did not respond to requests for comment.
DeepSeek's large language models rival Western systems in performance at a fraction of the cost as they focus on "inference" or producing conclusions. That optimises computational efficiency rather than relying solely on raw processing power.
"When DeepSeek launched, many misjudged that computing power demand might stagnate or decrease. In reality, more advanced AI models drive deeper integration into daily life, exponentially increasing inference-level compute need," said Nori Chiou, investment director at Singapore-based White Oak Capital Partners.
A DeepSeek-induced global rout in AI stocks that began January 24 saw Nvidia shares lose as much as a fifth of their value at one point but they have since regained most of that ground and are down just 3% for the year to date.
Though wider deployment of DeepSeek AI models is expected to help Chinese chipmakers such as Huawei(HWT.UL) better compete in the domestic market thanks to the models' focus on inference, Nvidia's H20 chip remains the industry standard in China.
Analysts estimate Nvidia shipped approximately 1 million H20 units in 2024, generating over $12 billion in revenue for the company.
The H20 is the primary chip Nvidia is legally permitted to sell in China and was launched after the latest round of U.S. export restrictions took effect in October 2023. Washington has banned exports of Nvidia's most advanced chips to China since 2022, concerned that advanced technologies could be used by China to build up its military capabilities.
Numerous Chinese companies have announced plans to use DeepSeek's models. Among them are Tencent, which has said it will beta test integrating the models into its highly popular WeChat messaging app, and automaker Great Wall (601633.SS), opens new tab which has integrated DeepSeek's model into its connected vehicle system.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Daily Mail
39 minutes ago
- Daily Mail
Trump's tariff czar reveals the four countries on verge of trade deals…including a 'nasty' one
President Donald Trump 's top trade negotiator shared that the U.S. may be close to finalizing tariff deals with several major partners in coming weeks. U.S. Trade Representative Jamieson Greer told CNBC that an imminent deal could be struck with the E.U. after Trump threatened increasing the 'nasty' union's rate to 50 percent earlier this month. Greer also shared that he has been in nearly constant contact with India 's top trade representative, claiming to have calls with them 'every day.' In addition, the trade rep. revealed that he plans on forwarding negotiations with Malaysia and Vietnam at the upcoming Organization for Economic Co-operation and Development (OECD) meeting next week. He also said he will be talking to E.U. representatives at the gathering. Though not all news has been positive; China, Greer said, has 'violated' a trade agreement made with the U.S. in Geneva earlier this month - echoing a major complaint highlighted by Trump the same day. A handshake agreement between the world's two largest economies came earlier this month and was widely seen as a way to tamper tensions between the pair. China was hit with a tariff rate in excess of 145 percent earlier this year before the agreement, but the rate then came down to around 30 percent. Trump said he expects to talk to Chinese President Xi Jinping during an Oval Office press conference with DOGE leader Elon Musk on Friday. 'I'm sure that I'll speak to President Xi and hopefully we'll work that out, but its a violation of the agreement,' he said. The president said he is sure they will talk as the countries remain at odds over trade. Earlier in the day the 78-year-old commander in chief sent off a salvo on Truth Social lambasting China for violating the Geneva deal. 'Two weeks ago China was in grave economic danger! The very high tariffs I set made it virtually impossible for China to trade into the United States marketplace which is, by far, number one in the world,' his post began. 'We went, in effect, cold turkey with China, and it was devastating for them. Many factories closed and there was, to put it mildly, 'civil unrest,'' it continued. 'The bad news is that China, perhaps not surprisingly to some, has totally violated its agreement with us.' The agreement made between Greer, Treasury Sec. Scott Bessent and some top Chinese officials in Geneva in May stipulated that the two countries would unwind tariffs and trade restrictions on certain critical minerals. Greer accused China of slow-walking that process during his interview Friday. 'The Chinese are slow-rolling their compliance, which is completely unacceptable and it has to be addressed,' he stated. Soon after, a spokesperson for the Chinese embassy in D.C., Liu Pengyu, hit back at the claims. 'Recently, China has repeatedly raised concerns with the US regarding its abuse of export control measures in the semiconductor sector and other related practices,' they said in a statement. 