NVIDIA

DeepSeek is good, pretty and very cheap. And above all, the weapon to create a Chinese hardware industry independent of Nvidia

by

The arrival of DeepSeek-V4-Pro It hasn’t caused that much of a stir. like the one caused by DeepSeek R1 a year and a half ago, but we may be facing an even more important model. If that version revealed to the world that China was advancing spectacularly in this race, this other one is beginning to allow us to glimpse something else more interesting. What most people see is a very decent model and above all “low priced”. Which hide the company It’s another more important thing: achieve independence from Nvidia and US hardware. what has happened. Last Friday, those responsible for DeepSeek announced something surprising: their promotional offer with a 75% price cut to use their DeepSeek-V4-Pro model will be maintained permanently. That makes this model offer very decent features (but not exceptional) for a really low price: 1M entry tokens 1M tokens output DeepSeek-V4-Pro 0.435 0.87 GPT-5.5 5 30 Opus 4.7 5 25 Gemini 3.5 Flash 1.5 9 Good, pretty and very cheap. It is true that the performance of DeepSeek-V4-Pro is inferior to that of rival models from OpenAI, Anthropic or Google. Artificial Analysis tests indicate that the DeepSeek model is at a very good level, but it is also much cheaper than its competitors. This is especially relevant for agentic tasks that consume many tokens and that with this model become accessible and very affordable. According to Artificial Analysis, DeepSeek is close to the performance of the best models in the industry, and although it is slower in its responses, it is also much cheaper than the frontier models from OpenAI, Anthropic or Google. A different strategy. How is this company going to make money? It does not have subscription plans like its local competition (GLM, Kimi) or the western one (ChatGPT Plus, Claude Pro). It also does not have voice or image models. It does not have an AI agent for programming that competes with Claude Code. It publishes the open weights of its models and shares its technical innovations with the industry (and with its competitors). For those who closely follow the company and these decisions, the strategy is clear. DeepSeek’s goal is not to win the AI ​​model race. Their goal is to build a Chinese AI hardware industry that doesn’t depend on Nvidia or TSMC… and get paid their share in that process. Hardware independence. China has a structural problem in this AI race: sanctions and vetoes imposed by the US make you unable to access the most advanced chips nor to ASML UVE photolithography. And since China cannot currently compete in terms of computing power, what its companies are doing is ensuring that their AI models need less computing power to achieve similar results. Efficient architectures. The Mixture of Experts (MoE) and Multi-head Latent Attention (MLA) architectures are two key weapons in this strategy. The first already existed but was adapted by DeepSeek for their model: with it only part of the total parameters of the model are activated to answer the query without losing precision. What MLA does is compress the attention information (the so-called KV Cache) with which the model maintains the context of a conversation, reducing it by 90%. Both techniques allow us to reduce the need to use high-speed HBM memories, something that is also striking in order to reveal DeepSeek’s probable strategy. The importance of KV Cache. As the GDP analyst explains in Xthat use of MLA allows that for one million tokens, DeepSeek-V4-Pro only needs 5.48 GB of HBM memory. Competitors like Zhipo AI, which develops GLM 5, need 60 GB for the same, while Alibaba’s Qwen 3 needs 89 GB. This advantage allows DeepSeek to offer much lower prices to obtain performances similar to those of its competition, but it also means that DeepSeek models can run on Chinese memory chips that cannot compete in speed with HBM modules. Goodbye HBM, hello NAND and SSD. These innovations open the door to the use of NAND memories and even SSD drives to process this data, and there YMTC enters the scenea Chinese Flash memory manufacturer that is slowly becoming a global giant. Also CXMTwhich manufactures DRAM memoriesbecomes an alternative here and the reason is equally interesting: DeepSeek introduced a memory search module in LLMs called Engram which is also intended to avoid excessive dependence on HBM memories. How to bypass the CUDA monopoly. Nvidia continues to have a fundamental element in CUDA to maintain its market dominance, but here DeepSeek too has proposed an alternative. Is called Tile Kernels and these are software cores created with TileLang (a variant of Python for this field) that allow governing advanced AI chips (GPUs). Huawei as an invisible ally. Those responsible for Huawei recently indicated that its new Ascend AI supernodes fully support DeepSeek v4 models. Precisely this provides another fundamental advantage to the company, which thus avoids (at least in part) total dependence on the use of Nvidia chips and prepares to further strengthen Huawei’s relevance in a market in which until recently Jensen Huang’s company was queen and mistress. Open models to attract the hardware industry. US companies continue to maintain their closed and proprietary models, but DeepSeek is one of the many Chinese startups that publish them with open weights. With this, what she and the others intend to do is not only attract AI developers and users, but also create a hardware ecosystem that adopts these architectures. DeepSeek invites its rivals to use techniques such as MoE or MLA precisely so that all these advances become a de facto standard and hardware manufacturers also adopt them and integrate them in an optimized way into their designs. A round of 10,000 million to advance. The company is also preparing a financing round in which they intend to raise 10,000 million dollars and with which they would achieve a valuation of between 45,000 and 50,000 million dollars. Still far from the mammoth valuations of OpenAI or Anthropic (already close to a billion dollars) but certainly … Read more

