NVIDIA

NVIDIA fears that China will hinder the sale of H200 chips, so it is asking for advance payment without exchanges or returns

by

The fact that NVIDIA can market H200 chips in China It’s going around a lot these days and it’s no wonder. And after the Government’s uncertainty about whether it ends up allowing them in the country or not, the company has imposed unusually strict payment conditions for customers who want to buy these chips in China. According to information According to Reuters, the company now requires full payment up front, with no cancellation, refund or configuration changes options once the order is placed. Why it matters. NVIDIA has billions at stake in China, the world’s largest semiconductor market. Chinese technology companies have placed orders for more than 2 million H200 chips valued at about $27,000 each, well above the company’s available inventory of 700,000 units, according to account the middle. But the regulatory situation is a powder keg: the United States has just authorized the sale with a 25% tariff, while China has not yet confirmed whether it will allow imports. Regulation. The Biden administration had banned the export of chips advanced AI to China, but Donald Trump reversed that policy last month allowing H200 sales with the aforementioned 25% tariff that goes directly to the US government. However, China has not yet given the official approval. According to BloombergBeijing plans to approve some imports this quarter, but only for select commercial uses. The military, sensitive government agencies, critical infrastructure and state-owned companies would be left out for security reasons. Protection. The payment terms transfer all of NVIDIA’s financial risk to its customers, who must commit capital without certainty that Beijing will approve the imports or that they will be able to deploy the technology as planned. According to account The average, although NVIDIA has always required advance payments from Chinese customers, deposits were sometimes allowed in lieu of full payment. Now the company is especially strict due to the lack of regulatory clarity. A recent scar. NVIDIA has reason to be cautious. Last year it had to write down $5.5 billion in inventory after the Trump administration abruptly banned the sale of the H20 chip to Chinathe most powerful product that it could then offer there. Although the United States has reversed that decision, China has since banned H20 shipments. This experience explains why the company prefers to ensure collection before any unforeseen regulatory event. Overwhelming demand. Chinese tech giants like ByteDance and Alibaba see the H200 as a significant improvement. This chip, currently NVIDIA’s second most powerful, offers approximately six times the performance of the now locked H20. According to Bloombergboth Alibaba and ByteDance have privately communicated to NVIDIA their interest in ordering more than 200,000 units each. Delivery times. NVIDIA plans to fill initial orders with existing stock, with the first batch of H200 chips expected to arrive before the Lunar New Year holiday in mid-February, according to account Reuters. The company has also approached TSMC to increase H200 production to meet demand in China, with additional manufacturing planned for the second quarter of 2026. The local competition. Meanwhile, NVIDIA’s Chinese rivals are gaining ground. And just as inform Bloomberg, local manufacturers such as Huawei have developed AI processors, including the Ascend 910Calthough its performance still lags behind the H200 for large-scale training of advanced models. On the other hand, Cambricon Technologies It also plans to significantly increase its production of AI chips in 2026, thus expanding its market share and filling the gap left by NVIDIA. What’s coming now. In the coming days it will be known if China makes a final decision on H200 imports. Jensen Huang, CEO of NVIDIA, declared at CES this week that customer demand for H200 chips is “quite high” and that the company has “activated its supply chain” to increase production. Huang also noted that he doesn’t expect the Chinese government to make a formal statement about approval, but rather that “if purchase orders come in, it’s because they can make them.” Cover image | NVIDIA and Arthur Wang In Xataka | There is a new player in the race for the autonomous car and it is one that should worry Tesla a lot: NVIDIA