'China once again urges the US to immediately correct its erroneous actions, cease discriminatory restrictions against China and jointly uphold the consensus reached at the high-level talks in Geneva.' The spokesperson's complaint comes after a Reuters report that the U.S. ordered some companies to stop shipping goods to China without certain licensing. Products ranging from semiconductors, chemicals, software and aviation equipment are among the restricted items, sources shared. The president announced a 90 day pause on some tariffs earlier this year, with many steep rates coming back into effect this summer if U.S. trade partners do not sign deals. Vietnam was hit especially hard by the White House's 'Liberation Day' tariff announcement in April. On a massive board showcasing the rates to be imposed held aloft by Trump on stage, Vietnam was slapped with one of the highest tariffs at 46 percent. Malaysia, meanwhile, was hit with a 24 percent rate under Trump's reciprocal tariff plan while India faces a 27 percent rate. Earlier this week, Trump was dealt a devastating blow, albeit a temporary one. A three-judge panel at the U.S. New York-based Court of International Trade ruled that parts of the president's tariff regime was unlawful. The ruling argued that Trump had wrongfully invoked an emergency law to justify the levies. Erupting with frustration, some in the White House vented that the move was akin to a 'judicial coup' against the president's executive authority. The pause was short-lived, though, after an appeals court issued a stay on the panel's ruling that have allowed the tariffs to remain. The appeals court demanded to hear from the Trump administration and the litigating parties, Democratic-led states and a group of small businesses. The case remains ongoing, though the White House has multiple tools at its disposal to implement tariffs in other ways if the court rules against it.
Geeky Gadgets
an hour ago
- Geeky Gadgets
Deepseek R1-0528: The Open Source AI Model That Could Topple Big Tech Giants
What if one model could dismantle the dominance of AI giants like OpenAI and Google? That's exactly what some are saying about Deepseek R1-0528, the open source disruptor that's sending shockwaves through the artificial intelligence industry. With performance that rivals proprietary titans like GPT-4 and Gemini 2.5 Pro—at a fraction of the cost—Deepseek isn't just another player in the game; it's rewriting the rules entirely. But this isn't just about cheaper AI. It's about a deeper shift in power, accessibility, and the very future of how we innovate. Could this be the moment where open source finally topples the closed gates of Big Tech? Wes Roth explores how Deepseek R1-0528 is reshaping the AI landscape and what it means for you. From its new cost efficiency to the ethical dilemmas it raises, this model is more than just a technological achievement—it's a statement. You'll discover the methods behind its performance, the growing divide between open source and proprietary AI, and the geopolitical tensions fueling this revolution. Whether you're a tech enthusiast, a business leader, or just curious about the future of AI, this story has implications that reach far beyond the tech world. After all, when the foundations of an industry shake, everyone feels the tremors. Deepseek's Disruptive Impact Performance That Matches Industry Leaders Deepseek R1-0528 has demonstrated its ability to compete with some of the most advanced proprietary AI models available today. Benchmark tests reveal that it outperforms Google's Gemini 2.5 Pro in specific tasks while maintaining parity with OpenAI's GPT-4 (03) in others. However, what truly sets Deepseek apart is its cost efficiency. By offering significantly lower API costs, it makes high-performance AI accessible to a broader audience, including small businesses, independent researchers, and startups. This affordability has the potential to provide widespread access to AI adoption, allowing organizations with limited budgets to use innovative technology. For you, this means access to powerful AI tools without the financial barriers often associated with proprietary systems. Open source vs. Proprietary AI: A Growing Divide The rise of Deepseek R1-0528 highlights the increasing divide between open source and proprietary AI models. Open source solutions like Deepseek challenge the business models of closed-source companies by delivering similar capabilities at a fraction of the cost. This shift could lead to a more equitable distribution of AI technology, allowing smaller players to compete with industry giants. However, it also raises concerns about the sustainability of proprietary research and development. Proprietary models often rely on substantial funding to push the boundaries of innovation, and the growing popularity of open source alternatives could disrupt this ecosystem. For you, the choice between open source and proprietary AI may come down to balancing cost, performance, and long-term support. Deepseek R1-0528 Just Broke the Entire AI Industry Watch this video on YouTube. Take a look at other insightful guides from our broad collection that might