Micron

Google has made AI consume up to six times less memory. Micron, Samsung and SK Hynix are paying dearly

by

we carry months wrapped in the memory crisisbut maybe there is a way out. Last week Google Research published a study in which he revealed a technique called TurboQuant. This is a compression algorithm capable of compressing the working memory of AI models up to six times without appreciable loss of quality or performance. Great news for end users, who see a light at the end of the tunnel, but terrible news for manufacturers, who this golden age could end. Let’s explain what KV cache is.. To understand TurboQuant you have to understand what that memory is that it manages to compress. When a language model processes a long conversationyou need to remember the context. Each token that is processed is stored in the so-called KV cache, a type of working memory that grows as we chat. The longer the conversation, the more memory the model requires. Compressing what is a gerund. It is one of the main bottlenecks in the AI ​​inference stage (that is, when we use the models), and one of the reasons why data centers they need as much RAM or HBM memory. TurboQuant uses a vector quantization method to compress this cache while maintaining the precision of the model. Pied Piper. As soon as this Google study appeared, the analogies began with the plot of the series ‘Silicon Valley’. In it, the fictional startup in the plot managed to develop an extraordinarily efficient compression algorithm called Pied Piper that threatened to revolutionize the technology industry. These days, multiple references to the series appeared on social media, which had already been referred to as visionary for reflecting what is happening with spectacular accuracy even when the series was a comedy. Six times less memory. The Google Research paper states that this method is capable of reducing the KV cache six times without an appreciable difference in performance in long conversations. The researchers will present their results at an event next month and explain the two methods that allow it to be put into practice. If they confirm what they’ve already teased, the implications are huge: less memory for inference means data centers can do the same thing with much less hardware/memory. Google’s DeepSeek moment. The discovery has some analysts calling this Google’s “DeepSeek moment.” A year ago, the Chinese startup DeepSeek launched an AI model that competed with the best but had cost much less to develop. That shook the industry, and now we return to a technical achievement that points to the same thing. In AI, doing the same with less is crucial, given the enormous resources that this technology requires. There are those who already have done evidence preliminaries with TurboQuant and have confirmed that the method does indeed work. Micron, Samsung and SK Hynix pay dearly. The impact of this technique can be enormous, and this has already begun to be noticed in the stock market valuations of DRAM memory manufacturers and HBM. Companies like Micron, Samsung, SK Hynix, SanDisk and Kioxia fell noticeably last week from their recent highs. On March 18 it was around $471, and today its shares are at $357, a staggering 24.2% drop. The same has happened with the rest of the manufacturers, which were already falling since that date, but have accelerated that fall with the launch of TurboQuant. But. The technique can theoretically be applied only to the inference phase, but the training phase of AI models is not affected by this compression technique. Therefore, huge amounts of memory will still be needed during the training phase. Besides we will have to wait for AI companies to actually start applying said system if it is confirmed to work, and that will be when we can see the real impact. Theoretically this will give a lot of room for maneuver to big tech, which will be able to reduce token prices even further, but it remains to be seen if they do so. RAM memories drop in price. The impact of TurboQuant has also been clear in the prices of memory modules, which have dropped appreciably in price. For example, the Corsair Vengeance DDR5 32 GB 6000MHz (2x16GB) modules were at 489.59 euros on Amazon until a few weeks ago according to CamelCamelCamel, but right now they are at 339.89 euros, a notable discount. It is true that not all components are falling equally, but there are indeed cases in which reductions seem to be occurring. In Xataka | The RAM crisis is destroying all of Valve’s plans with its Steam Machine

