{"id":789,"date":"2026-11-02T07:40:45","date_gmt":"2026-11-02T07:40:45","guid":{"rendered":"https:\/\/www.junkatanuma.com\/?p=789"},"modified":"2026-11-02T07:40:45","modified_gmt":"2026-11-02T07:40:45","slug":"the-old-generation-ai-frenzy","status":"publish","type":"post","link":"https:\/\/www.junkatanuma.com\/?p=789","title":{"rendered":"The &#8220;Old-Generation&#8221; AI Frenzy"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">The AI boom we are witnessing today may one day be remembered as the frenzy of an &#8220;old generation.&#8221;<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">At the end of October 2026, Kioxia was reported to be forecasting net profit 24 times higher than a year earlier. Semiconductors and memory are in the middle of an explosive surge in demand. Data centers keep multiplying, and the electricity and materials needed to run them are chronically short. More capacity, more speed, more power. There is no end in sight to this cycle.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">But having no visible end is not the same as being able to go on forever. At the center of this frenzy is the latest technology, and at the same time, an old idea. In this essay, I want to deliberately call what looks like the cutting edge the &#8220;old generation.&#8221;<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Numbers in a Frenzy<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">What is driving profits is not so much endlessly growing demand as supply that cannot keep up.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">In its July 2026 results, Kioxia announced that net profit for April\u2013June was 46 times that of the same period a year earlier. It projected 31 times for July\u2013September. Press reports attributed this to a sharp rise in selling prices amid a scramble for memory. What is growing is price more than volume.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The same Kioxia posted a record net loss of 243.7 billion yen in fiscal 2023, only three years ago. The company itself states in its risk factors that the flash memory market cyclically swings between improvement and deterioration.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The market knows this too. Even as record results kept coming, Kioxia&#8217;s share price had fallen by more than half from its June peak as of July. The numbers are in a frenzy, but surprisingly few people fully believe them.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Dependence on a Single Idea<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">This frenzy rests on a single idea: make it bigger, and it gets smarter.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Add more data, enlarge the model, pour in more computation, and performance rises as predicted. This rule of thumb, known as the &#8220;scaling laws,&#8221; has guided AI development for the past several years. In 2019, AI researcher Richard Sutton called it &#8220;the bitter lesson.&#8221; Attempts to build human knowledge into machines, he argued, have ultimately lost to the brute force of computation.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The bitter lesson gradually turned into a formula for success. The industry began to treat increasing computing resources as an end in itself. The shortage of memory and power is a natural side effect of this idea.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Efficiency is improving. Performance per watt of semiconductors gets steadily better every year. Yet total consumption does not fall. The 19th-century economist William Stanley Jevons observed that as steam engines became more efficient, coal consumption increased. Greater efficiency widens the range of uses, and the total swells instead. The same thing is happening with AI. As long as the premise stays the same, efficiency does not calm the frenzy; it fuels it.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Innovation in Means Only<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Elon Musk is bolder than anyone in his means, yet more orthodox than anyone in his premise.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">In February 2026, SpaceX merged with xAI in a deal valued at about $250 billion. The stated goal is to deploy large-scale AI data centers in space. Musk has also proposed a giant semiconductor plant called &#8220;Terafab.&#8221; If there is not enough power on Earth, go to space, where sunlight never runs out. If there are not enough chips, build the world&#8217;s largest factory yourself.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Both ideas remove constraints, but neither asks why so much is needed in the first place. The statement announcing the merger spoke of scaling up to create a &#8220;sentient sun.&#8221; Expansion of scale has itself become the mission. Musk embodies the idea of scaling in its purest form.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Even in space, a physical wall awaits. The CEO of Voyager Technologies, which is developing a commercial space station, has pointed out that cooling remains an unsolved problem.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Space is often thought of as a cold place. In reality, it is a place where things struggle to cool down. Think of a thermos. It keeps tea hot for hours because the space between its walls is a vacuum. Without air or water to carry heat away, heat has nowhere to go. A data center in orbit becomes a machine generating heat inside a giant thermos.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Data centers on Earth carry heat away by blowing air or circulating water. Those methods are not available in space. The only option left is radiation, releasing heat as infrared light. But the heat that radiation can remove increases only in proportion to surface area. In space, the only answer is to deploy vast radiator panels.