Memory Farms in the Valley of Heart's Delight

Before it was Silicon Valley, the Santa Clara Valley had another name. Promoters in the early twentieth century called it the Valley of Heart's Delight: miles of prune, apricot, peach, and cherry orchards between the bay and the hills, among the largest fruit-growing and canning regions in the world. Old photographs show the valley floor in bloom from end to end. The original silicon valley companies were located in the blockhouses of a chicken farm.

The orchards are almost entirely gone. In their place came the electronics industry: Hewlett-Packard, then the semiconductor firms, then Fairchild in 1957 and Intel in 1968, until a trade journalist named Don Hoefler gave the place its new name in 1971. The valley traded fruit for silicon in about a generation. It also picked up a legacy along the way: Santa Clara County has more federal Superfund sites than any other county in the country, most of them left by the chip plants' solvents in the groundwater.

Now a third change is under way, and it's arriving faster than either of the first two. The buildings are called data centers, but the name undersells what's inside. What fills them now is less a store of data than dense concentrations of AI accelerators and the high-bandwidth memory stacked around them. They're warehouses of memory devices, drawing power on the scale of small cities and shedding the heat that power becomes. The city of Santa Clara already hosts one of the densest clusters of them in the country, drawn partly by the rates of its municipal utility, which run well below what PG&E charges its neighbors.

Utilities across California and the country report requests for gigawatts of new data-center load. Memory makers have been shifting production toward the chips AI needs, and prices for ordinary memory have climbed as a result. These buildings are memory farms, and they're being planted in the same soil that once grew the orchards.

This essay asks what a social libertarian should make of that. The quick answer is to have no opinion about it, since it's someone's business and businesses shouldn't need permission. That answer is mostly right, and it isn't enough.

Organic and inorganic growth

This collection of essays has looked at this valley before. Toward an Authentic Santa Clara Valley described it as two places laid over each other. One is a mosaic of distinct towns, from the wealthy hillside enclaves to San Jose's ten council districts to the rural reaches of Morgan Hill and Gilroy, each with its own weather and position on the Pacific fog belt, its own watershed, and its own sense of itself. The other is Silicon Valley proper, "sprawling with a constant influx of new people and a kind of growth of which few can be proud or feel at home." That essay borrowed Murray Bookchin's distinction to name the difference. Authentic politics is organic, "the activity of a public body–a community–just as the process of flowering is an organic activity of a plant." Party machinery is "inorganic as the state itself–an excrescence of society that has no real root."

The same distinction holds for places, and not only for politics. The Jute Mill of Kolkata set two parts of one city side by side. Barabazar grew without a master plan, into fine-grained, overlapping, mixed uses and a dense street life. Salt Lake City and New Town were laid out from above, on straight streets and straight blocks, with far less diversity and far less life. Both are Kolkata. One grew, the other was installed.

The Santa Clara Valley's first transformation was mostly the first kind, though not entirely. The electronics industry grew out of Stanford's engineering school, small firms spinning out of other small firms, and talent clustering because talent was already there. Federal defense contracts also poured into the region during the Cold War and steered a good deal of what got built, so the story was never purely organic. Still, the orchards gave way one parcel at a time, as owners sold to buyers who could pay more for the land. The valley that resulted was built by thousands of separate decisions.

The memory farms are arriving differently. They're built by a handful of very large firms, each racing the others, on timelines measured in quarters. Increasingly they come with help from the state: special utility rates, tax incentives, permits sped up by order from above, and grid upgrades whose cost is spread across every ratepayer. Many are announced under code names, through shell companies, with local officials bound by nondisclosure agreements until the deal is done. Federal orders in 2025 directed agencies to speed data-center permitting, and states compete to offer the best package. That is not one parcel at a time. It's closer to New Town than to Barabazar.

Money without limits, development without roots

The subsidies and secret deals are the visible part of the imposition. Underneath them is something larger: the money that's paying for all of this, and where it came from.