Micron knew that the RAM crisis was going to be great for them. The reality that has gone even better

by

As it could not be otherwise, the companies that are benefiting the most from the RAM crisis They are precisely those that have the product and, therefore, they are the ones that set the price. Micron is one of those few companies that is profiting from the excessive demand of this key component for any gadget, a demand caused by the AI ​​fever. The figures from its latest financial report have even exceeded expectations. Although there are some nuances to comment on. Let’s go to trouble. What has happened? Micron just published the results of its second fiscal quarter with numbers that have left analysts speechless. Its revenues have almost tripled those of the previous year, reaching $23.9 billion, well above Wall Street estimateswho expected about 20,000 million. Earnings per share have skyrocketed to $12.20, compared to the $9 projected. And for the third quarter, the company anticipates revenue of approximately $33.5 billion, almost ten points above what the market expected. Those who share the benefit. Artificial intelligence has changed everything in the memory market. The data centers that power AI models require massive amounts of high-performance memory, and the available supply cannot meet that demand. Micron, together with Samsung and SK Hynix, forms the trio that controls practically the entire supply world of high-bandwidth memory (HBM) chips, which are basically one of the key components to run the long-awaited NVIDIA GPUs. Those who buy at any price. Micron’s own CEO, Sanjay Mehrotra, counted to CNBC that the company can only cover between 50% and two-thirds of what its main clients need. Put another way: there is a queue of buyers willing to pay whatever it takes, and Micron simply doesn’t have RAM for everyone. According to SK Group President Chey Tae-won, the global shortage could last another four to five years due to structural bottlenecks in semiconductor production. What’s coming Aware that what is happening now will not last forever, Micron is investing at a speed that has made the market nervous. The company plans to exceed $25 billion in capital spending in 2026 alone, and has already announced that in 2027 that number will rise another $10 billion. Among other operations, it has closed purchasing a plant of Taiwanese Powerchip for $1.8 billion, which will begin producing DRAM wafers in the second half of 2027. The company has also started mass shipments of its new HBM4 memory of 12 layers, which will be directed to the new Vera Rubin platform from NVIDIA. Precisely how much NVIDIA will depend on Micron for this new generation compared to its rivals is the big open question for all investors. Everything is going well for them, but the shares are going down. There has been a bit of a cold reaction in the stock market, as shares have fallen around 5% in the session after the results, despite the fact that the numbers have beaten all forecasts. The reason is the same thing that happened with NVIDIA a few weeks ago: When expectations are very high, even good results can disappoint. From Goldman Sachs they counted that the value could move in a narrow range in the short term after a “very solid quarter with guidance well above consensus, in a context of already elevated expectations.” That has not prevented banks like Wells Fargo or Barclays from updating their upward forecasts to $550 and $670 per share, respectively. The big photo. Micron has accumulated a revaluation of more than 60% so far this year, and has become the most profitable value on the PHLX (Philadelphia Semiconductor Index). Mehrotra affirms that Micron is “the invisible layer that powers AI today.” But it seems that the company is slowly losing that cloak of invisibility. In Xataka | NVIDIA has been pining for months to sell its H200 to China: it just received the news it was waiting for

Micron has emulated TSMC and is spending $1.8 billion on a RAM factory. Don’t clap yet