NVIDIA already has its own Autopilot. And Tesla has reason to worry

by

NVIDIA has presented at the CES 2026 Alpamayo, a family of open source AI models designed specifically for autonomous vehicles. The system not only detects obstacles and plans routes, it “reasons” about complex situations and explains your driving decisions. Mercedes-Benz will be the first to implement it in the CLA, which will arrive in the United States in the first quarter of 2026. Why is it important. Tesla has kept its FSD system completely closed since 2016, and now NVIDIA is betting on releasing the weights of the model, the framework of simulation and more than 1,700 hours of driving data. This strategy can make NVIDIA “the Android of autonomous mobility” and allow any manufacturer to access capabilities comparable to Tesla’s without requiring years of internal development. The contrast: Tesla sells its FSD as a proprietary system integrated only into its cars, generating recurring income from your own clients. NVIDIA wants to sell chips to the entire industry, providing the base technology for others to build their systems. The first model earns more per individual sale, but the second can scale exponentially if multiple manufacturers adopt the platform. In detail. Alpamayo 1 is a 10 billion parameter model that processes video and generates both a trajectory and the logic behind each decision. Jensen Huang has described it as the “ChatGPT moment for physics AI.” The Mercedes CLA will integrate 30 sensors (cameras, radar, ultrasonic…) and will be marketed as a “Level 2+” system, similar to Tesla’s FSD in that it requires constant attention from the driver. Between the lines. NVIDIA’s move seems really good from a regulatory point of view: By generating a “reasoning traceability” that explains every decision, it reassures regulators who are often terrified by black-box models. And by releasing the code, it hooks startups and manufacturers in your CUDA ecosystem. If you can’t develop autonomy yourself (most traditional manufacturers can’t), you just use Alpamayo… and run it on NVIDIA chips. The threat. For Tesla, this means the dreaded commoditization of a technology that has been its main differentiator. If Mercedes delivers FSD-like capabilities in March based on a system that any brand can buy, Tesla’s sales pitch weakens. Elon Musk You have already commented on this announcement on your X profile: “It’s easy to get to 99%, then it’s very difficult to solve the rest.” It also seems like an implicit admission that Tesla hasn’t solved that final problem either. Yes, but. Open source does not guarantee success or similarity with Android in telephony. Actual implementation, integration with specific sensors and validation in real conditions remain complex. Tesla has been accumulating millions of kilometers of driving data for years. NVIDIA offers 1,700 hours, a tiny fraction in comparison. The question is whether that data advantage for Tesla offsets the distribution advantage NVIDIA can get by partnering with multiple manufacturers. Time and the market will tell. In Xataka | If it seems expensive to change the battery in an electric car, wait until you see what it costs in a Ferrari LaFerrari: more than 200,000 euros Featured image | Pixilustration

NVIDIA has paid $20 billion to “license” Groq’s technology. He actually bought it

by

NVIDIA has reached an agreement to “license” assets from Groq and will pay 20 billion dollars for said assets. The company—not to be confused with Elon Musk’s chatbot, Grok—has been designing and manufacturing AI chips for model inference for years. The quotes around “licensing” are important, because this is not a deal: it is a stealth acquisition. what has happened. on Wednesday the news appeared that NVIDIA had agreed to sign a licensing agreement with AI startup Groq. This news was confirmed by those responsible for Groq themselves. on your blogin which they talked about a “non-exclusive license agreement for inference technology to accelerate AI inference on a global scale.” But what both companies say is one thing and what this really is is quite another. How to buy a company without buying it. As part of the agreement, the company’s CEO and co-founder, Jonathan Ross, will go to work for NVIDIA, as will Sunny Madra – its current president – and other senior executives who “will join NVIDIA to help NVIDIA advance and scale this licensed technology.” At Groq they point out that they will continue to operate as an “independent company” led by Simon Edwards, who was their chief financial officer (CFO) and will now become the CEO. NVIDIA keeps (almost) everything. In September Groq raised a financing round of 750 million dollarswhich placed its valuation at $6.9 billion. Disruptive, Blackrock and other companies participated. Alex Davis, CEO of Disruptive, indicated on CNBC that NVIDIA will keep all of Groq’s assets except for one: Groq’s newly launched cloud business. NVIDIA’s biggest “pseudo-acquisition”. This operation is by far the most important for NVIDIA, which bought the Israeli company Mellanox —which designs chips—for $6.9 billion in 2019. In an internal email obtained by CNBC, NVIDIA CEO Jensen Huang explained that “although we are adding talented employees to our ranks and licensing Groq’s intellectual property, we are not acquiring Groq as a company.” The phrase is significant but sensitive, and NVIDIA may want to escape regulators’ scrutiny with this type of pseudo-acquisition. They already made another pseudo-acquisition before. Last September NVIDIA made an identical move by “betting” 900 million dollars by server startup Enfabrica. As in this case, they called to that operation a licensing agreement for its technology, but as in this case what happened is that the CEO of Enfabrica, Rochan Sankar, and other employees, ended up being part of the NVIDIA staff. What is Groq?. Although the name is confused with that of the xAI chatbot, this AI startup does something very different from that model. Groq was founded in 2016 by a group of former Google engineers led by Jonathan Ross and Douglas Wightman. Ross was one of the designers of Tensor Processing Units (TPUs), and Wightman was part of the Google X team and would end up becoming Groq’s first CEO until his departure in 2016. What Groq does. The company has designed AI chips that are specifically specialized in inferring AI models, or in other words, accelerating the execution of those models. While NVIDIA and other companies are especially focused on chips for model training, an equally critical phase, they are not as prepared for inference. Chatbots at full speed. That’s where Groq comes in, who allows extraordinary acceleration of inference and ensure that when we chat with models they “write” at very high speeds. This is when very high token/s speeds are obtained, far above other infrastructures. Not only that, Groq is also cheaper thanks to its specialized chips, so if you want your chatbot to respond at full speed, Groq chips are a fantastic option. How to be a monopoly without saying it. This investment by NVIDIA demonstrates its intention to diversify its business and not stay stuck in its own solutions. The huge operation gives it a major competitive advantage because none of the big AI companies today had focused specifically on inference chips. Groq did from the beginning, and with this “deal” it seems clear that NVIDIA’s dominance in this sector can be strengthened. Is, some analysts saya defensive move rather than a strategic one, and they may be right: Google is getting stronger and stronger with its TPUsand that now Groq is basically part of NVIDIA – although they don’t want to say it that way – will allow it to compete better against the aforementioned Google and the rest of the rivals that are beginning to challenge that dominance. Image | Groq | NVIDIA In Xataka | AMD’s problem is not that it doesn’t make good GPUs for AI. It’s not even close to NVIDIA