There was a time when Nvidia was a gaming company. That business is now pocket change for the owner and lady of AI

by

In 1993, Nvidia was founded with the goal of creating graphics chips for video games. For almost three decades Nvidia has been basically that: a semiconductor company for gaming that yes, I had ambition in the field of professional computing. But things change: Nvidia’s gaming business has generated $6.4 billion the first fiscal quarter of 2027and although it is a healthy business, for Nvidia it is something else: It’s almost pocket change. Gaming no longer (almost) matters. In any other company in the sector, this income (29% more than last year) would already be extraordinary, but at Nvidia they are almost a footnote, because gaming represents less than 8% of the company’s total income. The other $75.2 billion came from the data center business, which grew 92% from the previous year. AI has made Nvidia’s original business almost irrelevant in relative terms. Stratospheric numbers. Nvidia has earned $81.6 billion in the first fiscal quarter of 2027. It is an absolutely colossal figure that should be put into perspective: it is so large like GDP from Croatia, Panama or Uruguay. The company led by Jensen Huang has managed to grow 85% in revenue since a year ago, almost double. The surprising thing is that it has also done so when it seems increasingly difficult to grow at this rate. The graph shows year-over-year growth in revenue in percentage. In 2026 the trend is bullish again. Source: FT. This is non-stop. The company exceeded Wall Street expectations, which projected revenues of 78.86 billion, but Nvidia also states that its forecast for the next quarter is to earn 91 billion dollars, 12% more than the current one. It’s true that growth is slowing in percentage terms, which is normal at this point, but in absolute terms the company continues to add billions of dollars of additional revenue each quarter. Data center numbers. Those $75.2 billion in data center business aren’t just GPU sales for hyperscalers. It also includes the company’s networking solutions business, which has grown no less than 199% year-on-year to $14.4 billion: it has tripled. The reason is logical: the demand for interconnection infrastructure for the large clusters that are being created everywhere is enormous, and Nvidia provides an ideal solution for those who buy its AI chips. Beware I, Anthropic is coming. On the call with investors, Jensen Huang gave a singular fact: Anthropic has made virtually no use of Nvidia solutions to train and serve its AI models, but that is going to change. The company’s CEO highlighted that the computing capacity they are going to deploy for Anthropic this year and next is going to be “quite significant.” Or what is the same: they are going to continue selling like hotcakes even if the competition tightens. Nvidia is also an investor in startups. Nvidia’s strategy is also being curious on a financial level, because it is not content with growing its business, it is betting on AI startups. It has invested more than 26,000 million in investments in this type of companies, and that does not include the recent agreements with OpenAI or in listed companies like corning. Beware II, China is coming. All these numbers, attention, are being achieved without the help of the Asian giant. In December, the Trump administration authorized Nvidia chip exports to China (with a 25% government fee). Theoretically that should make Nvidia generate notable income thanks to said authorization. Huang explained that at the moment these revenues are zero and that there is some uncertainty about whether China will finally allow its chips to be imported. In the second fiscal quarter of 2027, income from China is not assumed, but if that market finally opens, we will have even more extraordinary numbers. Buying back shares. Nvidia has returned about $20 billion to shareholders this quarter between buybacks and dividends. The board of directors has approved investing $80 billion more in share buybacks, thus multiplying by four what had previously been authorized. That’s a clear sign of Nvidia’s confidence in its future, which will also benefit shareholders: the dividend has passed from $0.01 per share to $0.25 per share. Previously, Nvidia offered specific data on gaming revenue. From now on, stop doing so to put that division within the Edge Computing category. Gaming no longer appears in the accounts. Typically Nvidia’s financial reports divided revenue into data centers, networking, gaming, professional visualization, automotive, and a few other fields. Now that Nvidia is a fully AI-focused company, it has changed its revenue pooling structure. Everything related to gaming, PCs, consoles, workstations, robots, cars and other devices is part of the “Edge Computing” category. Gaming, we insist, no longer (almost) matters. In Xataka | For the first time in 30 years, Nvidia will not present new GPUs for gamers in 2026. They earn much more with AI

Chinese Big Tech can now buy Nvidia GPUs. The problem for Nvidia is that they don’t need it now

