[Technology & Economy] Why next to Tokyo Tower? Investigating the "hard-to-articulate unease" surrounding the expansion of AI data centres into the city centre.
//Summary - Level-C1//
The text argues that next‑generation AI requires ultra‑low latency, forcing data centres into city centres despite enormous costs. As AI begins interacting with the physical world—autonomous driving, AR, voice agents, and rapid machine‑to‑machine communication—distance becomes a critical bottleneck. This shift mirrors the extreme proximity race of high‑frequency trading. Yet urban residents bear noise, heat, infrastructure strain and rising energy costs, while foreign tech firms extract profits. The author proposes targeted taxation and mandatory community benefit agreements to ensure that AI’s expanding physical footprint enriches local communities rather than exploiting them.
//Summary//
An AI-dedicated data centre is rising right next to Tokyo Tower—but why in a location with the highest land prices in Japan?
The answer is that "next-generation AI requires physical proximity." But is this race truly new?
In fact, the world of high-frequency trading in finance has seen a level of madness for over a decade—the kind where mountains were blasted away to shave off a single millisecond. There is an industry that spent 45 billion yen to lay straight fibre-optic cables, only to find that insufficient and switch from light to microwaves. A "milder" version of that madness is now arriving in the heart of Tokyo.
This video explains the three key factors behind the rush to build data centres in the city centre, using real-world examples from high-frequency trading:
1. The decisive difference between text-based AI and next-generation AI
2. The value of "physical proximity" pushed to the absolute limit by the financial industry
3. The sense of unease regarding the structure that allows foreign firms to capitalise on Japan's low-cost infrastructure
1)
I am an engineer working in the field of facilities, such as plants, factories and data centres. As a cat who also enjoys learning about economics, history, share investment and technology as a hobby, I thought I’d share with you the practical experiences and knowledge I’ve gained through these pursuits.
2)
The term ‘cloud’ might well be the greatest fantasy ever conjured up by the IT industry. We are led to believe that the data we exchange daily via our smartphones, or the magical-sounding text generated by ChatGPT, is being processed in some fluffy, safe place high up in the sky. However, the reality of the cloud is nothing like a cloud at all. It is a ‘giant concrete fortress’ that devours vast amounts of electricity, emits intense heat, and keeps its cooling fans whirring with a deafening roar.
3)
And now, such a fortress is about to be built right in the heart of the city where we live. Currently, the Australian company ‘NEXTDC’ is constructing a massive AI-dedicated data centre on a prime site in Shiba Park, Minato Ward, right next to Tokyo Tower, a symbol of the city. Its output is 28 megawatts. This is equivalent to the electricity consumed by approximately 20 medium-sized office buildings.
4)
Anyone with even a modicum of knowledge of business or property should feel a strong sense of unease at this. It is common sense that data centres should be built in the suburbs or cooler regions, where land is cheap, vast plots can be secured, and the electricity grid has spare capacity. In Japan, the fact that places such as Inzai City in Chiba Prefecture are referred to as the ‘Ginza of data centres’ is proof of this.
5)
So why, then, would anyone go out of their way to place a data centre at the foot of Tokyo Tower, in an area with Japan’s highest land prices and highest population density? Surely it cannot be profitable. As someone who extensively researches data centre-related matters, I, too, found this deeply unsettling.
6)
However, behind this location strategy—which at first glance might seem ‘madness’—lie the ‘rules of next-generation AI’ that will dominate the coming era, and the ‘cruel truth of physical distance’ to which global capitalism has been driven.
In this article, whilst unravelling the mystery of the baffling construction next to Tokyo Tower, we delve into the frightening reality that the evolution of AI is forcing an update to our urban landscapes and living infrastructure.
7)
Chapter 1: With Text AI, ‘Even if the servers are in the US’, nobody is inconvenienced.
As a starting point for solving this mystery, I would first like to clarify ‘the AI we are currently using’. Generative AI systems such as ChatGPT and Claude, which the world is currently going wild for. When we type a question into our smartphones here in Japan, that data travels via undersea cables across the Pacific Ocean to massive data centres situated across the vast expanse of the United States. There, ultra-high-performance GPUs calculate the response, which then crosses the Pacific once more before the text appears on our screens in Japan.
8)
Just for this round-trip communication alone, no matter how fast the speed of light is, a physical ‘latency’ of approximately 100 milliseconds (0.1 seconds) one way—and 200 to 300 milliseconds round-trip—is inevitable. When you factor in processing time, it’s not uncommon to wait 1 to 2 seconds.