Each round of money creation over the last two decades came with a narrow justification. In 2008 it was a financial system on the edge of collapse. In 2020 it was a pandemic that shut the economy down. The Federal Reserve's balance sheet went from under a trillion dollars before 2008 to about four and a half trillion after the first rounds of "quantitative easing," and then to roughly nine trillion by 2022. The broad money supply grew by about forty percent in the two years after early 2020. Interest rates sat near zero for most of the period from 2009 to 2022. Each step was presented as an emergency measure, temporary, and limited to its purpose.

Money doesn't stay where it's aimed. Richard Cantillon noticed this in the eighteenth century: new money doesn't reach everyone at once. It enters at a particular point, through banks, bond markets, and the holders of assets, and the people nearest that point spend it first, before prices rise. The effects run far beyond the stated intention. Asset prices outrun wages. Those who already own stocks, buildings, and land grow richer for reasons that have nothing to do with anything they produced. Ordinary prices follow last, as they did when consumer inflation reached 9.1 percent in June 2022. Hayek and Mises had a name for the deeper effect: when credit is made artificially cheap, capital flows into projects the real savings of a society can't support, chasing the largest future return and not what the people already living somewhere need. They called it malinvestment.

The memory farms are where a large share of that money has gone. A decade and more of near-free money concentrated enormous capital, and enormous stock-market valuations, in a handful of technology firms. Rates rose after 2022, but the war chests built in the cheap-money years didn't shrink. Those firms are now spending hundreds of billions of dollars a year on AI infrastructure, racing each other, with lenders and private-credit funds financing the buildings on top of that. The scale of the buildout isn't set by what the Santa Clara Valley, or any valley, needs from its computers. It's set by what capital accumulated in that era is looking for: the next large return.

That money also lands on land. Cheap credit capitalizes into the one thing nobody can make more of. The parcels near substations and transmission lines, the ones with water rights and zoning that allows heavy industrial use, are bid up by buyers who can borrow against the promise of what they'll hold. The gains go to whoever already held the ground. The costs go to everyone else who needs that land, that power, and that water.

So the memory farms are doubly unrooted. They arrive on terms set by the state, with special rates, socialized costs, and permits sped up from above. And they arrive at a scale set by money created in Washington, not by anything the valley grew. That's why they have so little organic heritage. They weren't planted here because the valley needed them. They were planted here because this is where the money went, and where the power was cheap.

What a social libertarian doesn't oppose

It's worth being exact about the starting point, because the easy positions on both sides are wrong. A social libertarian doesn't oppose a kind of business. Data centers aren't a sin, and building one isn't a harm to anyone in itself. A landowner who wants to put servers on a parcel, pays for the power, water, and land they use, and harms no neighbor in the process owes no one an apology. The instinct to ban a category of building because it's new, large, or ugly is the same instinct this collection has argued against in zoning, in licensing, and in the "sin" trades. It doesn't become a better instinct because the business is a tech company.

But two conditions change the question, and the memory farms meet both.

The first is scale. Some enterprises are big enough in their need and their impact to shape a community's destiny. A shop doesn't do that. A factory town's single factory does. When one kind of building can claim a large share of a region's electricity, set the price of power for every household, draw on the same water as the farms and cities, and absorb the grid capacity a region needs for everything else it might want to become, it has stopped being one business among many. It has become a decision about what the place is for. That decision belongs, in some real measure, to the people who live there.

The second is imposition. A business that arrives by paying full price for what it uses is growing organically, however large it gets. A business that arrives on special rates, socialized grid costs, tax breaks, sped-up permits, and secrecy is being imposed, and the imposition isn't the market at work. It's the state choosing a winner and charging everyone else for the choice. A social libertarian's objection isn't to the servers. It's to the subsidy, the preemption, and the nondisclosure agreement.