A Singapore company has purchased 136,000 AI GPUs from NVIDIA. What is not clear is what he has done with them.

by

In the last three years, an unknown Singapore company has become the largest buyer of NVIDIA chips in Southeast Asia. This singular activity has caused alarms to go off, especially now that the trade war between the US and China means that the “illegal trafficking” of these components is extremely monitored. The suspicion. The company, called Megaspeed, is being investigated by the US government. The objective is to find out exactly if there are ties that unite this company with the Chinese government and if the NVIDIA chips that the company has purchased have ended up in China despite the veto and prohibition that said cards can end up there. The Singapore government is also checking whether Megaspeed has violated local laws, they say. on Bloomberg. Megaspeed denies the major. In a statement sent by mail to that newspaper, those responsible for Megaspeed declare that the company “is based in Singapore and operates fully in accordance with applicable laws, including United States export control regulations.” At the moment there is no evidence. An NVIDIA spokesperson indicates that its request for information from Megaspeed shows no evidence that there was a violation of the terms of those transactions. In their visits to Megaspeed’s data centers they confirmed that “the GPUs are where they are supposed to be.” Furthermore, according to its data, Megaspeed has owners and operates entirely outside of China, and there is no Chinese shareholder. But it does serve Chinese tech giants. Megaspeed has a “neocloud”, cloud infrastructure dedicated to offering computing capacity for AI projects. It has several data centers in Southeast Asia, and the company rents NVIDIA chips to Alibaba. This is an option that the US government does continue to allow: no buying chips, but access to those from suppliers from “non-vetoed” countries. Delicate situation. The question is whether Megaspeed has really done things right or whether it has ended up serving as an intermediary for NVIDIA chips to end up in Chinese technology companies. It would also be disturbing if in the end Megaspeed did have ties to companies or the Chinese government. This discovery comes just as President Donald Trump has stated that he would approve the sale of certain NVIDIA chips to China, something that until now was prohibited. Confusing data. Although Bloomberg admits that they have found no evidence that Megaspeed’s NVIDIA chips have ended up being sent to China, doubts remain. They have analyzed documents with records of commercial transactions, appointments and job offers from both Megaspeed and some of its collaborating companies, and have detected “inconsistencies” between the inventory of chips and those that should really be installed in their data centers. Megaspeed has thousands of NVIDIA GPUs. And the problem is that this company has a huge number of company chips. Since it was founded in 2023 and until November 2025, Megaspeed has imported at least 136,000 NVIDIA GPUs according to Malaysian and Indonesian customs records. More than half are Blackwell chips, which Trump said I would not approve of them being exported to China. Most of those newer GPUs were purchased six months ago, but NVIDIA employees who visited the data centers did not definitively clarify that those that were exported actually ended up where they were supposed to be. The suspicion: a mysterious data center in China. On the Megaspeed website it says that they have three data centers in Malaysia and Indonesia. There is also mention of a room under construction in an unspecified “specific area.” The problem is that Megaspeed showed an image of a render with a data center in Shanghai financed in part by Megaspeed’s original parent company, a Chinese company. Not only that: Megaspeed has a kind of corporate twin in China with an identical website that shows that in reality the employees of the Singapore company are its employees. All of this raises clear questions that remain unresolved and that raise even more suspicions. In Xataka | The US believed it had dealt a mortal blow to China when it deprived it of NVIDIA. He only accelerated one plan: ‘Delete America’