by

The United States and China are immersed in a trade and technological war that has caught the line of fire to the AI ​​giant: Nvidia. The situation is that Nvidia must prioritize AI companies from the United States to guarantee the supremacy of this country, but as a company it would be interested in taking a bite out of the giant Chinese market. And the problem is twofold: it has not been able to do so for a long time due to trade vetoes, but now that it seems that it can sell its famous H200 to China, it turns out that China has turned the page. More or less. green light. Nvidia has gone from having a monopoly on AI GPUs in China to have a 0% quota. These are the words of the CEO, Jensen Huang, and the reason is the aforementioned trade restrictions between the powers that prevented Nvidia from selling its most powerful products to the Asian giant. Huang has spent months insisting on Donald Trump’s government to allow them to sell with a very clear logic: China is going to develop its alternatives and what better way to make a profit until then. The situation is gone relaxing at the end of last year and at the beginning of this to get to the point where we are now. According to Reutersthe US Department of Commerce already allows ten Chinese companies and distributors such as Foxconn and Lenovo acquire that long-awaited H200the company’s second most powerful AI chip. Good news for the company. Or they should be if it weren’t for the fact that the Chinese industry is going its own way looking home. Alibaba, ByteDance, JD.com and Tencent are the Chinese giants that can supposedly already buy H200. Up to 75,000 chips each, to be exact. However, it is noted that they have not yet made any shipments. Here there is a mix between very restrictive bureaucracy and, above all, that emphasis on national development. Tencent, for example, noted in September last year that they had no intention of producing AI chips, but that they were going to invest a lot of money in domestic partners. For example, they are in the process of adapting their infrastructure to be able to connect Huawei’s Ascend platform (particularly the Ascend 950 series) as the main training tool for large models. A few days ago, Tencent’s strategy director already pointed out that that strategy was still in place and that the company expects a significant increase in spending on AI GPUs designed in China. Manufacturing at home. Alibaba and Bytedance have a different approach. If Tencent is focusing on acquiring Huawei platforms, Alibaba and Bytedance are looking to create their own chips. Alibaba seeks to be the most powerful RISC-V chip created to date and it was reported that Bytedance wanted Samsung will manufacture its processor. In the end, whether buying from Huawei or developing the tool internally, the two approaches respond to the great national objective: that at least 50% of the data centers that belong to the State use at least 50% Chinese integrated circuits in their servers. That is one of the great Chinese technological impulses of recent years, one of the crucial points of the Five-Year Plan for the development of the country and, above all, the strategy that Nvidia had been warning the United States about for some time. The age of inference. Because this period of ostracism to which the US condemned China has served for the country to develop three very clear alternatives to Nvidia and encourage companies that are already working with models to develop their own hardware. This is important especially in the new AI framework we are entering, that of inference. Although the AI ​​will continue to train and GPUs will be needed for this, the next step is inference, the agentic era in which the processor or CPU is very important. AMD is moving there, same as Intel or ARMand precisely processors are something that Huawei is good at and in which the Chinese giants can shine as much as their American counterpart by developing chips tailored to their models and needs. Also, as pointed out in CNBChaving your own chips means you don’t have to fight with anyone else in a time when there is scarcity and, of course, if you don’t have to buy from an outsider, there is an improvement in the gross income margin. juicy cake. And this leaves Nvidia in that uncomfortable situation, one in which it wants to participate, but in which it seems that it is no longer needed as much as before. Because China is developing its chips for this new era of AI and Nvidia is running into a final boss called bureaucracy and the pressure groups of the ‘Make America Great Again‘. The first is due to the slowness of the export order processes, something that takes months when orders should be much more agile. The second are the aforementioned pressure groups that hold that any deals Nvidia makes with Chinese companies are less chips for American companies, something that should not be allowed. Meanwhile, Chinese companies are developing their alternatives and Huawei wants to flood the market with 750,000 chips this year, three times more than its shipments in 2025, and Nvidia is falling short of a $50 billion pie. In Xataka | The US has the best AI models. China has something else: AI too cheap to care about

For the first time in 30 years, Nvidia will not present new GPUs for gamers in 2026. They earn much more with AI

