token

For decades, companies have accumulated capital and talent. Satya Nadella Thinks They Need Something Else Now: Token Capital

by

The reflections of the CEOs of large companies must always be taken with some caution since they are not usually simple reflections thrown into the air. They seek to give someone their ear: investors, rivals, users or to calm the waters among its employees. Satya Nadella, has published in their profiles of social networks a text in which he redefined, without much dissimulation, what a company should be in the era of artificial intelligence. It was not just a statement, but rather it almost pointed to become a manifesto in which the focus of the race was changed by develop the best AI model (an area where Microsoft seems to accept the Copilot’s defeat) and defines that the future of companies involves generating a third pillar to their structure: token capital. The company of the future needs a new type of capital: token capital. Nadella does not conceive AI as a substitute for human employees since, according to his thesis, human capital is the basis for turning AI into a truly disruptive tool that is nourished by knowledgethe criteria, the relationships and the reading of patterns that employees provide. However, it brings a new element to the business equation: token capital. This new element is formed by the AI ​​that a company builds and owns, not in the models that third companies rent to it. Currently, most companies use AI as a subscription service in which they pay for a model that they use to perform tasks with it. However, when they stop paying for it, all that knowledge and evolution is lost and the company retains nothing of all the knowledge. time and resources you have invested in fine-tuning its use. Nadella maintains that this path leads to transferring value to a few suppliers and the only ones who accumulate advantage are those who sell the models, the knowledge of each company ends up being the raw material that feeds others. The loop that becomes active. The Microsoft CEO’s idea revolves around what he calls a “learning loop”: a system that feeds back with each decision made and each workflow completed. That is, it is a knowledge base that makes the company’s memory permanent and not lost when changing the AI ​​model or employees. “This cycle becomes the company’s new intellectual property,” highlights the Microsoft CEO. “I look at it as a hill-climbing machine.” The key is that this asset, unlike what happens today, is evolutionary and is built based on training with real company data and internal measurements of its response. The more you use it and tune it, the more value it has. And, Nadella argues, the company that builds it will soon have something that can’t be bought in any AI model marketplace: a tool that has been “trained” to do a very specific job in a custom context. AI as a tool, not a monopoly. There is a paragraph in Nadella’s statement that is striking coming from the CEO of a company valued at three trillion dollars. Nadella compares the current risk with what happened in the first phase of globalization: entire industrial sectors were emptied by outsourcing. The macroeconomic figures of the countries they endured the loss of industrial fabric, but the social fabric ended up suffering by adding tension in the labor market. His warning leaves no room for interpretation: “If all value is concentrated in a few models, political economy simply will not tolerate it. There is no social permission for an AI future that destroys entire industries.” The goal, he says, has to be an ecosystem where each company can build its own learning, not be another cog in an AI monopoly. Actually, this is not new, since it is the same principle with which Microsoft built its platform business in Azure cloudwhich used Microsoft’s infrastructure for companies to generate more value than the platform itself had. The problem that the manifesto does not solve. However, Nadella’s words also raise a series of contradictions with respect to the latest movements of Microsoft and other large technology companies. The CEO maintains that human capital becomes essential as token capital grows since it is the employees who make a company’s AI learn. However, his own company has been half doing the opposite. Microsoft fired to more than 15,000 employees during 2025, and in April 2026 it offered voluntary retirement packages to some 8,750 workers in the US, something it had not done in its 51-year history, linking these layoffs to your commitment to AI. It is not an exclusive case of Microsoft. In the first quarter of 2026 they are already more than 92,000 layoffs among the employees of large technology companies and the argument that all companies repeat is the same: AI allows us to do more with less people. In Xataka | Jensen Huang enters the Samsung salary controversy: “Workers should earn as much as possible”