by

Taiwan is becoming one of the technological hotspots worldwide. If the country was already at the center of the technology sector because it is the home of TSMCwill now take on more prominence in the new era of AI. Your crown jewel is investing an astronomical sum in the United States and, now, the American Micron ends to close a $1.8 billion deal in Taiwan. And you can guess the goal. Keep feeding the data centers based on RAM memory. Micron. In recent weeks, Micron has been one of the big names in the technology sector. However, Crucial may sound more familiar to you. It is, or was, Micron’s brand for consumer RAM, but also for storage. Their products are very well regarded when it comes to assembling a PC in parts, but They turned off the tap at the end of last year and the last shipments will occur in February 2026. Now, Micron is shifting its focus to something much bigger and more lucrative: artificial intelligence. Specifically, supplying those same components, but to large companies that are setting up gigantic data centers. In the end, a data center It is made up of hundreds of “computers” that need both storage and RAM. The operation. Given the context, we come to the news. As the company itself has confirmedhave just signed an operation worth $1.8 billion to take over the P5 factory of the Powerchip Semiconductor Manufacturing Corporation -PSMC- company in Tongluo, Taiwan. An operation like this must pass several filters, but the company’s intention is for the transaction to be closed by the second quarter of this same 2026. They have stepped on the accelerator, and as soon as they can, they will begin to do one thing: increase the production of DRAM memory. clean room. Micron has confirmed that it is just one of the operations it is contemplating in a global expansion movement “to meet the long-term demand of its customers,” and acquiring a semiconductor factory makes perfect sense. Beyond the fact that the components and machines are different, there is something that factories of this type share: clean rooms. It is an extremely… well, clean facility stripped of any external elements. Suspended particles are kept at extraordinarily low levels, temperature, humidity and pressure are highly controlled parameters and the air is filtered numerous times per hour. Static electricity is reduced as much as possible and, ultimately, it is a clinical space so that no impurities interfere with something as sensitive as the manufacturing of semiconductors. It is, in short, like an operating room (or stricter if possible). Example of a clean room “All in one hour“Creating something like this requires a considerable investment (which is why new companies are entering to compete in the RAM segment, as rumored with Asusit is tremendously complicated), and that is why Micron has taken over existing facilities that they will only have to adapt to their activity. Besides, take the example of TSMC. In Taiwan, all components TSMC needs are “an hour” away. This allows the assembly line to be efficient, minimizing time, maximizing production and saving money. The new Micron factory will be very close to the one they already have in Taichung, being able to emulate that way of working that has led TSMC to excellence. Consumption RAM for when. Micron is expected to begin optimizing the manufacturing process in the new plant by the second half of 2027, but thanks to the context we gave before, we know that these “customers” are not those who want to assemble a PC in parts or even assemblers such as Asus, MSI, Lenovo or Gigabyte: they are the ‘Big Tech’ that are setting up data centers. In a recent interview, Christopher Moore, vice president of marketing for Micron’s client and mobile business, said the problem and the RAM bottleneck is elsewherebut also stated that this growth in data centers has gone from representing 30% of its market to 60%. He also stated that, although Crucial has disappeared, Micron will continue to supply memory to OEM manufacturers, but it is evident that the bottleneck is affecting, that prices are through the roof and that things are not looking good if you had to renew PC.E And, according to Micron’s vice president, it will continue until 2028. At least. Images | Maxence Pira, Hunter Trick In Xataka | Google doesn’t have rockets, but it is going to install data centers in space. SpaceX and Blue Origin rub their hands

Everyone blames the manufacturers for the lack of memory. Micron says real bottleneck lies elsewhere