by

In 1995 Nvidia presented its NV1 chipsits first multimedia card and the one that would start its particular revolution in the world of gaming. Since then, every year the company has presented a new model intended for this segment. In 2026 that tradition will be broken. what has happened. What has happened is AI. The rise of this industry has been of such magnitude that it has had a critical impact on the technological field and, little by little, on the social field. Nvidia is at the center of this particular revolution, because the company bet early on the ability of its GPUs to be used as AI chips and that bet has been rewarded. Gamers in the background. Such has been the explosion in this field, that Nvidia has decided that what is important is no longer gamers, but AI chips for data centers. From a financial and business point of view, the logic is overwhelming: the profit margin of AI chips round 75%, especially thanks to price control that allows the company to set prices to its liking thanks to the fact that it currently has almost no competition. Data centers win by a landslide. There is another element that favors it: volume. Not only is the price per unit higher, it is that the volume handled in data centers is much higher than that of gaming GPUs. Analysis like that of App Economy show how the market started timid, but in the second fiscal quarter of 2024 revenues began to skyrocket and the data center fever has made Nvidia the company with the largest market capitalization in the world. No GPUs for gaming in 2026. After the launch of the RTX 5000 in January 2025, this year Nvidia was expected to announce the “SUPER” versions of said family. These models they were going to tell with denser GDDR7 memory modules, which would allow the memory configuration of the original models to be increased. The memory crisis and the total focus on the catalog of GPUs for AI has meant that Nvidia has not announced them, and for the first time in more than 30 years there will be no renewal of the gamer catalog for this year. And the RTX 6000 even further. If the news is already bad for the SUPER versions of the RTX 5000, things are even more terrible for the theoretical RTX 6000, which will have Rubin architecture and from which a notable jump in performance is expected. According to the latest datathese graphics cards will not begin to be manufactured until the end of 2027, which would mean that they would not arrive until 2028. The current situation suggests that it is likely that they will not even arrive that year. Do we really need more powerful GPUs? On Reddit a user did an important comment when it became known that NVIDIA would probably not release new graphics for gamers. “On the one hand it makes me angry. On the other hand I realize that I am playing ‘Rimworld‘ and ‘Terraria‘”. It refers to very popular games that can be played even with integrated GPUs such as those used by many Intel or AMD processors. Others they responded that Nvidia GPUs are so powerful that they are actually often necessary because game developers don’t really squeeze the hardware. Be that as it may, it seems that the current generation is usually more than prepared for the most demanding titles, and the urgency for a new generation is perhaps not so pressing. The April 2026 Steam survey makes it clear that the next-generation RTX 5000 coexists with a market in which the RTX 4000 and RTX 3000 remain very popular. Source: Steam. The data confirms it. If you go through the April 2026 Steam Survey you see how more than a year after its presentation, the RTX 5000 has almost 24% market share, while the RTX 4000 has 35% and the RTX 3000 has 16%. The rest of the users opt for previous solutions or from rivals like AMD, which is still far away in this battle. Many users have already invested in their RTX 3000 and 4000, and it seems unlikely that they will do so again for a new GPU, especially when in the recent times The prices of these cards have skyrocketed. There is nowhere to run. There is another problem with this Nvidia strategy of turning gamers into second-class users: there are not many alternatives, at least if we want maximum performance. AMD continues to fight in this market, but its graphics still fail to capture the interest of many users. Intel has done interesting releases recent years, but not in the high range in which Nvidia is a de facto monopoly. Your efforts They are not achieving great success either.and the company is not focusing on it either because it knows that now the money is somewhere else. In Xataka | If at some point NVIDIA has to choose between giving its best chips to the US or China, its choice is very clear.

Nvidia and Samsung are the names of AI. Quietly, someone is eating up the server processor market

by

Artificial intelligence is about to enter a new era. After plunder the internet and drink all human knowledgetraining is no longer the obsession of the big AI companies and the inference is about to take the baton. That inference will reach its climax with the explosion of AI agents and that implies a change in balance: GPUs will continue to be key, but CPUs will take on a greater role. Inference requires other types of resources other than training and that is why Nvidia is preparing with its platform Vera Rubinbut also the rest of the industry. Intel has already said that is moving its production lines towards the Xeon, the server processors, while ARM It is seeing green numbers because a few months ago it presented a powerful processor for AI. The one that is also seeing the ‘stonks’ grow is AMD. Although its name sounds less than that of Nvidia, AMD is very present in the AI ​​race. It has secured the best memory for its new platform, it has a GPU for training and also the processors EPYC for servers. These are precisely the ones that are giving you joy. AMD EPYCo record According to analysts Mercury Researchboth ARM and AMD have had a spectacular quarter. Both have continued to eat market share from Intel (which is why it seeks to respond) and, in the case of AMD, in the first quarter of this year they have reached a record 46.2% revenue share in x86 CPUs for servers and 30% of the CPU segment. Here are two numbers to keep in mind. The first is that AMD was already coming from a fairly comfortable position, with a 41.3% revenue share in servers in the last quarter of 2025. Thus, it seems that this growth to 46.2% is not too big, but the second number that must be taken into account is the one that allows us to see the company’s leap in this segment. It is estimated that the company It had only between 1% and 2% CPU share for servers in 2018. Since then, AMD has been doing things very well both in consumer computers with their Ryzen as in servers with its EPYC, which has allowed it to eat Intel’s share by leaps and bounds. And just as important as the quota are the company’s results, not because they interest us in terms of money, but because it gives us an idea (just like what is happening with Nvidia, SamsungSK Hynix or Micron) of how far we are from being able to see competitive prices again in the consumer market. Because it is estimated that this part of the business focused on data centers left 5.8 billion dollars in AMD, an increase of 57% year-on-year. It surpasses Intel (5.1 billion), being the first time this has happened in the data center sector and, in addition, AMD projects a growth of more than 70% year-on-year in the data center segment. In this particular battle, we have already commented that Intel is not sitting idle and has new processors for data centers, a great projection being the great american foundry and we will have to wait to see the efforts to reconvert their production lines to return to the Xeon. What is evident, according to estimates, is that the server processor market is experiencing an impressive increase due to this new generation of AI and is wait that goes beyond 30,000 million in 2025 up to 170 billion dollars by 2030. Landing this for us, the users, this implies one thing: if it was already expensive to build a PC due to RAM and SSDnow other components such as processors or motherboardswho are also reorienting the business. In Xataka | The US confesses its worst nightmare: if China invades Taiwan and controls TSMC, the US economy will collapse