Put those together and the position becomes clear. We don't oppose memory farms. We oppose memory farms that couldn't exist on their own terms, at a scale that decides other people's futures, installed by arrangements those people never agreed to.

Three things that seem to be in tension

There are three claims here that look as if they can't all hold.

A free society doesn't ban kinds of business. Whether a data center gets built is between its owner and the people it trades with.

This kind of business can reshape a region's destiny. Its demand for power, water, land, and grid capacity is large enough to set the terms of everyone else's life.

A community has a say over its own destiny. The people who live with the consequences should have standing to decide them.

The first and third seem to collide over the second. Either the community gets to say no, which looks like banning a business, or the business gets to come, which looks like overriding the community.

The way out is that the first two were never in conflict with the third. Most of what makes a memory farm reshape a region's destiny is exactly what's imposed. Strip away the imposition and the scale disciplines itself. Charge the full land value of the site, the full cost of the grid capacity it claims, and a real price for the water it draws, the natural rents Single Tax for the Living World argued belong to everyone. End the special rates and the socialized upgrades. Hold the permits to the same process as anyone else's, in public. A memory farm that still pencils out on those terms is paying its way, and the community is being compensated for what it gives up. One that doesn't was never really a market decision.

That's the Georgist answer to the data-center boom, and it's more useful than a moratorium. A moratorium bans the business by type. Charging full rents removes the subsidy and lets the business decide whether it still wants to come.

How to assess community opposition

That still leaves the hardest practical question. When a community objects to something this large, how much weight should the objection carry? Opposition to any project gets loud, and loudness isn't the same as harm. Some of it is a few neighbors who would object to anything, and some is a campaign run from elsewhere. Some of it is a real community recognizing that its future is being decided for it. A social libertarian needs a way to tell these apart. Counting signatures doesn't do it.

Three Blind Assessors used the old story of the blind men and the elephant to describe land assessment: each method touches one part and mistakes it for the whole. Community opposition to a memory farm looks the same. The neighbor hears the cooling fans at night. The ratepayer sees the electric bill rise. The farmer watches the water table. The planner sees grid capacity disappear that the next housing project needed. Each is touching one part of the same large thing, and each sounds, to the others, like a special interest. Assessing the opposition means putting the parts together, the way a good assessor reconciles several methods into one value.

A few tests help.

Count stakes, not voices. List who bears each cost, and how much: noise and heat for the nearest parcels, rates for every household on the grid, water for every user of the basin, lost capacity for everything else the region might build. Opposition from people carrying a real, identifiable cost deserves weight in proportion to that cost. Opposition from people carrying none deserves a hearing and not a veto.

Ask whether the opposition is organic. Bookchin's distinction applies here too. Opposition that grows out of the community itself shows up in the places that community already governs itself: the neighborhood association, the water district board, the utility commission, the city council's public comment. That kind of opposition is the body politic responding to something happening to it. Opposition mobilized from outside, by a national campaign or by a party looking for an issue, is inorganic, however loud it gets. So is support mobilized the same way, the industry coalition or the regional booster group speaking for a public it didn't consult. Neither side's inorganic voice should be counted as the community's.

Check whether the costs are priced. Where the project pays full price for land, power capacity, and water, much of the opposition is already answered: the community is being compensated through the rents it collects. Where those costs are subsidized or spread across ratepayers, the opposition is pointing at a real transfer, and it should be taken seriously.

Watch what the project needs in order to happen. A project that requires a code name, a shell company, and a nondisclosure agreement for local officials is telling the community it expects opposition and would prefer to arrive before that opposition forms. Secrecy is evidence. It doesn't prove harm, but it shifts the burden onto the project.

Ask whether the scale is the community's own. A data center sized to serve the region's own businesses and residents is one thing. A cluster sized to serve a global market, using the region's power and water as a cheap input, is another. The second is the one that reshapes a place's destiny, and the one where local consent matters most.

