You Used Fifteen Data Centres Before Breakfast

When you unlocked your phone this morning, a small piece of a data centre confirmed you were you. When you checked the weather, another confirmed the forecast. When you tapped your card at the cafe, three of them processed the transaction between your bank and the shop’s bank. When you streamed a podcast on the train, a fourth held the file, a fifth routed it to you, and a sixth billed the streaming service for the bandwidth. By the time you sat down at your desk, you had touched more than a dozen. You did not see any of them. That is the point of a data centre. It is a piece of infrastructure that works by being invisible.

The buildings themselves are anything but invisible. Each one is the size of several warehouses, cooled by industrial systems that run continuously, powered by electricity in quantities that used to be described as for a small city, and increasingly built in places where the local community did not know the buildings were coming. Arc Brief has spent thirteen issues describing what happens in the places where these buildings arrive. This issue is the one that describes what happens because they exist.

The count below is that morning, in data centres. Fifteen touchpoints. Fifteen buildings or clusters of buildings. Each icon marks an action on a screen that required physical infrastructure somewhere on the planet to handle it. Click any icon to see what it touched.

Most conversation about data centres treats them as an abstract technology story. That framing hides how directly they touch the reader.

Netflix does not have movies stored in a warehouse somewhere waiting to be posted to you. Every scene of every show sits in a data centre and streams to your screen frame by frame the moment you press play. The recommendation you see when you open the app was calculated in a data centre in the last few seconds. YouTube works the same way. Spotify the same. TikTok the same.

WhatsApp carries approximately one hundred and fifty billion messages a day. Each one travels through a data centre before it reaches the recipient. iMessage, Signal, Telegram, WeChat — all the same. Gmail stores approximately one and a half billion active users’ email histories on data centre servers. If the buildings holding your email vanished tomorrow, so would every message you ever received.

Google Maps does not have a map of the world on your phone. It has a small window into a much larger real-time map that sits on servers, updated every second by traffic data flowing in from millions of other phones. When it tells you the road ahead is congested, that answer was computed in a data centre a few hundred milliseconds ago. Uber, Grab, Ola, Didi — every ride booked on every one of them matches driver to passenger in a data centre before the notification arrives on either phone.

Every online banking transaction — every card tap at every terminal, every UPI transfer in India, every PayNow in Singapore, every OSKO in Australia — moves through data centres run by the banks and the card networks. When your card is declined at a coffee shop because you tapped it too fast twice, that decision was made by a computer in a building you have never seen.

ChatGPT, Gemini, Claude, Copilot — every AI response you have ever received was computed inside a data centre. The building was almost certainly not in your country. The electricity that powered the computation might have been enough to run a household for the day. This is not a criticism. It is a description.

These are the everyday examples. The heavier examples are where the counterfactual matters more.

Genome sequencing — the technology that has, in the past decade, turned some previously untreatable cancers into treatable ones — produces roughly two hundred gigabytes of data per patient sequenced. That data has to be stored, compared with reference genomes, run against clinical databases, and interpreted by algorithms trained on millions of prior cases. Every step of that pipeline runs in a data centre. Without it, personalised cancer treatment as it exists today does not exist.

Drug discovery now uses machine learning to model how candidate molecules will interact with human proteins, compressing what used to be five years of laboratory work into months of computation. The computation happens in data centres. Recent approved treatments for previously untreatable conditions — some rare paediatric diseases, some late-stage cancers — depended on this pipeline.

Weather prediction for typhoons in the Philippines, cyclones off Queensland, and monsoons across South Asia runs on supercomputer clusters inside data centres. The twelve-hour warning that lets Manila prepare for a landfall was computed by machines that would fill a factory. Without them, warnings drop from twelve hours to two. Evacuation planning becomes impossible. Deaths from tropical storms rise sharply.

Public health surveillance — the systems that detected the unusual respiratory cluster in Wuhan in December 2019, that would detect the next one, that monitor for antibiotic resistance and vector-borne disease patterns — all run in data centres. So does the vaccine cold-chain monitoring that keeps doses viable through the tropical distances of Southeast Asia. So does the aircraft routing that keeps six thousand flights over the Pacific from colliding at any given moment.

None of this is dramatised. None of it is future. All of it is a Tuesday.

The bill for the infrastructure above is not primarily paid in money. It is paid in electricity, water, silicon, and land.

Every photo you have taken since your first smartphone sits on a data centre server somewhere, backed up at least three times, in geographically distributed facilities so that no fire or flood loses it. Every email you have ever received. Every location ping your phone sent while you carried it. Every message on every messaging app. Every corporate document produced by any company legally required to keep records for seven or ten years. Every video ever uploaded to YouTube, still available a decade later.