But think about it for a moment. Have you ever been using ChatGPT and got so angry that you thought, ‘It’s too slow to be of any use!’? Most people probably feel that ‘whilst I’m thinking of what to write, I have to wait a little, but it’s within acceptable limits’.
9)
The reason is extremely simple: ‘It is because the speed at which AI generates and returns text is overwhelmingly faster than the speed at which humans recognise and read text on a screen.’ The moment the AI begins to clatter out text, we try to follow it with our eyes and comprehend it in our minds. Thanks to this ‘bottleneck’ of human cognitive speed, the one-second delay occurring in the background is neatly masked.
10)
In other words, it is only in the current use of AI—where ‘text is entered on a screen and read by a human’—that it makes no difference whatsoever if the server is located on the other side of the globe. The model of placing servers in the vast, low-cost countryside of the United States to provide services worldwide was, in fact, highly rational. However, this peaceful era is soon to end.
11)
Chapter 2: ‘A One-Second Delay Can Be Fatal’—The Arrival of Next-Generation AI
Why must we go to the lengths of paying exorbitant land costs to locate a data centre right in the heart of Tokyo? The primary reason is that ‘the next generation of AI will no longer be a tool for typing and reading text’.
Next-generation AI will operate in real time alongside the real world, directly engaging with our five senses. The moment we enter this realm, the ‘200-millisecond delay’ incurred when communicating with the US will become a fatal flaw. In what specific scenarios would such latency lead to catastrophic failure?
12)
Autonomous driving and robotics: When an autonomous vehicle is travelling at 60 km/h, a delay of just 0.1 seconds results in a deviation of approximately 1.7 metres. If, at the moment a child darts out, the camera feed is sent to a server in the US to decide whether to ‘apply the brakes’, the vehicle will undoubtedly hit the child. In life-critical fields, only delays of ‘10 milliseconds or less’ are permitted.
13)
AR glasses (smart glasses) overlay information onto the user’s field of vision. The moment you turn right, the AI-generated information must slide to the right, in sync with the real-world scenery. If a time lag occurs between the visual information and the AI’s processing, the human vestibular system goes into panic mode, causing severe ‘VR sickness’.
14)
Natural conversation with voice agents: In human conversation, if the other person’s response is delayed by more than ‘200 milliseconds’, people feel a sense of unease, wondering, ‘Has the connection dropped? Are they listening?’ If a round trip to the US takes 300 milliseconds, a smooth conversation is absolutely impossible.
15)
Endless ‘behind-the-scenes’ meetings between AIs: This has the greatest impact. For example, if you give an instruction such as ‘Use this proposal to analyse the competition and create presentation materials’, behind the scenes, the AI responsible for drafting the outline and the design AI will begin a series of back-and-forth discussions at breakneck speed, asking ‘What do you think of this?’ or ‘Could you fix this part?’. If this communication has to travel across the Pacific and back, just the 50 exchanges would result in a 5-second delay due to the transmission alone. In the coming era, the ‘time spent by AIs conversing amongst themselves behind the scenes’ will become overwhelmingly longer. Unless those conversations can be conducted entirely within Japan—and indeed, within the same city—the system will be useless.
16)
Furthermore, there is the legal barrier of ‘data sovereignty’. Regulations are increasingly restricting the cross-border transmission of highly confidential data—such as data from the financial and healthcare sectors—to overseas servers. All these factors lead to the conclusion that ‘the AI’s brain must be physically located right next to the user’. That is precisely why they are targeting the very heart of Tokyo.
17)
Chapter 3: Blasting Mountains and Casting Aside Light: The ‘Madness’ of the Financial Industry
‘The closer you are physically, the faster it is. That’s why, even if it costs more, we place it close to the user.’ Put into words, this seems like a matter of course, but in fact, there is an industry that has been pushing this ‘battle against physical distance’ to its absolute limits for well over a decade—long before the AI boom began.
This is the world of ‘high-frequency trading (HFT)’ in the financial industry.
18)
They buy and sell shares millions of times a day at a speed of one ten-thousandth of a second—a speed imperceptible to the human eye—skimming off the tiniest of spreads. If an order can be delivered just ‘one millisecond’ faster than a rival’s, it results in a difference in annual profits amounting to several billion yen. To ‘buy that one millisecond with money’, they have made investments that defy all reason.
