

Clip farms have made data centres the internet’s favourite villain, one viral post, one protest, one moratorium at a time. Every one of those posts is served from a data centre.


They are building data centres in space to escape a fight they are losing on the ground. That is not a metaphor.
In the first half of 2026, the richest companies on Earth committed real capital to putting AI compute into orbit and under the sea, and when they explain why, the engineering comes second. The first reason, said out loud, is that up there and down there, nobody objects.
We spent a fortnight inside the accounts that farm this fight. This is what we found on the ground. It opens with the science fiction, because the science fiction is the tell. It moves to where the argument is actually being had, which is not where the industry is having it. Midway, it goes inside a clip farm, with a real operator, recorded, walking through the machine one move at a time. And it closes on the uncomfortable conclusion for anyone with capital in the sector: being right is not enough, because the room where perception becomes policy has moved, and the industry stayed behind. The companion brief The Wrong Numbers shows the facts are on the sector's side. This one is about why that has not been enough.
Clarity kills momentum. You want tension. Strong framing.A clip-network operator, recorded · read the intercept →


The wildest ideas in data centres are pulling in public attention, serious money and, as they travel, a little misinformation. Space and the sea are both real, both funded, and both distorting on the way to your feed.
The escape is real. SpaceX has filed to put up to a million data-centre satellites in orbit; Google, Blue Origin and the startup Starcloud are close behind; China is scaling a state-backed constellation and has already run a model in space. The pitch, said out loud, is that up there nobody objects.
The sea is the same move. China switched on a wind-powered underwater data centre off Shanghai this year; Microsoft proved the idea with Project Natick, then walked away. And this spring Peter Thiel led a nine-figure round into Panthalassa, an Oregon startup building wave-powered, ocean-going compute nodes. That last one is exactly where attention, money and misinformation meet.
Five flagship projects moving compute off the planet and under the sea. Each opens to its filing detail, the money and the claim, tagged Verified, Announced or Estimated.
While the industry issues statements, the argument is being won and lost in the replies. And the industry is not in the replies.
The same fight, on three different feeds. The panic runs hot at the top; the correction sits at the bottom, unread.
Instagram and TikTok posts are real, accounts hidden. The X example is typical of the format.







