Zuckerberg has given the keys of the future to a child prodigy of 97

There is the elite talent ecosystem in AI like a anthill in the rain by The 100 classes that Zuckerberg is disbursing (In some case even more) to be able to take a goal to not only scientists of a very high level, but to the type of people who are able to move the needle that matters most: that of the stock market capitalization of the entire company. In those, surely he is not doing all the noise that the person deserves in whose hands the real power of the goal has ended. It is the one who solved the most mundane and decisive problem: how to feed it. Alexandr Wang, 28 years oldHe did not create Chatgpt or Gemini, but he did built a fundamental infrastructure to make possible proposals. And now Zuckerberg has opted 14.3 billion dollars to this – insulting – young is the piece he needed to win the superintelligence career. Son of outstanding scientists in New Mexico, Wang is the founder of Scale AI, a company that does the dirty work, the one that nobody sees but everyone needs: label and prepare the data that train the most advanced AI models of the world. It is not glamorous, but it is indispensable. Like himself He explained“AI is reduced to three pillars: computation, data and algorithms. The computation is promoted by people like Nvidia, algorithmic advances come from great laboratories such as OpenAi, and the data is fed by Scale.” His great success was to understand something that in 2025 seems obvious but that a few years ago was: that the bottleneck of the AI ​​does not only go to algorithms or calculation power, but of quality data, something that We already count on Xataka In … 2019! And Wang, instead of pursuing a viral AI application, started selling shovels in full gold fever. And it turns out that everyone needed shovels. On Monday, Zuckerberg sent an internal memo that sounds more like a declaration of war than with simple reorganization. And it is fair to recognize that when Zuckerberg gets in visionary mode, he usually is right. Nine years ago he nailed him. Wang is a technical genius, but only explains a part of his meteoric ascent despite his youth. The least obvious is that Early understanding where the real value is. When researchers much more experienced than he focused on competing creating better algorithms, he realized that the future would be the one to control the data. “You can have the best algorithms in the world, but if you don’t have quality data, you have nothing,” said Wang. Scale AI became the data provider for OpenAi, Google, Microsoft and, of course, goal. Basically, It was placed in the modern AI supply chain center. Something that has earned more than 14,000 million. His business philosophy is also pure maturity and balance, not always present in entrepreneurs under 30 (Ahem). Keep personal control over each hiring. Also about each data delivery to its customers. “The quality is fractal, the high standards are filtered throughout the organization”, He said. It is also true that if Scale AI was in creditors, we would say that he could not delegate. As the scenario is very different, we praise its involvement. It is, in any case, obsessive to the extreme, but in the correct specific aspects and not with all the width of an organization. “If you really exceed something, if you go to 10 miles per hour of more and constantly go further, then you will improve faster than any other,” said. Pise the accelerator so as not to lift the foot from there. Meta needed Wang because his previous strategy has not finished working: Call 4 It has been a disappointment in a market in which a disappointment is lapidary. Eleven of the fourteen researchers of the flame team They have left target. Several of them have gone to Mistral. Zuckerberg has realized that he is losing a career he really wants to win. And his solution has been to buy the most valuable data ecosystem in the world and give the keys to its creator. Meta reorganization under the umbrella of ‘Meta Superintelligence Labs’ is much more than cosmetics. It is a tacit admission that AI cannot be another department within the finish line. It has to be a goal. And he has put a traditional one twenty -year -old Yann Lecun. It is a total commitment: that in the world of superintelligence, it matters more to understand how the AI ​​business works that the algorithms work. Wang has been selling shovels for years while others cavan. Now Zuckerberg has given him the treasure map. In Xataka | If in 2020 they would have asked us which technology would be more attractive in 2025 we would not have said ‘goal’. But here we are Outstanding image | Scale ai

The Soc Xring O1 is an “impossible” prodigy. This is how Xiaomi has dodged the demanding US restrictions

Xiaomi’s Soc Xring O1 is making a thunderous noise. His debut at the 15S Pro smartphone, the PAD 7 Ultra tablet and the Watch 4S clock is being preceded by performance figures that They place it ahead of the Snapdragon 8 elite of Qualcomm and Dimensity 9400 of Mediatek. He is not but that we are bad if we are in mind that he is the first flagship designed by Shanghai Xuanjie Technologies, the Xiaomi subsidiary specialized in the design of semiconductors. However, in this article we will not investigate the performance of this chip or compare it with the alternatives proposed by Apple, Qualcomm or MediaTek. What I propose is that we verify how Xiaomi has achieved that the US Industrial and Security Office allows you to manufacture it in N3E (3 Nm) node of TSMC. This Chinese company is not included in the “blacklist” of the US, but this is only part of a story in which Xiaomi has proven to have a huge expertise. TSMC lithography has a leading role in the performance of this chip During the presentation of the soc Xring O1, Xiaomi revealed some details Very interesting about its architecture, and there is no doubt that ‘Xuanjie’ engineers have done a very good job in this area. However, Lei Jun, the founder and general director of this company, He recognized yesterday During the event that its purpose from the beginning was to put to tune a soc produced in the most advanced lithographic node available. And it is that a well -executed architecture and an avant -garde integration technology are the two ingredients that must necessarily intervene in the recipe of a flagship. As I mentioned a few lines above, Xiaomi is not included in the “blacklist” of the US, but this condition is not enough to access the N3E node of TSMC, which is One of the most advanced in operation currently. This last company is from Taiwan, but both the machines of The Dutch company Asml which uses to manufacture its integrated circuits such as wafering processing techniques that they use incorporate American technologies. This is the circumstance that empowers the US when intervening in the ASML and TSMC business. The US government decides who has access to its technologies even if they reside outside its borders In this context, the US Industrial and Security Office of the US Commerce is responsible for controlling the export of US technologies that can be used for military purposes, and therefore, they have the ability to compromise national security and condition US foreign policy. All this means, in short, that the US government decides who has access to its technologies even if they reside outside their borders. Xiaomi’s Soc Xring O1 is on the role as advanced as the best SOCs for smartphones that have Apple, Qualcomm or MediaTek and can be used for a very wide range of applications that go beyond their integration into a mobile phone or a tablet. In The current tension situation They hold the US and China it is surprising that TSMC has been able to manufacture this chip for xiaomi In one of its most advanced lithographic nodes. The interesting thing is that he could because the engineers of ‘Xuanjie’ have calibrated the design of their soc in such an ingenious way that they have managed to throw a very competitive performance without breaching the regulation of the US. In this delicate balance lies the authentic merit of Xiaomi. In TSMC practice it has been able to manufacture it in its N3E node because the socket Xring O1 meets three fundamental conditions. The first of these is that it has been produced in Taiwan and not in any of the countries of Group D: 5, which are the nations to which the US considers a threat, such as Iran, North Korea, Syria, Myanmar, Venezuela, China or Russia. In addition, this SOC incorporates 19,000 million transistors, so it is below the threshold of 30,000 million transistors. If I had exceeded this TSMC limit, I could not have manufactured it for a Chinese company. The interesting thing in this area is that the engineers of ‘Xuanjie’ have managed to set up an architecture capable of delivering very competitive performance and efficiency using a relatively moderate number of transistors. The third and last condition that this Xiaomi chip meets is that it does not live with HBM memories (High Bandwidth Memory), which are the ones that work side by side with the GPUs for artificial intelligence (AI). We only have to have the opportunity to try it thoroughly in one of the Xiaomi devices that are going to bet on it to check if it is really up to the expectations it has generated. Image | Xiaomi More information | Nomad semi In Xataka | China is about to have the ability to make 5 Nm chips, although it faces a difficult solution problem

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.