Huawei is not the only one seeking to challenge Nvidia. There are four other “little dragons” knocking on the door

by

“AI” may be one of the words of the year, but “funding round” is a concept that wouldn’t be far behind in the competition. The unicorn is a OpenAI that, if in 2024 it prepared for exceed 100 billion dollarstoday It is bigger than Coca-Cola or Samsung. He has achieved it thanks to money injected by third partiesand Chinese companies want to follow the same strategy as American companies with only one goal in mind: erase the United States from the equation. It’s the ‘Delete A’ plan. Biren. Talking about Chinese artificial intelligence is talking about deepseek and a few other models, but above all hardware companies like Huawei. Their GPUs are the ones that are helping for the Chinese AI field to flourish, and within those GPU companies is Shanghai Biren Technology. As we read in SCMPhas begun a financing round that seeks to raise more than 620 million dollars. Founded by Nvidia and Alibaba veterans, Biren has to his credit BR100one of China’s promises of raw performance to power the demanding data centers needed to train the artificial intelligence. And, unlike others that have opted for Chinese markets, Biren has chosen Hong Kong to attract international capital more easily. They are not the only ones in this race. Moore Threads. If Biren has Nvidia veterans on his team, Moore Threads is directly led by Zhang Jianzhongwho headed Nvidia in China. Perhaps, it is China’s most accurate response to Nvidia itself, and the reason is that it seeks replicate Jensen Huang’s business model combining 3D graphics, for a growing Chinese ecosystem of gamers, and GPUs for AI. To their credit they have the recent architecture Huaganga series that promises 50% more computing density compared to the company’s previous generation of chips, while being ten times more energy efficient. That efficiency is key to keeping AI operating costs at bay, something of vital importance for a China focusing on cheaper artificial intelligencebut functional as soon as possible. And saying that it is Nvidia’s great Chinese rival is not shooting with blank bullets. On the one hand, they are Huashan chips focused on massive clusters of up to 10,000 cards to train LLMs. On the other hand, the chips Lushan that feature hardware ray tracing for the video game market. New Moore Threads GPUs support major gaming APIs little dragons. When Moore Threads debuted on the Shanghai stock market earlier this month, Its shares skyrocketed 500% on the first day, demonstrating that the Chinese market wants to have “its Nvidia”. Biren and Moore Threads are two of the legs of the table. The other two are MetaX (formed by former members of AMD and focused on computing power) and Enflame (a company backed by Tencent and who develop AI systems in the Cloud for Tencent itself). Are known as the “four little dragons of AI” (although other startups are known the same), four of the most promising GPU startups in China that, together with Huawei that has taken giant steps with its AScend 910Dthey have only one objective. “Delete A“Delete the United States. In 2022, when it was still recent the veto of Huawei by the United States in it escalation of the trade war between the US and China, China’s State Assets Supervision and Administration Commission launched Document 79. It was an initiative to encourage the creation of technology that would turn its hardware companies into heavyweights in the global industry. However, there was something else. According to Wall Street Journalthis document has an unusual level of secrecy and an underlying idea: delete United States. Hence the ‘Delete A’ or ‘Delete America’. As? Making all state-owned companies operating in strategic sectors (such as finance, telecommunications, defense or energy) replace foreign software and hardware with domestic alternatives. When? Before of 2027. To do this, national options must be given, and hence the boost to Huawei and startups like these “little dragons.” Although it has also given headaches to companies that have not been able to access Nvidia chips such as Nvidia H20 because they must opt ​​for native solutions, less powerful or optimized in some aspects. Chinese sovereignty. And this development is not just a whim of China, but a necessity. Huawei, Enflame, Moore Threads and Biren, among many others, are on the Entity List of the US Department of Commerce. This prohibits trading with Western companies and access that foreign technology, although more recently the United States has loosened the rope, allowing Nvidia can sell its H200 chips to China… under certain conditions. It is a clear movement resulting from “if China is going to have the technology anyway, let’s take advantage while we can.” And it is because Huawei is working on a open alternative to Nvidia’s CUDA technologythe real ace up the company’s sleeve. Because it is no longer about technical muscle, but about the “language” that the AI ​​speaks. And when China manages to develop this “interpreter”, that is when they will have taken the real leap forward in the development of their tools and in the search for that sovereignty. Images | BirenMoore In Xataka | Big tech is starting to pawn grandma’s jewels for AI: it’s a worrying symptom