The compounding cost of that default — of never deleting anything, of always backing up, of retaining indefinitely for compliance — is measured in electricity that could have run cities, in water for cooling that came from local watersheds, in silicon that had to be mined and manufactured, in land that had to be built on. One person’s cloud footprint uses a small but non-trivial amount of these resources every year. Multiplied across billions of people and hundreds of thousands of companies, it produces the infrastructure Arc Brief has been covering — the campuses in Sydney, Singapore, Tokyo, Mumbai, Vizag, Johor, Jakarta.

The AI layer sits on top of all of it. Every AI model you have used was trained on some portion of this stored data using electricity in quantities that would have powered towns. Every response those models generate uses more. This is not an argument against AI. It is an argument for seeing what AI actually costs at the infrastructure layer — because the cost is real and someone is paying it, whether or not the reader has noticed.

For most of the past decade, data centre coverage sat in the trade press. General readers skipped it. The stories used industry vocabulary — hyperscale, colocation, PUE, WUE — that did not translate. The buildings themselves were geographically remote, aesthetically anonymous, and belonged to companies most readers had never heard of. It was reasonable to conclude that this was somebody else’s story.

It was not. It is the reader’s story. The community that protested in Visakhapatnam is protesting the construction of the physical infrastructure that would host the AI models the reader now uses daily. The farmers in Mekaguda who filed a High Court petition are describing the specific facility that runs some portion of the enterprise cloud services their state’s IT industry depends on for its salaries. The residents of Kampung Cahaya Baru in Johor whose homes filled with construction dust are living next to the buildings that will host the workloads Singapore’s economy is offshoring across the causeway. The €5.8 billion in Irish data centre projects stranded in Dublin represents a set of facilities that would have hosted a meaningful slice of European enterprise IT — and their absence means those workloads are being routed to buildings somewhere else, in APAC, at scale. The $24 billion Blackstone paid for AirTrunk was for the Sydney-founded operator whose facilities host workloads that touch several actions the Australian reader took this morning.

Every one of those stories is a story about the trade-offs being made on the reader’s behalf. Which communities absorb the water cost. Which grids absorb the load. Which governments do the community consultation and which skip it. Which operators are held to a standard and which are not. Whose retirement savings finance the buildings and on what terms. These decisions are being made now, at pace, at scale, and largely without the reader’s attention.

The point of this publication has been that the reader’s attention is warranted.

Every one of the enabling capabilities described earlier — the streaming, the messaging, the mapping, the banking, the diagnosis, the discovery — is now a settled feature of how we live. We did not vote on this. We adopted it, one service at a time, over roughly two decades. Life restructured around it. Family patterns restructured around it. Work restructured around it. Healthcare, education, and finance restructured around it. The infrastructure did not arrive to serve a future we had planned. It arrived alongside a life we had already begun to build.

That life is not something the region can put back down. The generation that never lived without a smartphone is now in the workforce. The professional geography that video calling enabled is now the geography of most APAC families. The healthcare pathways that AI-assisted diagnosis unlocked are now the standard of care in every teaching hospital in the region. The financial infrastructure that moved cash payments into digital rails is now the infrastructure most APAC economies balance their books on. Rejecting the buildings that make this life possible is not rejecting a choice about the future. It is rejecting a choice we already made about the present.

This is not an argument against the community objections that have run through Arc Brief’s earlier coverage. Those objections are about the how — how the buildings get built, on whose land, on whose water, at what cost, with whose consent. Those objections are correct, and this publication has spent the past year describing why. This is an argument that the whether question was decided already, incrementally, without a debate, by every reader who opened a streaming app or accepted a transfer confirmation. Pretending otherwise makes the how question harder to answer, not easier — because it frames every conversation about the sector as if the buildings could be simply refused, rather than as if their siting, their standards, and their governance are the choices that actually remain.

The debate that treats the sector as a yes-or-no question is therefore a debate about a choice that is no longer on the table. The choice that is on the table — and the only one that matters — is what kind of buildings host the life we already have.

Arc Brief has spent the past year describing how these buildings get built, who pays, what gets lost, and who bears the cost. This piece is the one that names why any of it matters to a reader who has never opened Arc Brief before.

The infrastructure will be built because the life it enables has already been built. The choice that remains for the reader is not whether the sector should exist. It is whether the reader is going to notice — and hold to some standard — the buildings that were always going to be there anyway. Every news story about a data centre — every protest, every court case, every deal, every grid decision — is one small piece of that decision being made in real time.

The buildings behind the screen are running your morning. They deserve at least the attention the screen gets.