Real posts from public feeds, mid-2026. The same few claims, restated until they read as fact.
The sci-fi does not stay in the business press. It gets harvested. Faceless accounts run the streaming era’s content-farm playbook on data centres: clip, caption, repost, farm the outrage. The reach dwarfs anything the sector’s own channels manage. One collective’s reel on a Google data centre hit 2.6 million views before India had it restricted, and 404 Media has traced hundreds of AI-generated “Life in [State]” pages inventing farmers who fight data centres, purely for ad revenue. Across the platforms, the genre runs to an estimated 50 to 150 million views a month.
And the comment sections are not neutral. They are overwhelmingly hostile, and they are where the received wisdom is set: that data centres drink the water, boil the planet and give nothing back. It does not stop at outrage. The wins get their own victory laps, packaged for the algorithm, a moratorium turned into a montage.
That buried comment in the thread above is the whole problem. The accurate answer exists, it is often posted, and it is structurally invisible, arriving late, low on emotion, and buried by an algorithm that never rewarded it. The industry's instinct is to write a better technical rebuttal. The battlefield does not run on rebuttals. It runs on feeling, format and frequency, and on those three the sector is not competing.
To see the machine from the inside, our field investigator approached one of its operators, posing as a small clipper looking to scale. What follows is taken from the recording, edited only for length. The operator is anonymised; where a claim can be checked, it is tagged.
The sector defends itself in the outlets it trusts. The audience it needs is somewhere else, and the politicians already know it.
Ask where the data-centre industry makes its case and the answer is the trade press, the broadsheet business desk, the panel at the conference. Ask where the public actually forms its view and the answer is the feed. Those are not the same place, and the gap is now enormous. Trust in mainstream media has fallen to 30 per cent in the UK and 25 per cent in the US, a ten-year low, while 5.66 billion people use social platforms and nearly two-thirds of Britons under 35 now get their news mainly from them. The sector is buying billboards in a town the audience left.
Politics has already made this migration. The 2024 US cycle was fought on podcasts as much as on television, with Donald Trump's Joe Rogan appearance alone passing 50 million views, and the cycle now remembered as the podcast election. Closer to this report's subject, Andy Burnham's recent media run, LBC, Times Radio, Novara Media, the Guardian's Today in Focus and BBC Newscast, shows a serious politician working the modern surface area on purpose. The people whose job is winning public consent have concluded that the room moved. The data-centre industry, whose licence to build depends on exactly that consent, is behaving as though it did not.
The same sector, heard clearly by one audience and barely at all by the other. The gap is not competence. It is which room each side is standing in.
CNBC, the BBC business desk, the FT and LinkedIn are excellent at reaching investors, executives and analysts. For that audience the sector’s case is clear, sober and largely accurate, and it lands.
The household next to the substation, the resident scrolling at night, and the councillor or MP forming a view get their information somewhere else, the feed, and there it is vast, emotive and often false. The investor hears the truth; the neighbour and the lawmaker are fed the misinformation.
Traditional platforms, the BBC, the broadsheet, the bulletin, are ceding the room to new media and to individuals: podcasters, creators and people with an audience of their own. The politicians already followed them there. The data-centre industry stayed in the room the broadcast is emptying. Scroll the migration, and why, in my read, the sector’s approach will not hold.
The 2024 US cycle was fought on podcasts as much as television. Trump’s Joe Rogan appearance alone passed 50 million views. The room where public consent is won had already moved, and the winners knew it.
Andy Burnham’s media run, LBC, Times Radio, Novara, Today in Focus, BBC Newscast, is a serious politician working the modern surface area on purpose. Not a statement desk. The feed.
The migration is to personality and affinity: creators, long-form, comments and duets. High-trust, hard to counter, and nothing like a right of reply in the trade press.
Data-centre leaders defend the build in the business desk, the trade press, the conference panel. Correct, sober, evidenced, and aimed at a room the deciding audience has already left.
Being right is not enough when perception becomes policy in a room you are not in. The politicians who win consent are already on the feed. Until the sector shows up there, the facts keep losing to the caption.
This is the turn the whole piece rests on. The engineering case is sound and the numbers are on the sector’s side. The distribution is not. Right on the facts, missing from the fight.
Perception is not a soft problem to be managed after the fact. It is the upstream cause of the hard ones: moratoria, refusals, and a licence to build that quietly evaporates.
The history of infrastructure is full of technologies that lost the public before they lost the argument, and had policy turn against them regardless of the evidence. Germany's nuclear exit within months of Fukushima, the EU's block on GMOs, England's seven-year de facto ban on onshore wind, the 5G health panic, are the canonical cases: in each, sentiment hardened, and law and permitting followed the sentiment, not the science. Data centres are now walking the same path, faster, because the feed compresses the timeline.
The evidence is no longer anecdotal. A Gallup poll this March found 71 per cent of Americans oppose a data centre in their area, 48 per cent strongly, which makes them more unpopular than a nuclear plant. The most-cited reason, named by 50 per cent, was their effect on resources, the water-and-power story the companion brief shows is largely mismeasured. That sentiment is already converting into policy: the anti-data-centre movement has driven 93 local moratoriums across 35 US states, 62 still active and seven outright bans, from Seattle and Denver to permanent bans in New Jersey and North Carolina. Ireland paused new grid connections for four years; Amsterdam now allows no new data centres at all; Virginia's Loudoun County, the densest cluster on earth, stripped them of by-right development. And brands have begun to monetise it. When an 89-year-old camera company decides its sharpest marketing angle is attacking data centres, the culture has moved, and the sector is the villain by default.
This is the mechanism the industry underrates. A hostile comment section is not noise, it is the leading indicator of the next refused planning application, the next moratorium, the next line in a manifesto. Perception is the top of the funnel; permitting is the bottom.
Not every complaint is equal. Graded against the evidence, the backlash splits into fights the sector can win on the facts, and fights it can only win by delivering. The mistake is answering all six the same way, with a statistic.
The same medium, aimed the other way. On the left, a real campaign that cast data centres as the villain. On the right, an Entelligencia reimagining: the same quiet infrastructure, framed as the thing that lets the analogue life run.

The industry’s problem was never that its facts are wrong. It is that it brings facts to a fight about feeling, on a battlefield it abandoned.
The instinct after reading this is to commission another white paper on closed-loop cooling. That instinct is the disease, not the cure: a message perfectly designed for a room that emptied a decade ago. The sector has every advantage it needs, capital, genuine good news, and opponents whose best arguments are, on the merits, wrong. It is simply not showing up.
Showing up means going where the audience is, funding creators and native formats rather than a corporate account posting PDFs. It means fighting feeling with feeling, the jobs, the resilience, the scan that ran on that rack, because emotion is not beneath the argument, it is the argument. It means being fast and native, a correction in the first hour and in the format of the attack, not a right of reply three days later. And it means attributing the demand honestly and delivering a real local win, so the villain narrative loses its subject.
Concretely, it means stopping some things. Cancel the next conference panel and shelve the next white paper, and move that budget into ten creators on retainer, a rapid-response desk that can answer inside the hour, and one campaign that reframes the building the way the section above does. The reframe is not a thought experiment. It is the brief.
My own read is blunter than the analysis above. The escape to orbit is the honest admission underneath all of this: the sector would rather spend billions leaving the planet than spend a fraction of it winning the room.
On the current trajectory, that choice is a forecast of who writes the rules, the people in the comment section, or the people who never joined it. Being right has stopped being enough. The room moved, and the industry stayed behind.
So the mandate is plain. Data-centre stakeholders have to start talking to the world directly: stronger social campaigns, native marketing, real creators, and the sector’s own voice in the feed where consent is now won and lost. Fund it like it matters, move at the speed of the clip, and show up. Or keep handing the room to the people who already did.
And it should not fall to the sector alone: governments ought to teach how the modern world actually runs, the data centres behind it, the energy it costs, and the real environmental and technology trade-offs, starting in the classroom, while the industry funds that public understanding and fights with the right tactics rather than the old ones.
Losing the Room is a Sociogencia field study inside The Next Hotspot, an interactive read on where the world actually builds AI infrastructure. Chapters land through the summer, the full edition on 4 August, and new original analysis is already going live.