19)
Level 1: Placing servers next to the exchange. First, they positioned their own servers ‘right next to’ the stock exchange’s main server. Astonishingly, in this world, a major row can erupt simply because ‘our cable is 30 centimetres longer than our rival’s (and therefore slower)’.
20)
Level 2: Blasting through mountains with dynamite. Next comes the speed of communication between cities. For communications between Chicago and New York, standard fibre-optic cables are slow because they meander along the roads. Consequently, one company dynamited a tunnel through the Appalachian Mountains—which stood in its way—to lay a ‘dedicated fibre-optic cable in a perfectly straight line following the curvature of the Earth’. The total construction cost was approximately 45 billion yen. The time saved was a mere ‘3 milliseconds’ for a round trip. It may seem madness to spend 45 billion yen for a mere 3 milliseconds, but the returns were so substantial that the project was a resounding success.
21)
Level 3: Discarding fibre-optic cables because they are ‘slow ’. The ultimate example was the decision to discard even straight fibre-optic cables because they were ‘slow’. As light refracts whilst travelling through glass, it is approximately 30 per cent slower than if it were travelling in a straight line through air. Consequently, they concluded that ‘it would be faster to transmit signals directly through the air via radio waves’. They began constructing a massive, straight line of microwave communication towers between Chicago and New York.
When performing the same task, the physically closer location will always win out. When humanity pours boundless capital into this law of physics, it reaches a level where mountains are blown up, and even the refraction of light is abhorred.
22)
Chapter 4: The Madness of Finance Has Come to Tokyo in the Form of ‘AI’
Once upon a time, this ‘madness of shaving off a millisecond’ was confined to the closed world of Wall Street. It had no impact whatsoever on us watching YouTube at home. However, this ‘extreme competition over physical distance’ has now descended into the eneral social infrastructure, disguised as AI. This is the very crux of what is happening right now.
23)
A massive data centre is under construction right next to Tokyo Tower. Why is it located in such a prime spot that land prices have skyrocketed? You’ve probably already guessed. It’s because ‘placing servers at the heart of Japan’s most densely populated area—right in front of users’ eyes—minimises physical latency’.
24)
Although they haven’t actually blown up a mountain, the concept is the same as that of the financial industry. What’s more, it is a ‘fortress specialised in AI inference’ that has even introduced state-of-the-art systems to pump coolant directly onto the semiconductors to dissipate the intense heat generated by AI. This is not just the story of a single company. Tokyo is currently buzzing with one of the world’s largest construction booms for data centres, with plans for small and medium-sized AI data centres in the city centre emerging one after another. One railway company, for example, has even begun a pilot project to house a data centre beneath an elevated railway line.
The ‘cloud’—which until now had remained out of sight—is taking the form of concrete structures and is physically moving in next door to our homes.
25)
Chapter 5: Exploited Infrastructure and a ‘Difficult-to-Put-into-Words Sense of Unease’
AI infrastructure is coming to our cities, and communication speeds are set to increase. So what’s the problem? In fact, alongside the rush to build data centres in city centres, serious friction is beginning to arise with residents. ‘What about the noise from the cooling fans?’ ‘Is the massive amount of electromagnetic radiation safe?’ ‘Will the local infrastructure be able to cope with the massive consumption of electricity and water?’ ‘Won’t these huge, windowless buildings spoil the landscape?’ ‘Isn’t this a waste of prime land, given that it doesn’t even create jobs?’ These are all legitimate concerns.
26)
And what I feel about this trend is a certain ‘indescribable sense of unease’. Logically, it shouldn’t be wrong, yet deep down I cannot bring myself to accept it.
The fundamental premise is that, from the perspective of foreign mega-tech companies, Japan is currently one of the ‘most lucrative countries in the world’. Partly due to the weak yen, land prices, construction costs, water, electricity and the salaries of highly skilled engineers are unbelievably cheap when viewed in dollar terms. Moreover, major power cuts are rare, public safety is excellent, and water resources are abundant. You’d be hard-pressed to find such high-quality infrastructure available at such a bargain price anywhere else in the world.
27)
Foreign-owned companies, having pooled capital from around the world, set their sights on Japan’s cheap, high-quality infrastructure and build massive fortresses in prime city-centre locations. They draw vast amounts of energy from Japan’s power grid, use Japanese water for cooling, and siphon off enormous profits back to their home countries. Meanwhile, the ‘hidden costs’—such as noise pollution, concerns about electromagnetic waves, the deterioration of the landscape, and the risk of soaring electricity bills due to supply shortages—are all borne by the residents who just happen to live there.