Google’s secret weapon against CUDA dominance is called TorchTPU. And it’s an NVIDIA waterline missile

by

Google has launched an internal initiative called “TorchTPU” with a singular goal: to make their TPUs fully compatible with PyTorch. For the not so initiated, we translate it: what Google intends is to destroy once and for all the monopoly and absolute control that NVIDIA has with CUDA. Why is it important. NVIDIA has become the first company in the world by market capitalization for two big reasons. The first, for its AI GPUs. And the second, much more important, for CUDAthe software platform that is used by all AI developers and that has an important peculiarity: it only works on chips from NVIDIA itself. So if you want to work in AI with the latest of the latest, you have to jump through hoops… until now. What happens with Google and its TPUs. Google’s Tensor Processing Units (TPUs) were until now optimized for Jax, Google’s own platform that was similar to CUDA in its objective. However, the majority of the industry uses PyTorch, which has been optimized for years thanks to the aforementioned CUDA. That creates a barrier to entry for other chipmakers, which face a huge bottleneck in attracting customers. Goal is in the garlic. Anonymous sources close to the project indicate in Reuters that to achieve its goal and accelerate the process Google has partnered with Meta. This is especially striking because it was Meta who originally created PyTorch. Mark Zuckerberg’s company has ended up being just as much a slave to NVIDIA as its rivals, and is very interested in Google’s TPUs offering a viable alternative to reduce its own infrastructure costs. Google as a potential AI chip giant. The company led by Sundar Pichai has made an important change of direction with its TPUs, which were previously reserved exclusively for it. Since 2022, the Google Cloud division has taken control of their sale, and has turned them into a fundamental revenue driver because they are no longer only used by Google: Tell Anthropic. A spokesperson for this division has not commented specifically on the project, but confirmed to Reuters that this type of initiative would provide customers with the ability to choose. All against NVIDIA. This alliance is the last attempt to put an end to that great ace in NVIDIA’s sleeve. In these months we have seen how companies like Huawei prepare your own alternative ecosystem to CUDAbut they also participate in a joint effort of several Chinese AI companies for the same purpose. Hardware matters, software matters more. CUDA has become such a critical component for NVIDIA that if other semiconductor manufacturers have not been able to compete with it, it is not because of their chips, but because they cannot support CUDA natively. We have a great example in AMDwhich has exceptional AI GPUs. In fact, they are superior to NVIDIA in certain sections, but their software is not as powerful. In Xataka | Google’s TPUs are the first big sign that NVIDIA’s empire is faltering

Huawei is building its own alternative ecosystem to CUDA. If it succeeds, NVIDIA will have a serious problem