China and Nvidia star in the “great technological divorce” of 2026. A bureaucratic hell that is erasing it from the market

by

Talking about Nvidia is talking about artificial intelligence glue. The GPU giant has invested millions financing cocompanies like OpenAI or Anthropicbut along the way has not forgotten startups or to make purchases for strengthen your position in the market. The problem is that it is missing out on a potential $50 billion market: China. Because Nvidia is eager to enter China, but it is trapped between bureaucracy, the Trump Government, Xi Jinping’s Government, and the smuggling of its graphics cards. The great divorce. In a very short time, Nvidia has gone from dominating the Chinese GPU market for artificial intelligence to losing it completely. The restrictions of the Trump Administration and the intensification of the trade war between the powers left Nvidia out of the game. Either it would adapt its GPUs and create less capable versions of those it sold in the West or it would not be able to sell in China. For a time, Nvidia was selling the H20 to adapt to the new rules, but it is something that has taken its toll. As AI needs demanded more powerful GPUs and own chinese industry with Huawei, Cambricon and Moore Threads was developed, Nvidia was being left out of the game. Official quota. In the middle of last year, Nvidia CEO Jensen Huang pressed Donald Trump to see reason: it was better for Nvidia to be able to enter China both to make money and to slow the accelerated development of the domestic industry, one that Western restrictions had given wings to. In the end, the US gave in previous tariffs of 25% and one condition: all GPU orders from Chinese companies to Nvidia would be reviewed one by one. There is a problem: the US body in charge of reviewing these export licenses has decreased by 20% in recent months, which is causing delays of months when it comes to fulfilling an order. From when a Chinese company asks for Nvidia GPUs until they are given an answer, the ‘chinese dragons‘They have already released some product. The result? Huang points out that Nvidia has gone from being a leader in China to have a 0% quotapainting the situation as a true drama and pointing directly to the strategies of both China and, above all, the United States as the cause of his company falling into the offside of the large Asian market. Furthermore, it is China itself that encourages its companies to, to the extent possible, use Chinese hardware that they is developing at accelerated rates. “Official” fee. But the fact that Huang claims that his market share in China is 0% does not mean that there are no GPUs for AI in China because it seems that there are H100, H200 and even B200 due to something very simple: smuggling. Despite the proprietary technological solutions they are developing, it is evident that a large part of the AI ​​industry is built with Nvidia GPUs and that implies that the tools are very well optimized for them. There are several occasions in which Nvidia AI chip smuggling networks have been reported, with modest seizures on occasions (just tens of millions of dollars) and somewhat larger seizures on others (hundreds of millions in a few months). Chinese companies obtain these chips through indirect routes from Hong Kong and Singapore and, although Nvidia tries to trace the origin, the clandestine flow and opaque chains make the task complex. trapped. Someone is lining their pockets and that someone is not Nvidia. And the problem is that Huang’s pressure had an effect, but the solution they gave him is not as agile as the market needs. Returning to the issue of bureaucracythe United States Office of Industry and Security, which is responsible for reviewing these export licenses, reduced its workforce by 19% in 2024. Specifically, those who develop standards linked to the semiconductor industry and review licenses have decreased by 20%. The result is an average of 76 days to resolve export requests, something that is extending so far this year and which is disastrous news for both Nvidia and others deeply involved in the AI ​​segment, such as AMD. From China, things are not much better, since companies must make it very clear why they need Nvidia AI chips and cannot meet their objectives using national alternatives. Jensen, almost excluded. In any case, it is evident that Huang does not like to be missing the AI ​​party in China, in the same way that he is going to miss the new trip of Donald Trump and other executives to a summit between Trump and Xi Jingping that will be held between the 13th and 15th of this month. Or so it seemed. This is an event in which conversations will focus on agriculture and commercial aviation, so a priori Jensen didn’t have much in mind. But of course, alongside Trump are CEOs like Elon Musk, Cristiano Amon or Tim Cook, among others. And, although it seemed that he was not invited, as we see in South China Morning PostIn a message from Trump on his social network, it was confirmed that Huang will finally accompany him on the trip. In the end, it’s about money. Jensen Huang doesn’t want China to have the best chips because He wants to save those for the United States.but it is a very large market in which Nvidia can offer chips strategically: it makes money while making companies opt for its product instead of that of the Chinese companies themselves. In Xataka | Nvidia’s superpower is not having money, it is making everyone work for it: Foxconn is the latest to join