capture your interest in Deepseek. How Deepseek Achieves Its Performance The methods behind Deepseek R1-0528's impressive performance have sparked significant interest and speculation. Experts suggest that the model may have employed knowledge distillation, a technique where synthetic outputs from proprietary models like Gemini are used to train new models. This approach enables open source models to replicate the performance of their proprietary counterparts without requiring direct access to their datasets. Additionally, bioinformatics tools have been used to analyze AI outputs, providing insights into model lineage and training strategies. These innovative methods demonstrate how open source models can achieve high performance while operating within resource constraints, offering a blueprint for future development in the field. Global AI Competition: The U.S. and China Face Off The release of Deepseek R1-0528 underscores the intensifying competition between the United States and China in AI development. Historically, the U.S. has been a leader in software innovation, but China's growing focus on open source AI and its advanced hardware manufacturing capabilities could shift the balance of power. For instance, China's expertise in semiconductor production may provide it with an edge in deploying AI systems at scale. Meanwhile, U.S. policymakers are exploring strategies such as R&D tax incentives to maintain their competitive advantage. For you, this rivalry could influence the availability, affordability, and direction of AI technologies in the years to come, shaping the tools and platforms you rely on. Ethical and Strategic Challenges The rise of open source AI models like Deepseek R1-0528 brings both opportunities and challenges. On one hand, open source AI aligns with the goal of providing widespread access to access to advanced technologies, making them available to a wider audience. On the other hand, it introduces risks related to AI safety and misuse. For example, the accessibility of powerful AI tools could lead to unintended consequences, such as their use in malicious applications. Governments and private organizations in both the U.S. and China are under scrutiny for their roles in AI development. Critics question whether concerns about safety and regulation are genuine or driven by geopolitical motives, adding complexity to the ethical debate. For you, understanding these challenges is essential to navigating the evolving AI landscape responsibly. The Future of AI Innovation As global competition in AI intensifies, the stakes for innovation and collaboration continue to rise. Open source models like Deepseek R1-0528 have the potential to redefine the industry by making advanced AI capabilities more accessible and affordable. For researchers, developers, and businesses, this represents an opportunity to innovate without being constrained by proprietary systems. However, the broader implications for national security, economic policy, and ethical standards remain uncertain. The balance between fostering innovation and addressing potential risks will play a critical role in shaping the future of AI. For you, staying informed about these developments will be key to using the opportunities and addressing the challenges that lie ahead. Media Credit: Wes Roth Filed Under: AI, Top News Latest Geeky Gadgets Deals Disclosure: Some of our articles include affiliate links. If you buy something through one of these links, Geeky Gadgets may earn an affiliate commission. Learn about our Disclosure Policy.
Daily Mail
an hour ago
- Daily Mail
Joe Biden 'got lost in his own closet at the White House while staff hid mental decline'
Former President Joe Biden got lost in his own closet at the White House, a Secret Service whistleblower claimed to MAGA Senator Josh Hawley. The Missouri Republican shared the intel with Fox News host Sean Hannity Friday, adding to the ongoing speculation that Biden's administration worked to cover-up his declining mental state. 'He [the Secret Service member] told me that Biden used to get lost in his closet in the mornings at the White House,' Hawley explained, adding that his source was assigned to Biden while he was the president. 'I mean, the guy literally stumbling around in the White House residence couldn't find his way out of his own closet. The president of the United States. I mean, this is outrageous. We were lied to.' Hawley said the alarming claims were made to him while he was investigating the attempted assassination of President Donald Trump during his campaign rally in Butler, Pennsylvania. He went on to criticize the ongoing scandal that the Biden administration worked to cover-up his cognitive decline, specifically with the use of an autopen - a device that automatically replicates a person's signature. 