by

When talking about NVIDIA, almost all the focus is on the hardware: the H100Blackwell, racksenergy consumption, nanometers… It is understandable, but it is a mistake. The defensive moat – the moat– NVIDIA is not the hardware. Is CUDA. CUDA is not an add-on to the chip, it is the de facto standard upon which most of the AI ​​code on the planet is written, optimized and debugged. Changing GPUs without changing CUDA does not exist. And switching from CUDA means rewriting years of work. That is why it is a moat. Why is it important. Huawei’s big bet is not to “make a Chinese H100.” It is to build a path for the developer to reach Ascend without feeling like you are changing planets. The restrictions are accelerating it. Exports have split the world in two: An ecosystem that revolves around NVIDIA. And another that China is trying to lift against the clock. In that second, Huawei is not just playing chips: is playing “ecosystem”in AI and outside of it. And therein lies the nuance: you can be years behind in chips and still reduce dependency if you get the software to swallow. In detail. Huawei is attacking the problem on three fronts, with a pragmatically Chinese logic: not to replace everything at once, but to open shortcuts. Native stack (CANN + MindSpore). It is your “pure” alternative: your own environment and your own tools to get the most out of Ascend. The cost today is high, there are complaints of instability, the documentation is rather messy, and the community is much smaller. PyTorch support. This is the most strategic move. Huawei does not try to make the world love its framework– Try to ensure that the world doesn’t have to leave PyTorch. torch_npu acts as an adapter to run PyTorch models on Ascend, but with one problem: it is not native and suffers with every PyTorch change. If PyTorch advances and your backend lags behind, the developer notices. Portability via ONNX. Here Huawei looks for its best window: inference and deployment, not training. ONNX works as a bridge format: you train where you can (often NVIDIA) and deploy to Ascend. It’s a less romantic and more useful approach: if shortages hit, moving inference to local hardware is an immediate relief. Between the lines. The real story is that Huawei is trying to replicate the “trick” that made NVIDIA great: turning its hardware into an experience. That’s why the tactic that explains everything appears: putting engineers in the client’s home to migrate code and optimize it. It is not scalable as a business model, but it is scalable as a transition model: you buy time while you mature tools, libraries and support. And there is another derivative: if China gets enough teams to adopt Ascend out of necessity, over time that can become habit and then infrastructure. Not because it is better, but because it is already integrated. Yes, but. Huawei has two limits that cannot be fixed with marketing: Hardware improvement rate: Roadmap analysis suggests relative stagnation and a gap that could widen, not close, if NVIDIA continues to accelerate cycles. Off-chip bottlenecks: memory (HBM), tools and industrial capacity. You can add “worse” chips, but you need to make a lot of them and build a lot of systems. And now what. If this movie continues, we will see two clear signs: Less hype of chips and more real migration stories: how many computers have moved to Ascendwith what frictions, with what performance losses. Less obsession with training in Ascend and more normalization of the hybrid pattern: I train where I can, I deploy where I must. NVIDIA will continue to be CUDA. Huawei is not “a chip.” It is an escape strategy. And the restrictions are the fuel that is making it inevitable. In Xataka | With HarmonyOS NEXT Huawei has achieved something incredible. Neither Samsung, Microsoft nor Mozilla achieved it Featured image | NVIDIA, Huawei