Corning has the solution to accelerate Nvidia chips even more

by

There are good reasons why a company of Nvidia’s stature would want to collaborate with a company like Corning, specialists in manufacturing the glass that protects our mobile phones. Corning offers more products than your Gorilla Glassand that is precisely what Jensen Huang’s company is interested in. And it is that Nvidia is going to invest about 3.2 billion dollars at the glass manufacturer with the intention of multiplying the optical connectivity production capacity on US soil tenfold. What’s on the table. The financial scope of the agreement has been revealed in parts. It was initially announced that Nvidia would receive warrants (stock purchase rights) to acquire up to 15 million Corning shares at a price of $180 per share, representing a potential investment of up to $3.2 billion. Added to this is a pre-financed warrant for another additional 500 million. But the CEO of Nvidia confirmed on CNBC that the company has also made “a prepayment of several billion dollars” to finance the construction of the new factories, a figure that was not part of the initial official announcement and whose exact amount has not been made public either. Fiber optics are the thing. The data centers that power AI They house hundreds of thousands of GPUs that must communicate with each other continuously and at high speed. For decades, this communication has been carried out using copper cables, and in fact for short-distance connections within the rack (from the server to the switch), they are still used, but fiber optics end up being superior in everything, both in terms of speed, energy consumption and lower signal loss. Which Nvidia has in mind. The technical term at the center of this agreement is co-packaged optics, which refers to the integration of glass fiber directly into chip systems, progressively replacing copper cables. Inside Nvidia rack systems (such as the Vera Rubin) there are currently about 5,000 copper cables that could be replaced by Corning fiber optics. Already at last year’s GTC, Huang rated this technology “essential for the deployment of AI.” The company has been preparing the ground for months: in March invested $4 billion in Coherent and Lumentumtwo companies specializing in lasers and components that convert data between light and electrical signals, which then travel through Corning fiber cables. Who else is in the race. Nvidia is not the only one betting on this technology. Its competitors Broadcom and Marvell They have already launched similar productsand Intel also develops its own co-integrated optics solutions. For its part, Corning already had Meta as a reference client. In fact, Zuckerberg’s company announced an agreement of up to 6 billion dollars for Corning to expand its optical cables plant in Hickory, North Carolina. The alliance with Nvidia now adds three more facilities and multiplies the company’s optical connectivity manufacturing capacity in the United States by ten, in addition to increasing its fiber production by more than 50%. ““Made in America”. The agreement comes precisely at a time when the Trump administration is pushing to relocate technological supply chains that have been built for decades in Asia (Taiwan, China or Vietnam). Huang counted to CNBC that “it is an extraordinary opportunity to reinvest and revitalize American manufacturing for the first time in generations.” According to the CEO of NVIDIA, the tech sector would not be the only one to benefit, since the construction and operation of these data centers generates demand for electricians, construction workers, chip manufacturing operators and infrastructure specialists. “The skilled worker shortage and demand are incredibly high,” mentioned Huang in the middle. Converted company. Corning has become another of those companies that have seen their business benefit from the AI ​​boom. And the signature accumulates an increase of more than 300% in the last year, driven by its repositioning towards the AI ​​market and moving away from its best-known image as a manufacturer of glass for mobile screens. In Xataka | If the question is whether using ChatGPT or Claude in English is more efficient and saves tokens, the answer is: yes

The CEO of Nvidia believes that we are in a new industrial revolution where AI will not replace us: it will micromanage us

by

Artificial intelligence has been available to users and companies for a few years now and we are at a point where they converge several ideas about AI and the future of work. There are several open fronts such as if AI will replace usif it will only be a tool or if, instead of freeing ourselves from the workload we carry, will add more to us. But the CEO of Nvidia, a Jensen Huang who has no trouble spilling his tongue, has another opinion. AI is going to micromanage us. Micromanager. A few days ago, Huang attended a talk at Stanford Business School. At these events, company CEOs usually leave motivational messages and talksbut I don’t know if in this case it would motivate someone who is looking for a job. During his panel, the Nvidia boss commented that, right now, “we are doing things faster, on a larger scale and we can think to do things we never imagined.” That part of the speech is fine, but he went on to note that “AI agents will harass you, micromanage you, and you will be busier than ever.” Like a good 1st century Roman baptisterywho wouldn’t like having an AI agent egging you on? Will create more jobs. Lately, Huang has chosen to blurt out headlines and vaguely elaborate. At the event, he also commented that these agents we have help us explore new avenues of work, do that work better and make it more profitable. He also addressed the great controversy, that of the supposed great replacement. On this, his opinion is that there will be some jobs that will be redundant because AI will be able to do the same as a human, but he considers that, in general, there will be humans with new jobs to adapt to. “I think we are going to create more jobs. There will be more people working at the end of this industrial revolution than at the beginning of it,” he says. Insecurity. It is curious that you compare it with the industrial revolution at a time when there is concern, above all, about the instability of the labor market. Huang ha commented that computer engineers are busier than ever and it makes sense, the problem is what happens next and what is happening with all those who are not dedicated to tasks strictly related to AI. In an article by Fortune published a few weeks ago, the issue of layoffs directly related to artificial intelligence was addressed. An example is Jerome Powell, president of the United States Federal Reserve, who warned that AI is quietly impacting the labor market as job creation is practically at zero. Another is that of Dario Amodei, CEO of Anthropic, who believes that “entry-level” jobs will be reduced by half in the next 18 months. And then Microsoft’s AI chief, Mustafa Suleyman, predicting that AI will cause many white-collar jobs to collapse in that same time frame. AND Meta is going to do without 8,000 employees as it transforms into an AI company. All this while, on short video networks there is a lot of content of young people saying that they have a university degree and are rejected at Target or McDonalds. The AGI has already arrived. Well no. HE esteem that, during 2025, some 55,000 people in the US will lose their jobs directly due to AI. It is only 4.5% of all layoffs, but a significant number that, if forecasts are met, will multiply by several figures over the coming months. For now, so far in 2026, esteem that technology companies have laid off 92,000 people, not all of them must be related to AI, but a scary number if we take into account that, during 2025, the total was 120,000 people. Just 28,000 less in just four months. But, beyond that, the prediction that an AI agent will not take our jobs, but rather will be a tiresome second boss, is not the only thing that Huang has commented recently without going much further. A few weeks ago, on Lex Fridman’s podcast, he already commented on things like that workers must be clear about the purpose of their work and that the tasks and tools they use to do it are related, but they are not the same. Also He commented that we had already arrived at the AGI (artificial general intelligence) giving an example that it has nothing to do with an AGI that, for now, remains theory. A black hole of money. Byan Catanzaro is the vice president of deep learning at Nvidia and has commented that AI currently costs more than human employees. “For my team, the cost of computing far exceeds that of employees.” It must be taken into account in this that AI is not an abstract entity: it is a huge investment in hardware, data centers and energy. According to the calculations According to Keith Lee, professor of AI and finance at the Swiss Institute of Artificial Intelligence, AI expenditures will be $5.2 trillion by 2030 in a conservative estimate and $7.9 trillion in a more aggressive one. But more interesting is what he comments about the fact that fixed subscriptions are not making money for companies because they do not cover operating costs. And that, at a time when companies like OpenAI and Anthropic should not take long to go public, is something to take into account because they will stop receiving millions from other private companies to have to respond to investors with their product and benefits. In Xataka | There are programmers from Meta and Microsoft competing to be the one who uses the most AI and wasting millions of dollars along the way