“token factories” with their own hardware

by

The AI ​​race is no longer explained just by looking at which company launches the most powerful model or who gets access to the most advanced chips. That part is still important, but it doesn’t tell the whole story. What is beginning to emerge is a much broader dynamic to ensure the necessary resources to continue competing. China is pushing that idea with a formula it has called “token factories.” In action. The most recent example comes from Wuxi, a city in eastern China’s Jiangsu province. According to Global TimesHonflex and the Wuxi High-tech Zone have promoted the first computing supernode there Huawei Ascend 384 of the province. The idea is to use that infrastructure as a starting point for a large-scale installation aimed at offering AI capacity measured in tokens to the market. Demand grows. If more and more applications use language models and AI agents, someone has to stably provide the capability to run them. Xinhua points out that at the end of March 2026, daily requests for tokens in China exceeded 140 billion, more than 1,000 times more than at the beginning of 2024 and 40% more than at the end of 2025. That is where the concept of “factory” makes a little more sense. Meaning of the label. In practice, AI data centers already function as token factories. They execute models, receive requests and return responses. What changes here is not so much the technical nature, but the way of converting it into an industrial product. It presents computing power as something measurable and sellable for those companies that need AI without building the entire infrastructure on their own. In detail. The Wuxi facility will start with four Huawei Ascend 384 servers. The promise here is to create a high-performance cluster based on domestic chips and models. In parallel, China Mobile announced on May 17 that it had built a computing center in Hubei for the center of the country with locally developed AI infrastructure and intelligent computing capacity exceeding 2,200 petaflops. A reading of technological sovereignty. In both projects, emphasis is placed on Chinese infrastructure, Chinese chips and national models. There are no mentions of American technology as the basis of the deployment, nor of NVIDIA chips, although the American company continues to be a global reference in AI hardware and has had a very relevant role in China. This framing fits with the Asian giant’s efforts to gain autonomy in a strategic technology. The initiatives also seem to point in that direction. The race continues. If we take the launch of ChatGPT in November 2022 as the starting point of this new AI race, little time has passed on the calendar, but a lot of time has passed in the industry. The US is not exactly the same actor as it was then, neither is China, and in between we have seen export restrictions, regulatory comings and goings, development of national alternatives and growing pressure to secure the technological base that allows us to continue competing. In this context, the concept of “token factories” appears. Now we have to wait to see if it will translate into a real advantage. Images | Xataka with Nano Banana In Xataka | If the question is whether AI data centers end up increasing temperatures in a region, the answer is: 2.2ºC

some Amazon employees use AI just to inflate their token metrics

by

He tokenmaxxing now has its most documented case. Some Amazon employees have been using MeshClaw for weeksan internal AI agent tool, to automate unnecessary tasks and thus inflate your consumption of tokens in the internal markers that the company has implemented. This is not the first time something like this has happened in Silicon Valley: Meta had its own leaderboard tokenswith a winner who took the title of Legend Token. And similar patterns have been documented at Microsoft. But the Amazon case adds a detail that makes it more striking: the tool used to cheat is the same one that Amazon has officially deployed to make its engineers work better. Why is it important. Amazon requires more than 80% of its developers to use AI tools each week and measures compliance using data consumption markers. LLMs. The company has said those statistics will not be used in performance reviews. Several employees have responded with variations of the same phrase: managers are looking at it. “When you track usage, you create perverse incentives and there are people who are very competitive with this,” one of them told the Financial Times. Yes, but. There is a more generous reading. Forcing a large organization to come into contact with new tools has a certain logic: if you force enough people to use them, someone eventually finds a really useful use for them. The problem is that that only works if there is real exploration. An employee who delegates to an agent the task of summarizing emails that no one will read is not learning anything, he is just inflating his metrics. The big question. amazon has committed 200 billion in AI infrastructure whose demand, in theory, is absorbed as it is deployed. If a part of that internal consumption is tokenmaxxing Purely, the figures that justify these requests are less reliable than they seem. The distinction between real adoption and inflated consumption matters because the former generates lasting demand while the latter disappears as soon as incentives change. Amazon has already restricted public access to device usage statistics. When the marker is no longer visible, the behavior it encouraged also changes. Go deeper. The Goodhart’s law He has been explaining this for fifty years: when a measure becomes an objective, it is no longer a good measure. Amazon hasn’t built a system to know if its engineers are using AI well. You have built a scoreboard, and the scoreboards are played. In Xataka | If the question is whether using ChatGPT or Claude in English is more efficient and saves tokens, the answer is: yes Featured image | Xataka

Anthropic has not raised the price of Claude. He has invented something better: token inflation