with smuggled NVIDIA chips, according to The Information

by

The Chinese artificial intelligence startup DeepSeek would have been training his next model with thousands of NVIDIA Blackwell chipsthe most advanced on the market and whose export to China is expressly prohibited by the United States. So The Information states itciting six sources close to the company, who claim that the chips would have arrived in the country through smuggling. ANDl alleged smuggling scheme. According to the media, the chips would have been acquired legally through data centers in countries where their sale is allowed. Once installed and inspected by NVIDIA or its authorized distributors such as Dell or Super Micro Computer, the servers would have been disassembled and the components would have been shipped to China in separate pieces, passing customs under false declarations. This method would allow no trace of the end user to be left. The response of NVIDIA. The company has flatly denied these accusations in a statement: “We have not seen any evidence or received notices of ‘ghost data centers’ built to deceive us and our OEM partners, which are then dismantled, smuggled and rebuilt elsewhere.” NVIDIA adds that, although this type of smuggling “seems implausible,” it investigates any information it receives about it. Why Blackwell chips are so valuable to DeepSeek. NVIDIA’s Blackwell processors began shipping in the final quarter of 2024, with companies like Google, Microsoft, and OpenAI being the first to receive them. These chips include specialized hardware to accelerate sparse computing (Sparse Computing), executing this type of calculations up to twice as fast as traditional methods. According to The Information, DeepSeek would have been using a technique called “sparse attention” that activates only certain parts of the model to respond to requests instead of the entire model, which significantly reduces inference costs. Blackwells would be especially useful for this approach, although their application in larger models is proving more complicated than anticipated. Geopolitical context. US President Donald Trump came to boast to Chinese leader Xi Jinping that Blackwell chips are “10 years ahead of any other chip” and that he would not allow China access to them. However, this week Trump authorized the sale of H200 chips from NVIDIA to China, a generation before the Blackwells, although Beijing is still considering whether to allow its acquisition. Of course, this measure could reduce demand for smuggled Blackwell chips in the Asian country. lThe difficulties of enforcing restrictions. Most NVIDIA chips are manufactured in Taiwan and sold through a complex network of distributors around the world. Jacob Feldgoise, analyst at the Center for Security and Emerging Technologies at Georgetown University account to the media that “the burden of proof to enforce and prosecute chip smuggling cases is quite high. Clear and convincing evidence is needed.” DeepSeek remains silent. The Chinese startup has not responded to the allegations. Previously, DeepSeek had trained its models with older NVIDIA chips: 10,000 A100 units stored by its parent company, hedge fund High-Flyer Capital Management, before US export restrictions took effect in 2022. The company’s research documents from last year indicated that they also had used hopper chipsthe generation immediately before Blackwell. DeepSeek faces several sticks from Washington: in April, the House Select Committee on the Chinese Communist Party published a report calling the startup “a profound threat” to American national security, accusing it of illegally using export-controlled NVIDIA chips. Qregulatory repression. NVIDIA confirmed this week that it has developed a verification technology location through software that could indicate in which country its chips operate, although it has not yet been launched. This tool would use the computing capabilities of your GPUs to monitor the performance and location of the processors. The company has clarified that this is read-only software that does not allow NVIDIA to remotely control the chips or disable them. “There is no off switch,” the company said. Cover image | DeepSeek, Xataka with Mockuuups Studio and NVIDIA In Xataka | If anyone thought that Europe had no role in the race for AI, Mistral has something to tell them

A Chinese startup claims to have created its own TPU to compete with NVIDIA. The only problem is that it is three years late

by

A Chinese startup called Zhonghao Xinying (known internationally as CL Tech) has come to the fore with a bold promise. The company claims to have developed an AI chip that not only circumvents Western intellectual property restrictions, but also outperforms NVIDIA’s A100 chip. Which is very good, but also a little bad. Chana arrives. The chip in question has been named “Chana”, and according to SCMP we are dealing with a GPTPU (General Purpose Tensor Processing Unit). Unlike NVIDIA GPUs, aimed at accelerating AI workloads, this is an ASIC, that is, an application-specific integrated circuit designed from the ground up for neural network workloads. promise. According to Zhonghao Xinying Chana, it offers up to 1.5 times the performance of the NVIDIA A100 based on the Ampere architecture. Not only that: it achieves that performance with 30% lower consumption. The startup highlights that the computational cost per unit would therefore be less than half of that offered by the A100 chips. A little history of the company. Behind Zhonghao Xinying is Yanggong Yifan, an engineer formed at Stanford and the University of Michigan. He worked on the development of several generations of Google TPUs and also on the development of Oracle chips, and in 2018 founded this startup in Hangzhou together with Hanxun Zhengan engineer who worked at Samsung for several years. They were joined by other engineers from Microsoft, Oracle, NVIDIA, Amazon and Facebook, they indicate. on Baidu. We are therefore faced with several of those cases of “boomerang talent” with Chinese engineers who are forged in the US and then return to China to create solutions for their own industry. Solutions that do not depend on the West. Yanggong affirms that its chip features “fully self-controlled IP cores, a custom instruction set, and a fully in-house computing platform. Our chips do not rely on foreign technology licenses, ensuring long-term security and sustainability from an architectural perspective.” But. Although the achievement is striking, it is necessary to put it in perspective. The NVIDIA A100 is a 2020 AI GPU, and even with the improvements that this Chinese startup promises, its performance is, for example, far from H100 chips with Hopper architecture that appeared in 2022. Not to mention of the latest Blackwell Ultra chipswhich are currently NVIDIA’s greatest exponent in terms of AI chips. There are also no details about who makes the chip, and one of the candidates it would be SMICwhich has 7nm technology. They are very far away, and they have another problem. The technical achievement of these engineers is certainly notable, but everything indicates that they are still far from what NVIDIA and its competitors are achieving. like AMD or Google with its recent TPU Ironwood. There is another element that works against them: Chinese manufacturers continue without having direct access to the most advanced photolithography on the market, and although it also there is progress from Chinese manufacturers in that sense, competing is certainly complicated without access to the most advanced technologies. Pressure. In 2024 the company achievement revenues of 598 million yuan (73 million euros) with a net profit of 85.9 million yuan, but in the first half of the year the income was only 102 million yuan and had losses of 144 million yuan. The firm has reached an agreement with its investors by which it will have to go public at the end of 2026, or else it will be forced to buy back shares. The financial pressure is therefore notable for the company, which must demonstrate in the coming months that its roadmap is truly competitive. In Xataka | China was no longer supposed to be able to get its hands on NVIDIA’s most advanced chips. Until he found a shortcut in Indonesia