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The International Space Station is a good example. Separately from its solar arrays, the ISS deploys large radiator panels. Even so, they can reject only about 70 kilowatts of heat. A large AI data center, by contrast, uses more than 100 megawatts, almost all of which becomes heat. That is more than a thousand times the ISS. By rough estimate, getting rid of that heat would require radiators covering the area of more than a dozen soccer fields. These would have to be launched by rocket and assembled in orbit.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Sunlight may be limitless in space. But using electricity always produces heat. Going to space removes the power constraint, only to bring a different one, cooling, to the fore. The constraint has not disappeared. It has merely changed shape.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Of course, there is another view. Some argue that freeing data centers from the terrestrial power grid is itself a transformation of infrastructure. Musk also has a record of overturning conventional wisdom, as with reusable rockets. Still, this is innovation that raises the ceiling on the supply side. It is not innovation that rethinks what intelligence is.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Evidence of Another Path<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The evidence that intelligence need not depend on scale lies inside our own heads.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The human brain runs on about 20 watts. With the power of a single light bulb, it speaks, reads the intentions of others, moves the body, and keeps on learning.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The brain and today&#8217;s computers rest on fundamentally different structural premises. In a computer, memory and processing are separate, and simply shuttling data back and forth between them consumes enormous power. In the brain, memory and computation are not separated. Moreover, most neurons are at rest, and only the parts that are needed are active. Rather than adding more memory, the brain reorganizes the very relationship between memory and computation.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Research is already moving in this direction too: computing inside memory, neuromorphic chips modeled on the brain, and &#8220;sparse&#8221; models that activate only the parts they need. But each of these remains small as an individual technology.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Yet looking only at the brain shows just half the story. The brain does not compute alone inside the skull. It computes together with the body and the world around it.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Take walking. A bipedal robot manages not to fall only by continuously calculating the angle of every joint. In 1990, however, a model consisting of nothing but legs, with no motors and no controller, walked down a gentle slope with a strikingly human gait. It is known as a &#8220;passive dynamic walker.&#8221; Most of the walking was carried out by the shape of the body itself: the length and weight of the legs, their pendulum-like swing. The brain did not need to calculate everything.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The octopus is even more extreme. Of its roughly 500 million neurons, about two-thirds are not in the brain but in the arms. Each arm moves with a degree of independent judgment. Intelligence is not gathered in one place but distributed throughout the body.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Roboticist Rodney Brooks said that &#8220;the world is its own best model.&#8221; What is in front of us need not be memorized; we can simply look again. Humans do not copy the world into their heads piece by piece. When needed, we turn our eyes, reach out a hand, and check. We use the world itself as an external memory.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Today&#8217;s AI goes exactly the other way. It has no body, never touches the world, and learns about the world only from text and images. Because it cannot go and check the world, it tries to copy all of the world inside itself. The vast data, memory, and power can be seen as the price paid by an intelligence without a body. If you try to receive the world with the head alone, the head has no choice but to grow without limit.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Japanese has the phrase <em>shinshin ichinyo<\/em>, the oneness of mind and body. The <em>kata<\/em>, or forms, of martial arts and traditional arts are likewise a system for entrusting knowledge to the body rather than the head. A master does not think through each movement. Without thinking, the body knows. The key to making intelligence lighter may not be reducing computation. It may lie in sharing computation with the body and the world.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">What is needed is a co-evolution of ideas: hardware, algorithms, the body, and the question of what we want AI to do. These must be reshaped not separately but in mutual influence. Instead of everyone sharing a giant general-purpose model in a distant data center, small intelligences embedded in machines and tools on site would work in a distributed way, each in touch with its own environment. Then the very order of magnitude of the scale required would change.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">The Path Where Humans Themselves Change<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The &#8220;other path&#8221; described so far is still about the machine side. But co-evolution has another party: human beings.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Discussions of technology always turn to what to build: new services, new chips, new models. Only things that can be built outside of people become products, earn patents, and attract investment. When human abilities themselves grow, that does not show up in any share price. That is why so few people consider this path.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Some did consider it seriously. Douglas Engelbart, known as the inventor of the mouse, defined the goal of his research in a 1962 paper as &#8220;augmenting human intellect.