Europe is taking its technological independence so seriously that it is aiming for the most ambitious goal: NVIDIA

by

Europe cannot continue to be the technological vassal of the United States. With that powerful message, the CEO of Mistral presented a few days ago a roadmap with which he considers that Europe can take the pulse in the technological race of artificial intelligence. The warning came just when several companies are defining the future of European technological sovereigntyand one of those companies is Euclyd. It is seeking 100 million euros, is backed by one of the ASML bosses and has a clear objective: to stop depending on NVIDIA. And it’s not the only one. Euclyd. We have already talked at length about ASML. Although when we talk about the technology industry we have names like Intel, TSMC, NVIDIA or Qualcomm more present, ASML is the Dutch company that manufactures the most advanced machines for manufacturing semiconductors. Without it, the technology industry would not be what it is to the point that China is investing everything in having its own ASML. Well, Bernardo Kastrup is the former director of ASML and, in 2024, he founded Euclyd. This startup is backed by former ASML CEO Peter Wennink, and, according to CNBCis looking for financing to raise the necessary capital to start mass manufacturing chips. 100 times more efficient than NVIDIA. In this new round of financing, Euclyd is seeking $100 million and the goal is to create inference chips for AI. These chips are designed so that the models use what they learned in the training phase and are optimized for high speed, low latency and, above all, much lower energy consumption than the training ones. And that is where the ambitions are maximum. Euclyd, based in Eindhoven, claims that its ‘Craftwerk’ chip system is 100 times more energy efficient for AI inference than NVIDIA’s Vera Rubin chips. This is very good, but the comparison is a bit bulky because Vera Rubin, which is the new generation from NVIDIA, is not a pure training or inference platform: it is optimized to do both. European movement. But hey, Euclyd is currently raising the money with an eye toward delivering inference chips to its first two customers by 2027. And it’s not the only one. There are others such as the British Olix, Optalysys and Tactile, the French Lago or the Dutch Axelera that have raised more than 800 million euros to date. That is from the private sphere, since Europe has the FAMES pilot program which has 830 million euros to finance this type of projects. It is an extremely modest amount if we take into account what is moving on the other side of the pond, but between financing chip companies, renewables and European data centersis a sign that the feeling that Europe must fend for itself is there. world movement. The interesting thing is that this does not respond only to Europe’s feeling of technological sovereignty. It goes further, pointing to the great whale of AI: NVIDIA. Whatever company we think of, surely part of its hardware – or all – belongs to NVIDIA. own Mistral reached a very juicy agreement with the company led by Jensen Huang to be able to acquire thousands of GPUs, but the industry is already seeing what happens when all the eggs are in the same basket. That is why NVIDIA has its potential greatest rivals among its clients. Goal, tesla either amazon They buy from NVIDIA, but at the same time they are developing their own chips. The Chinese giants want NVIDIA chips, but they also develop their alternatives with local companies. All of this is creating more shadowy companies such as Texas Instruments, Marvell or Broadcom to do business, since they are the ones those who turn to They do not want to depend so much on NVIDIA. Google. In fact, just as startups developing AI chips are appearing in Europe, in the United States an ecosystem of companies is developing that are raising billions of dollars. Two examples They are Cerebras Systems, which is valued at 23 billion or MatXfounded by former engineers from Google’s TPU development team. Google itself, whose TPUs are manufactured by Broadcomthis searching an agreement with Marvell to diversify its inference chip business. NVIDIA responds. There is a phrase that has always made me laugh, that of “you think the police are stupid”, and applies perfectly here. NVIDIA has also been realizing for some time that it must diversify and has stopped injecting obscene amounts of money to only a few companies to go on to support other smaller onesbut promising. This way you get clients in the curious circular AI financingas well as continuing to be the one who leads the segment. But in addition to investing in others, she invests in herself. In March, he invested 4 billion a photonics company to make optical interconnection systems for next-generation data centers. They are also investing more than 18,000 million in R&D and winning juicy contracts with both TSMC as with Samsungwho make the chips for the company’s AI platforms. In the end, if all markets have something in common, it is unbridled spending. Europe, China and the United States have embarked on a race in which there is no end in sight and that will perhaps have its greatest test when Anthropic and OpenAI go public this year. In Xataka | Europe thinks that it is the one who wants to become independent from US technology companies. It’s actually the other way around.