In a financial carom, Google has stood up to NVIDIA, leaving an unexpected winner in the crazy AI race: Larry Page

by

NVIDIA promised them very happy being the best-positioned AI chip manufacturer. At least it was until Google has started making chips. This new scenario has excited investors, who have rushed to buy Alphabet shares, making your price goes up up to 6.3% from one day to the next, and accumulating an advance of more than 75% since its August price. This increase in the value of Google’s parent company has also coincided with a dip in Oracle’s valuation, which has caused chaos on the podium of the world’s largest fortunes. according to Forbes. What AI gives you, AI takes away. A few months ago, Larry Ellison, founder of Oracle rose as the second largest fortune in the world, overtaking Mark Zuckerberg. His fortune reached 291.6 billion thanks to the good growth prospects posed by the construction of the data centers for AI. In fact, the Oracle founder’s fortune grew so much that he was close enough to the unattainable Elon Musk as to threaten its position on that list. Just as AI raised Larry Ellison to become the world’s second-largest fortune, AI he has taken that place away to hand it over to Larry Page, who reaches that position with a fortune of 261.5 billion dollars. Google rises, Oracle falls. He Google stock rally contrasts with the downturn suffered by the main architect of the cloud infrastructure in which AI lives, leaving up to 6.79% of its price in recent days. This decline has meant that Ellison’s fortune, with a strong influence of Oracle on its income balance, has suffered, falling to $256.7 billion, being displaced to third position. That same stock market momentum of Google has taken another founding partner, Sergei Brin, to fourth position, with a fortune of 242.4 billion dollars, while Alphabet shares brought the company closer to a market capitalization of almost 4 billion dollars. Mark Zuckerberg and Jeff Bezos didn’t even see it coming. The most pronounced falls in recent months have been those of Jeff Bezos and, above all, Mark Zuckerberg, who, accustomed to remaining in the Top 3 of the greatest fortunes, fall to fifth and sixth position in the ranking of Forbes. The decline in Mark Zuckerberg’s fortune is especially striking, due to the poor performance of Meta shares in recent weeks. Interestingly, Meta shares have broken their downward trend following Google’s announcement to get into the semiconductor business for AI and the rumors that Zuckerberg could change NVIDIA processors for the Tensor Processing Unit manufactured by Alphabet. Larry Page and Sergei Brin: same company, different fortunes. Although Page and Brin co-founded Google and share control of the company through their shares, both millionaires do not own exactly the same number of shares, and that detail makes a big difference in their assets. According to public statements of Alphabet before the US Securities and Exchange Commission (SEC), between the two magnates they concentrate 87.9% of Alphabet’s class B shares, which grant 10 votes per title. However, the figures show that Page has just over 389 million shares, while Brin account with some 362.7 million of these shares, which makes Page the main beneficiary of the rally in the shares of the company they founded. Brin has been more generous with science. The key to this gap is that Sergei Brin has been much more active than Page in donating and selling part of his stake in Alphabet, and that has reduced his share package over time. Brin has been targeting large volumes of Alphabet and Tesla shares to research donations of treatment against Parkinson’s disease, bipolar disorder or autism, after being discovered a genetic mutation which made him prone to developing that disease. In Xataka | Larry Page and Sergey Brin founded Google and became millionaires. Now they are dedicated to collecting gigantic airplanes Image | Flickr (Fortune Global Forum, TED Conference)

Log In

Forgot password?

Forgot password?

Enter your account data and we will send you a link to reset your password.

Your password reset link appears to be invalid or expired.

Log in

Privacy Policy

Add to Collection

No Collections

Here you'll find all collections you've created before.