by

“Don’t worry, it costs the same.” That was Anthropic’s message to announce the launch of its new AI model, Claude Opus 4.7. In that statement they made it clear that “the price remains the same as Opus 4.6: $5 per million entry tokens and $25 per million exit tokens“There was, however, fine print, because the model is better but to achieve it it reasons more, and that means one thing: more tokens. And the more tokens you consume, the more the AI ​​bill goes up. Anthropic already warned. It should be noted that in that official announcement Anthropic did not hide the facts. In one of the paragraphs he clearly explained how Opus 4.7 “thinks more” and that has a direct impact on token consumption (we highlight the difference in bold): “Opus 4.7 is a direct update to Opus 4.6, but there are two changes worth keeping in mind as they affect the use of tokens. First, Opus 4.7 uses an updated tokenizer that improves the model’s processing of text. This means that the same input can generate more tokens (approximately 1.0 to 1.35 times moredepending on the type of content). Second, Opus 4.7 performs deeper analysis at higher effort levels, especially in the later phases of agent scenarios. “This improves its reliability on complex problems, but also means generating more output tokens.” Or what is the same: when it responds, Opus 4.7 uses significantly more tokens than its predecessor, and that is important because the output tokens are much more expensive than the input ones. In the specific case of Opus 4.7, five times more expensive ($5 versus $25). What is a tokenizer and why does it matter?. Large language models (LLMs) do not process text directly, but rather convert it into units called tokenswhich are fragments of words, symbols or characters. The tokenizer is the mechanism that makes that conversion. Anthropic has decided to update the tokenizer in Opus 4.7, arguing that its new system improves how text is processed. The direct consequence: the prompt that previously generated 1,000 tokens now generates up to 1,350. And since it is billed per token, the effective cost rises even though the price per token has remained the same. Confirmed by third parties. Simon Willison, a well-known analyst and popularizer in this field, created a tool to measure the difference in token consumption with the Claude Opus 4.6 and 4.7 API. He took the official Opus 4.7 ‘system prompt’ and ran it through both models: With Opus 4.6 it generated 5,039 output tokens With Opus 4.7 it generated 7,335 output tokens This represents a growth of 1.46x tokens between Opus 4.6 and Opus 4.7, even greater than that indicated by Anthropic (1.35x). For images the difference is even more extreme since the token consumption is up to 3.01x. There is an important clarification here, because there is support for images of up to 3.75 Mpixels and that higher resolution causes consumption to increase significantly. Bill Chambers, another X user, published another tool called Tokenomics that also allows you to compare token consumption between both models with any prompt. The aggregate ranking of all users who have tried this tool shows that the average increase is 38.6%, very much in line with what Anthropic points out. And also think more. As we said, this new model applies two changes in its way of acting. The first is the aforementioned tokenizer: the same input is converted into more input tokens. The second is the fact that the model now “thinks more” before responding, which means more token consumption. Opus 4.7 arrives with a new “effort” level called xhigh, located between high and max. Anthropic has decided that now the default effort will be precisely xhigh for all plans, so both mechanisms contribute to this higher token consumption. As Anthropic itself indicates, “Opus 4.7 thinks more about high effort levels, particularly in later turns in agentic settings. This improves its reliability on difficult problems, but it does mean that it produces more output tokens.” Criticisms on networks. The reaction of users has been clear and there are various examples on networks such as X or Reddit in which said users criticize the changes. On Reddit a thread titled ‘Opus 4.7 is a serious regression, not an improvement‘It already has 3,200 votes and 800 comments that sum up that this new model ignores instructions, hallucinates and lies, It’s “dumber”has become too complacent or even lazyand “talks too much”, which also contributes to the cost of each consultation. Many complain that their Pro and Max paid limits are running out faster than before due to these changes. Some users claim that Opus 4.7 is the first sign that Anthropic may has gone too fast for the first time when launching a new model. Anthropic reacts. Criticism about the cost and behavior of the model has made those responsible for Anthropic try to clarify things. Borys Cherny retweeted a message from the company in which was spoken how the “/usage” parameter in Claude Code allowed us to show what kind of things our API or usage plan is spent on. This same engineer, who is the person most responsible for the development of the aforementioned Claude Code, also indicated that since his new model now uses more tokens, in Anthropic they had increased the fees of use of the models, although without giving specific details. The pattern that repeats. For weeks now the user community he complained about what noticed a “regression” in the behavior of Opus 4.6. Although it is impossible to verify or validate it, there were many users who complained on networks about how the performance of the model had gotten worse in your tests. Now they have just launched a model that promises to be better than the previous one, but that ends up costing more to use if you are not careful. Both events draw a pattern: that Anthropic is increasing its revenue without announcing price increases as such. What users … Read more

Frankish calls are so problematic in Spain that even national banks have moved token. ING has been the first