raise the rent of your NVIDIA

by

GPU prices are through the roof. I’m not talking about AMD’s RX 9000 or NVIDIA’s RTX 5000, since those are for gamers. I am referring to the GPUs that, suddenly, They are the only ones that matter: GPUs for AI hyperscalers. The Big Tech of AI They have paralyzed the entire consumer market coppressing production of the few component manufacturers that exist in the segment and causing a brutal shortage. Good luck if you want to buy an SSD or RAM, but it is also impacting companies. Valve can’t release the Steam Machine and Apple just remove option from Mac Mini and Mac Studio with the largest amount of RAM. Simply put, either there isn’t one… or what there is is tremendously expensive. And the irony is that this situation is starting to impact the AI ​​business itself where some now have to pay to rent NVIDIA GPUs at almost double the price. It is the GPU as a service model. Sky-high prices for cloud GPUs Here we must differentiate between hyperscalers and AI companies that do not have their own facilities. Amazon, Microsoft, Meta, Tesla or Google, among others, are hyperscalers. They build gigantic data centers which they fill with tens of thousands of GPUs (which are usually from NVIDIA, since it is the one that dominates this market) to meet their needs. In them they carry out the training and inference tasks of their models, but some have turned to become service providers. Amazon Web Services, Google Cloud and Microsoft Azure They maintain a parallel business, that of huge NVIDIA GPU lessors. They buy huge lots of H100, H200 and A100 that they integrate into their infrastructure and simply rent their computing capacity to whoever is willing to pay the price. It’s like cloud gaming itself NVIDIA with GeForce Now: A company that has an interest in AI, but cannot build a data center, has the possibility of pay to rent that computing capacity to large tenants. So far so good because it is a win-win for all parties, but the problem comes when scarcity hits. On this playing field There are not only Google, Microsoft and Amazon. There are other companies more focused on the cloud GPU business, such as CoreWeavewhich a few months ago already increased rental prices by 20%. It coincided with the early stages of the RAM and SSD crisis. And the price increase was not the only change. From the previous year of permanence contract, the requirement increased to three years. In an article by business insider we can see more clearly the price of this demand exceeding supply. Carmen Li He is CEO of Silicon Data, an analysis firm, and has commented that NVIDIA’s veteran H100s have risen 20% in the last three months, from $2.20 per hour to $2.64. The B200 are along the same lines: from $4.40 per hour to $5.35. He problem comes with the H200, since rental price increases of 48% are being experienced here. From $2.75/hour a couple of months ago, they have gone to $4.08/hour. It’s almost double for the same product. because those who need it most want even more power for their latest models since so much money is being injected into this sector that more and more companies without data centers need more computing power for their products. The component manufacturers for these GPUs they can’t meet the excessive demand, which is causing Waiting times for new chips of between 36 and 52 weeks and, therefore, since there is not a GPU for everyone, cloud computing rental prices… increase. Between the big three and Meta more than 650,000 million will be spent in AI infrastructure this year and Carmen Li point that, since this demand for AI exceeds any expectation, not only is there not enough for everyone, but the old GPUs sold by hyperscalers When they renew equipment they depreciate very, very little.. In the second year of use of an H100, it can be sold for 85 cents. In the third year, for 84 cents. According to several voices in the sector, it is a tsunami that has already taken the consumer market by storm, but that with the rise of Agentic AI it will get worse. Because it is no longer just training and basic inference, but agents that execute several steps autonomously, consuming more computing capacity per request than traditional queries to a chatbot. Translation: that a market that some they point that should be stable is becoming something like electricity or energy, a roller coaster of prices that plays with the rules of the savage capitalism. In Xataka | Using Netflix in 2018 was much better than now: we have normalized degrading experiences

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.