&#8221; He envisioned evolving not only tools but also the skills, methods, and even language of the people who use them. Tools and people would lift each other up.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">But industry chose the opposite path. Being usable by anyone, without learning anything, became the mark of a good tool. The smarter tools become, the less people need to train themselves. Today&#8217;s generative AI is the culmination of that trend. The machine produces the answer, and people merely receive it.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Examples of human abilities co-evolving with tools are in fact close at hand. A person who has mastered the abacus can calculate mentally, moving imaginary beads in their head without the device. The tool has entered the body and turned into an ability. Literacy, playing a musical instrument, and simultaneous interpretation are the same. Each is the result of training that has reorganized how the brain and body are used, and each runs on 20 watts.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The <em>kata<\/em> of martial and traditional arts can be seen as a technology that has passed down this co-evolution for centuries. <em>Kata<\/em> does not store knowledge outside. It transfers knowledge into a person&#8217;s body and changes that person. This is the exact opposite of a data center trying to copy the world inside a machine.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">AI, too, could be designed differently, not as an intelligence complete outside of people. It could draw out human abilities and eventually be set aside, like an abacus. A tool that makes people grow the more they use it, rather than one that dulls them. Co-evolution of ideas, ultimately, means machines and people changing together.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">History Repeats<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Looking back at the history of technology, there are many cases where the cutting edge became obsolete along with the very idea behind it.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">ENIAC, completed in 1946, used about 18,000 vacuum tubes and consumed about 150 kilowatts. At the time, it was the cutting edge in the literal sense. But with the arrival of the transistor, the very idea of adding more vacuum tubes came to an end. The answer did not lie on the extension of the vacuum tube.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Another precedent is the dot-com bubble. Communications demand was believed to grow without limit, and optical fiber was laid in excess. When the bubble burst, most of it lay unused as &#8220;dark fiber.&#8221; Yet it later became the cheap foundation for video streaming and the cloud, uses that those who drove the frenzy had never imagined.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Today&#8217;s AI frenzy will probably not end calmly either. When it becomes clear that massive investment cannot be recovered through revenue from AI services, capital spending will be cut all at once, and memory prices will plunge. The semiconductor industry has long called this the &#8220;silicon cycle.&#8221; Its swings may simply be larger than ever before.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">But this is not mere collapse. Overbuilt data centers and factories will remain after the frenzy as an inexpensive foundation. On top of it, an entirely different idea may grow. New ideas usually emerge only when the old way hits a wall. Today&#8217;s shortages of power and materials are likely signs that the wall is coming into view.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The frenzy is only a prelude to the next movement. Seen this way, calling today&#8217;s cutting edge the &#8220;old generation&#8221; is not sarcasm toward it, but hope for the next generation.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">References<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/www.sakigake.jp\/news\/article\/20260731CO0145\/\">Kioxia forecasts net profit of 1.27 trillion yen; AI demand continues (Akita Sakigake Shimpo, July 31, 2026, Japanese)<\/a><\/li>\n\n\n\n<li><a href=\"https:\/\/eetimes.itmedia.co.jp\/ee\/articles\/2405\/15\/news189.html\">Kioxia posts record 243.7 billion yen loss for FY2023 (EE Times Japan, Japanese)<\/a><\/li>\n\n\n\n<li><a href=\"https:\/\/edinetdb.jp\/company\/E35948\/earnings\/FY2024\">Kioxia annual securities report: business risks (EDINET DB, Japanese)<\/a><\/li>\n\n\n\n<li><a href=\"https:\/\/biz-journal.jp\/company\/post_393440.html\">SpaceX merges with xAI for about 39 trillion yen (Business Journal, Japanese)<\/a><\/li>\n\n\n\n<li><a href=\"https:\/\/www.businessinsider.jp\/article\/spacex-tesla-xai-merger-explained-elon-musk-business-empire-ipo-2026-1\/\">On the SpaceX\u2013xAI merger statement (Business Insider Japan, Japanese)<\/a><\/li>\n\n\n\n<li><a href=\"https:\/\/innovatopia.jp\/spacetechnology\/spacetechnology-news\/79791\/\">Voyager CEO disputes Musk&#8217;s space data center plan (innovatopia, Japanese)<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>The AI boom we are witnessing today may one day be remembered as the frenzy of an &#8220;old generation.&#038;#8221 [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":790,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-789","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=\/wp\/v2\/posts\/789","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=789"}],"version-history":[{"count":1,"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=\/wp\/v2\/posts\/789\/revisions"}],"predecessor-version":[{"id":791,"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=\/wp\/v2\/posts\/789\/revisions\/791"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=\/wp\/v2\/media\/790"}],"wp:attachment":[{"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=789"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=789"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=789"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}