'We need to find out who actually signed off, so to speak, on all those autopen signatures and all of those pardons and all of those clemencies,' Hawley said, adding that it amounts to 'one of the worst constitutional crises of our country's history.' contacted Biden's team for comment. The senator's comments come just after Biden announced he was diagnosed with Stage 4 prostate cancer, and amid the mounting backlash of CNN anchor Jake Tapper's new book that claims to blow the lid off of the White House campaign that hid Biden's health decline from the American people. His office revealed his frightening diagnosis on May 18, adding that it is an 'aggressive form' of prostate cancer. The cancer has metastasized and spread to his bones, a statement on the 82-year-old's diagnosis revealed. The announcement came just one week after reports revealed that a 'small nodule' was found on Biden's prostate that needed 'further evaluation.' Just as his health battle was revealed, Tapper's new book, 'Original Sin: President Biden's Decline, Its Cover-Up, and His Disastrous Choice to Run Again,' co-authored with Axios' Alex Thompson, hit the shelves. Early reviews of the book, which was released May 20, were quickly overshadowed by allegations that Tapper himself was part of the cover up and is now seemingly trying to rewrite history. People have gone on to label Tapper a fraud and 'phony' over his self-righteous rants about the former president's health cover-up, claiming that his role as an anchor at a left-leaning network helped conceal the alleged secret. 'He was and still is down for the cause. We are not fooled or amused,' a user wrote. 'No retractions, no apologies, and no consequences for those who pushed a false narrative for years,' another stated. Someone else commented: 'Bullsh**, he's lying. He knew all along that it was bullsh** and still carried the democrat party's water.' While facing a slew of backlash for his new book, Tapper recently sat down with Thompson on Stephen A. Smith's show Wednesday to address the heat he's been facing. Smith specifically asked the pair what they think about people calling the book a 'money grab,' leading Tapper to say that 'very few people write books to make money.' 'No, our only agenda was to find out what happened. Our only agenda was to write this book,' Tapper replied. 'In terms of the left it's a money grab, I mean that's not the case. It's not - this is, I don't want to get into where my money comes from. You know, CNN is my main job and that's where my main source of income, by far... Very few people write books to make money,' he babbled on. 'You write a book because you want to say something. You want to have a legacy, you want to uncover something, you have a message you want to get out there,' Tapper stated, adding 'I don't think people write books about Joe Biden to make money' as Thompson smiled. 'That's a good point,' Smith chimed in. On Tuesday, President Donald Trump's daughter-in-law Laura Trump revealed that Tapper called her to apologize for a condescending on-air dispute in 2020 where he dismissed her concerns about Biden's mental acuity. Laure revealed the phone call on Fox but said his apology was 'a little too late' -conveniently taking place just weeks before the publication of his tell-all book about the very topic. One of the most prominent examples of this was back in October 2020, when Lara pointed to a speech of Biden's and said: 'What we see on stage is a very clear cognitive decline.' Tapper immediately interjected, savaging her with a haughty response that she had 'absolutely no standing to diagnose somebody's cognitive decline' and accusing her of instead 'mocking his stutter.' While Tapper has publicly apologized for his response and admitted a sense of 'humility' over being proven wrong, Lara has now revealed he reached out to her privately two months ago. 'Jake Tapper called me about two months ago and he said "I have this book coming out,"' she told Fox News host Laura Ingraham. Tapper told her 'I know everybody is saying I should apologize to you. I plan, when the book comes out, to go on TV and say you were right and I was wrong.' While she appreciated the acknowledgment and that Tapper has since stuck to his word by apologizing publicly, she noted that 'it feels a little bit too late to me.' On Friday, Biden hit back at Tapper's book revelations after a female reporter asked about his mental capabilities. 'You can see that I was mentally incompetent and I can't walk and I can beat the hell out of both of them,' he joked to a female reporter, leaning down to speak to her after speaking at a service in honor of Memorial Day. The former president down played revelations from the new book and chatter from Democrats that he should have never run for re-election, saying: 'Why didn't they run against me then? I could have beaten them.' 'I don't have any regrets,' he added. The remarks were the first ones Biden made since his cancer diagnosis. His voice sounded hoarse as he spoke to reporters at Veterans Park in Wilmington, Del. He attends the ceremony to honor veterans every year. He and his family are optimistic about his diagnosis, Biden said, adding he was taking pills as part of his treatment. 'The expectation is we're going to be able to beat this,' he said. 'It's not in any organ.... my bones are strong. I feel good.'