by

For months, memory shortage It has established itself in the technological debate as one of those phenomena that do not seem to need too many explanations. If RAM is missing and prices risethe immediate conclusion is that someone is privileging AI and leaving the consumer aside. That idea has resonated strongly, especially after visible decisions that have affected the domestic channel and have reinforced the feeling of abandonment. But when you get down to how memory is manufactured and kept stable today, the diagnosis becomes less obvious: the bottleneck doesn’t seem as obvious as it seems. A controversial decision. In this climate of widespread suspicion, Micron has become a preferred target, shared with other large manufacturers, but for a very specific and recent decision: the announcement of the end of Crucial consumer products. The company recently announced that will stop selling RAM memory and storage under that historic brand, with shipments expected through February 2026. For many users, that move was interpreted as a direct consumer recall just when memory is short. Micron justified that decision by noting that AI-driven growth in data centers has skyrocketed demand and that Crucial’s exit seeks to improve supply and support to its strategic customers in higher-growth segments. The market has changed size. From Micron’s perspective, the problem is not a renunciation of consumption, but an abrupt change in the scale of the market. Christopher Moore, vice president of marketing for the client and mobile business, He said in an interview with Wccftech that the company continues to have a relevant presence in PCs and mobile devices, while serving data centers. What has altered the balance is the growth of the data center business, driven by AI, which has gone from representing around 30% of the market to approaching, according to its figures, 50% or even 60%. That leap, he defends, has left the entire industry without sufficient margin. Variety also creates scarcity. For Micron, the bottleneck is not so much the lack of factories as how the existing ones are used. Moore explains that producing memory is not about making a single type of chip seamlessly, but rather about switching between multiple densities and configurations depending on what customers ask for. Each change, for example going from 12 GB to 16 GB modules or from 16 GB to 24 GB, forces lines to be readjusted and reduces the total output volume. In a context of skyrocketing demand, this variety, which was previously acceptable, becomes a direct brake on production. Micron’s new Idaho factory under construction Faced with the temptation to think that new factories will solve the problem, the manufacturer asks for patience. Moore explains that expanding memory capacity is not an immediate process, because it requires not only building facilities, but equipping them, validating them and certifying each product with customers. The company laid the first stone three years ago in its ID1 plant in IdahoUnited States, whose entry into operation is scheduled for mid-2027. Even so, it warns that there will be no significant impact on supply until the entire qualification process is complete, which it places in 2028. Crucial is gone, the channel is not. Moore assures that, although Crucial has disappeared from the consumer showcase, the company continues to provide memory to major PC and mobile device brands through channels less visible to the end user. This OEM channel, in which Micron supplies memory directly to integrators and manufacturers, concentrates a very relevant part of the market and ends up being incorporated into commercial designs and equipment. From their point of view, the consumer continues to receive Micron memory, even if it no longer does so under a recognizable label. With this panorama, the lack of memory ceases to be a problem of isolated decisions and is revealed as the result of several overlapping tensions. AI-driven demand for data centers that has changed the scale of the market, operational limits on production and long lead times to expand capacity explain why supply will remain tight for years. Micron places the relief horizon no earlier than 2028 and, until then, the consumer will live with fewer options and pressured prices. The bottleneck, the company insists, is not only in who buys the memory, but in how it is manufactured. Images | Micron In Xataka | The situation with RAM prices is so desperate that there are already those who build their own memory at home

The RAM memory crisis seemed to have its months numbered. Micron has a completely different perspective

by

They say that there is never a bad time to do something you really want and that, many times, the only thing that stops us is finding the right excuse not to start. That idea can work in many areas, but today it doesn’t fit very well if what you have in mind is build your own PC. At least not without assuming that the current context clearly works against it. We are witnessing firsthand how the so-called “memory crisis” is pushing upand notably, the prices of NAND memories, key in SSD units, and of the DRAM used in computers and laptops. We have more and more reasons to be patient, because an actor as relevant as Micron already warns that the challenges for the sector will persist for quite some time. The memory crisis is still far from resolved The company has put a date on the table and it is not what many expected. In its communication of resultsSanjay Mehrotra, CEO of Micron, spoke of “tight conditions” in DRAM and NAND and said that they are expected to “persist through and beyond 2026“In other words, if today we are already seeing how memory is becoming more expensive, Micron is warning that it does not point to a clear return to normality in 2026. That nuance matters, because it changes the horizon of anyone who is thinking about building or upgrading a PC in the short term. There are numbers that help to understand why this phenomenon does not remain a simple market swing. The firm once again reported record revenues of $13.64 billion in the last quarter, compared to $8.71 billion in the same period of the previous year, driven by the push for AI. That does not mean that there is excess product in all segments, because capacity and industrial priorities do not adjust to the pace of demand. Micron says it expects to increase its DRAM and NAND shipments by 20% next year, but acknowledges that boost isn’t enough to catch up. To understand why the domestic market is more exposed, it is worth looking at the photo of the factory. Micron is pushing its business toward HBMa memory designed for high-performance systems in data centers, and that has an opportunity cost. It is a technology that uses three times more silicon wafers than conventional DRAM, which means that, with the same capacity base, fewer units can be produced for the rest of the segments. It is not that consumption memory disappears, it is that it has less priority. Micron is pushing its business toward HBM. The first consequence is already being noticed by those of us who look at prices to build or expand a PC. Memory is what is becoming more expensive and the effect is seen, for example, in DDR5 kits. From there, the tension begins to filter through the rest of the chain, not only due to price, but also due to availability. The decision with Crucial also fits into this shift in priorities. Micron will stop selling consumer products under that brandwhich means one less player in this market, and greater pressure for those who are still in the race in the domestic sector. If Micron makes one thing clear with its roadmap, it is that standardization is not around the corner. The company is accelerating investment and capacity, but with a calendar that moves in yearsnot in weeks, and that forces us to look at 2026 with another face. For those who are thinking about buying or building a PC, the prudent reading is simple: it is advisable to assume that memory will continue to be a sensitive component, both in terms of price and availability, for a good period of time. Images | Micron | Samsung In Xataka | RAM is so expensive that smartphone manufacturers already have a plan: return to phones with 4 GB of RAM