28)
Of course, they have purchased the land legally, pay the standard rates and pay their taxes. I am not advocating xenophobia along the lines of ‘foreign capital, get out’. From a business perspective, it makes perfect sense.
And yet, I am left with a sense of unease. The ‘cheap and stable infrastructure’ that our forebears painstakingly built up, and which we have maintained through our taxes and utility bills. It is the massive capital from outside that reaps the sweetest rewards and amasses immense wealth, whilst the noise, heat and strain on the infrastructure left behind are burdens that the people living in those towns must bear.
29)
This sense of unease at the fact that ‘whilst no rules are being broken, we are being structurally exploited’—this is the true nature of the ‘uneasiness’ I feel. In the US state of Virginia, which is leading the way, electricity bills for ordinary households are already rising due to the enormous power consumption of data centres. I believe Japan will undoubtedly follow the same path.
30)
Chapter 6: Solutions to Clear Up This ‘Uneasy Feeling’
So, are we simply powerless to do anything but stand by and watch this ‘infrastructure exploitation’ unfold? Certainly not. Having researched the matter, I would like to present two ‘personal solutions’ I have developed to rectify this imbalance.
31)
1. ‘Returning Profits to the Local Community’ through a Dedicated Tax System: The first approach involves the government changing the rules. For example, just as there is an ‘accommodation tax’ for hotels, a dedicated tax—such as a ‘data centre tax (special local tax)’—could be imposed on data centres that consume a certain amount of electricity or more. The revenue from this tax would then be directly allocated to ‘subsidising electricity bills for residents’ and to the development of local infrastructure. By reviewing the electricity tariff structure that currently favours large corporations, we can require businesses that place a burden on the local area to bear a commensurate cost, and return the savings to residents. This would create a convincing structure where ‘the more profit they make, the lower our living costs become’.
32)
2. Making ‘Community Benefits Agreements (CBAs)’ mandatory: The second approach involves requiring foreign-owned companies to enter into a ‘give-and-take’ agreement with the local community before purchasing land and constructing buildings. There is a concept known as the ‘Community Benefits Agreement’ (CBA), which has been introduced in some parts of the United States. In exchange for authorising the construction of a data centre, it imposes obligations to contribute to the local community, such as ‘providing the latest IT equipment free of charge to local primary schools and offering educational support’ or ‘converting part of the facility into green spaces or cafés open to the public’. The aim is to redesign what would otherwise be merely a ‘nuisance facility’ into one that brings tangible benefits to the local community.
33)
It is not realistic to indiscriminately drive out foreign-owned data centres as if they were ‘Black Ships’. What we should aim for is to take the initiative in crafting shrewd rules, stating: ‘If you wish to use Japan’s excellent infrastructure, you must pay this fee (in the form of benefits to the local community).’
34)
Epilogue: The Era of ‘AI as Our Neighbour’
A massive fortress rising next to Tokyo Tower. Current text-based AI could operate just fine from the United States. However, for next-generation AI—such as autonomous driving and communication between AI systems—physical proximity is essential. The ‘madness of low latency’ demanded by the financial sector will become the new standard.
35)
The ‘cloud’ is no longer up in the sky. Emitting immense heat and noise, it is moving in next door to your home. Whilst we are set to reap the overwhelming benefits of AI, we are also being challenged to consider how we will interact with our ‘new neighbour’ and establish the rules. The time has come to face up to this structural unease head-on, without resorting to the simplistic stance of ‘opposing foreign investment’, and to begin a robust debate.
36)
Next time you’re walking through town and spot a sterile, windowless building, I’d like you to use your imagination for just a moment. Imagine that, on the other side of those thick concrete walls, a ‘0.001-second race’ is unfolding with a frenzy intense enough to blow up a mountain. And imagine that, rather than being exploited, we can harness that energy to enrich our cities.
[Technology & Economy] Why next to Tokyo Tower? Investigating the "hard-to-articulate unease" surrounding the expansion of AI data centres into the city centre.
https://note.com/keisoumainte/n/ne1d0f5f4ecd8
Why is a massive foreign-owned data centre being built right next to Tokyo Tower? We discuss the reasons behind this trend of data centres moving into urban areas and the challenges involved.
https://www.youtube.com/watch?v=szb5nT0tTjI
Terrifying! AI Data Centres and their exact locations
https://www.youtube.com/watch?v=vWIAdEYDsMo