by

Spain has a problem with fraudulent so -called. From those that They get through companies like indeed to offer us alleged jobs to incessant calls for commercial purposes. Such is the focus that is put in Spain with spam and scam, that the Government has had to move cardforcing these to have to be done with prefix numbers 900 and 800, prohibiting conventional 600 and 700. If this measure will be sufficient or not to determine the time. At the moment, the situation is at the point that even Spanish banking has begun to move card. Who calls me? Is the name of The tool that has launched ING Direct in Spain. To use it, you have to access its application (it is focused for customers of it and not as an open service), and enter the corresponding aid section – who calls me, the phone number we believe suspicious. Ing will match whether or not the number corresponds to one of its agents, showing you at the time if it should be reported by fraud attempt or if you can save that number as an official contact. It is important to note that this security measure only serves the numbers that ING has registered. Other SCAM attempts cannot be detected from here. It is not the first attempt. Monzoa British digital bank implemented a different measure, but also focused on fraudulent so -called. If during a call we acceded to the bank’s app, it showed us a message indicating that it was not they who were calling us. A quite effective way to avoid making any transaction if someone gets through an agent. For the bad luck of the Spaniards, There is no Spanish bank who is doing something similar. Why is government measures not enough. The measures materialized by the Ministry for Digital Transformation and Public Function are an important first step, but not a universal solution. No one prevents scammers and unregistered companies for calling for fraudulent purposes from conventional numbers, through Sim Swapping o Use of disposable Sim cards. According to the Ministry, since the Government took the initiative to end the so -called fraudulent, 14 million fraudulent calls have been blocked in Spain, since operators have the obligation to block numbers that do not correspond to any user or service. What is being done. Google has been betting on anti spam functions for years and anti -arud in your phone app. This, automatically block the calls thatprior analysis by AI, may have indications of fraud. He does the same with messages, warning us when the sender is suspicious. On the side of iOS, the wwdc 25 was the great moment for iOS 26 Begin to try to compete until now basic on Android. One of them is that of incoming calls from unknown numbers (anyone, there is no analysis of fraudulent behaviors). From this version, the user can choose that they all leakes so that the interlocutor is obliged to explain who he is and why he calls. It is killing gunflowsbut an aggressive and powerful filter. It is not the only open front. Fraudulent calls are not the only method of popular attack. SMS are another entrance door to scams And, although Google’s app blocks some, They are still a deception of the most common. There are no magical solutions. Spam and SCAM have it more difficult than ever, but the rear doors are inevitable. As we always recommend in Xataka, the sender, URL to which he directs the message, and never give personal data by phone, should be proven more than once. Image | ING In Xataka | Spain pays less than ever in cash and yet there are more and more tickets circulating. We have a suspect

The EU moves token and approves its retaliation tariffs to the US: we already know when the counterattack will begin to be applied

by

The commercial war between the United States and a good part of the world is no longer a latent threat: it is happening, and Europe is part of the board. In recent times, the Donald Trump administration has launched three direct offensives against the European Union. First imposed 25% tariffs on steel and aluminum; Then he did the same with the cars, And now he has added the so -called “20% reciprocal tariffs”. Brussels has decided to answer. First firm step. The answer has already begun to take shape. The Member States of the Community Block The first commercial countermeasted package against the United States has just approved. The proposal was treated this Wednesday and received a majority support. From the European Commission they have not left doubt: they consider that Washington’s tariffs are “unjustified and harmful” and that “cause economic damage to both parties, as well as the world economy.” {“Videid”: “X8WLH9Q”, “Autoplay”: False, “Title”: “United States vs. China: The chips war”, “Tag”: “Webedia-prod”, “Duration”: “1611”} The details, still to be known. The Commission has officially confirmed the approval of these measures, but has not detailed the concrete percentages or the products that will be affected. That information will be announced in the next few days through a formal ad. However, media such as The New York Times, which accessed the documentation related to the voting, They assure that these are 10% and 25% tariffs on a wide range of categories. Appliances, vessels and even dental thread. These percentages coincide with the information published this weekwhere it transpired that the European package would reach consumer and leisure goods, such as appliances, motorcycles, recreation vessels and cards, in addition to food products such as sausages, poultry and other agricultural products. Personal care articles would also be included, such as dental thread. Date indicated in the calendar. Although formal steps are still missing, such as the publication of the act of execution, the European Commission has already advanced the key date: “Rights will begin to be raised as of April 15,” According to the institution itself. That is, the countermeasures will be effective within a few days. The dialogue continues on the table. The movement does not imply a total closure to the dialogue. From Brussels they have stressed that Member States maintain the intention of negotiating with the US administration, although any agreement must be based on “balanced and mutually beneficial” conditions. In that line, the Commission has confirmed that the countermeasures “can be suspended at any time.” In Xataka China has no intention of backing: it will put tariffs of 84% to all US imports Who will really pay these tariffs? The answer is that you possibly impact consumers. “Tariffs function as taxes applied to imports”, Remember the Tax Foundation. “In practice, that additional cost is rarely assumed by companies: it ends up impacting directly on the consumer pocket.” If nothing changes, we will see more expensive American products in the European market. Images | European Parliament | The White House | Alexandre Lallemand In Xataka | The international wine market was already broken, but a single idea has put it against the ropes: 200% tariffs (Function () {Window._js_modules = Window._js_modules || {}; var headelement = document.getelegsbytagname (‘head’) (0); if (_js_modules.instagram) {var instagramscript = Document.Createlement (‘script’); }}) (); – The news The EU moves token and approves its retaliation tariffs to the US: we already know when the counterattack will begin to be applied It was originally posted in Xataka by Javier Marquez .

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.