We believed that Nvidia was the company that had benefited most from AI. Micron is ridiculous

by

Micron is at historical maximums in Nasdaq, and rightly. The American manufacturer is taking a lot of benefit from the FEVER through the AI ​​and the data centers. The demand for memory chips is growing extraordinarily, but that has two faces. A good for micron, and another bad for customers and consumers. They all love micron. Citigroup analysts They promoted these days Micron’s target price from $ 150 to 175. The reason: according to its data, the company will have financial results “much better than consensus” when these are presented on September 23. Micron is doing so well that it even exceeds the growth of Nvidia. Source: Bloomberg The chips devastate. Yeah A week ago The shares were around $ 125, yesterday they closed at $ 150 and before market openings that figure is $ 155. This year the value has already grown by 81%, exceeding 33%Nvidia growth, although it is also true that the company led by Jensen Huang grew especially in 2024 (approximately 170%). Other companies such as Broadcom (55%), SK Hynix (91.88%in the South Korean bag), or TSMC (31%) also show an outstanding growth in the bags. Micron’s “Compute Networking” division is the one corresponding to the data centers. As can be seen, sales in that segment are already more than half of all of the last quarter. Source: Paul/Note. The commitment to HBM memories goes well. Micron has dedicated many resources to boost the manufacture of HBM memories, used precisely in the accelerators (GPUS) that are used in data centers. Independent analysis confirm the increasing weight of both these memoirs and the AI ​​segment in the micron business Micron will raise prices. According to Citi analysts, workloads for the inference of AI need more DRAM and NAND memories, and demand is spectacularly. The problem is that this demand will overcome the supply, and Micron will take advantage of the occasion to do something logical (for her): upload prices. Up to 30%. This is what it indicates Trendforce And also Some media In China, according to which Micron has notified its distribution channel partners today that the prices of their storage products will rise between 20% and 30%. In fact, the quotes of the DDR4, DDR5, LPDDR4 and LPDDR5 memories have been suspended among others: “All prices agreed with customers will be canceled and the quotes will be suspended. All products are expected to stop quoting for a week. “That involves not only industrial and consumer memories, and the chips for the automotive industry will rise in price by 70%. Sandisk and TSMC have already announced up. Both TSMC and Sandisk announced Price increases For memory chips in the past days. That will affect its great clients –apple, Nvidia, among others – and as indicated In Techpowerup It is a clear confirmation that manufacturers want to maintain their gross margins. In Sandisk there have been 10% prices due to the “growing demand” of the AI ​​market, data centers and mobile devices. At the moment, they indicate In Trendforcethat climb has encountered resistance from customers. In Xataka | Intel’s recent history is that of a failure. Now he has found a niche from which to resurface: HBM memories

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.