Decide at the altitude of the cost. The noise belongs to the neighborhood. The water belongs to the basin. The rates belong to whoever owns the utility, which in Santa Clara's case is the city. Each cost should be decided by the body that bears it, in public, and not preempted by a state or federal order written to speed the projects through. That's the municipalism this collection keeps returning to, applied to the largest thing currently arriving in the valley. The earlier essay on the valley ended with a list of ordinary actions: attend the commission meeting, comment on the agenda, serve on the water and sewer district boards, put a position in writing. Those are where a memory farm's costs should be weighed, by the people who carry them.

None of these tests says "no data centers." They say that a community facing a change large enough to decide its future gets to see the whole elephant before it's moved in, and gets paid for what it gives up.

Why the valley is the place to see it clearly

The Santa Clara Valley is a useful place to think this through, and not only because it's where the question happens to be arriving. Historically it's one of the most uncontrolled, and least community-focused, development experiments anyone has run. Nearly everything this essay is asking about has already happened here once, at speed, with the results on the record.

In the twenty years after the Second World War the valley went from orchards to suburbs and factories faster than any plan could follow. San Jose's city manager in those years, Dutch Hamann, annexed land in every direction to win the race for tax base, and the city grew from about seventeen square miles to well over a hundred. Subdivisions went in wherever an orchard sold. Chip plants went up on the old farmland with almost no rules about what went into the ground. Then came wave after wave of capital, from the semiconductor era through the personal computer, the internet, and social media, each arriving faster than the towns could decide anything about it.

The experiment produced real gifts, and they were mostly what came from leaving people free to try things. Startups could be founded without permission, engineers could leave one firm and start another, and software was nearly unregulated. That freedom is a large part of why the valley changed the world. It also produced bills, and those came due later and fell on people who never agreed to them. Solvents from the chip plants of the 1970s and 1980s left Santa Clara County with more federal Superfund sites than any county in the country, and the plumes are still being pumped out of the groundwater. Jobs grew far faster than housing, so home prices became among the highest in the nation, and commutes stretched into the Central Valley. The orchards that gave the place its name are almost entirely gone.

At each stage the growth answered to capital and to the cities' race for tax base, not to the communities it landed in. The lesson isn't that the valley needed a master plan. The places that got master plans, like Kolkata's New Town, didn't do better. The lesson is that the valley's growth outran any way of pricing what it cost and any forum where the people bearing those costs could weigh in. The freedom to build was mostly right. What was missing was a way to charge for what the building used up, the groundwater, the land, the road capacity, and to collect that charge for the people who lost it.

The memory farms are the newest round of the same experiment, and they belong to it more than anywhere else. They run on chips this valley invented. Robert Noyce worked out the integrated circuit at Fairchild in 1959. When he and Gordon Moore founded Intel in 1968, their first business was memory, and Intel's 1103 of 1970 became the best-selling memory chip in the world. The accelerators filling the new buildings are designed by NVIDIA, headquartered in Santa Clara. In a rural county in Georgia or the desert outside Phoenix, a data center arrives as a wholly foreign industry. Here it's the valley's own crop coming home.

That's what makes the valley a clear test. Take away "foreign industry imposed on a place that never wanted it," and what's left is the question the valley's history already answered once: what happens when a fast-growing industry draws on shared power, water, and land faster than anyone can price those costs or decide them in public? The valley has seen the answer in its groundwater and its housing. It shouldn't need to learn it a second time.

How seriously to take opposition here

The valley's history cuts both ways for the community opposition it will produce.

Some of it deserves more weight here than it would elsewhere. Opposition from engineers and operators who know exactly what these buildings draw, and how the load estimates get made, isn't the fear of the unfamiliar. It's expert local knowledge, the kind Hayek said a distant planner can never fully gather. And the valley remembers its own experiment. A community still pumping the last one's solvents out of its groundwater has standing when it asks who pays for the costs that show up later.

Some of it deserves less. This valley has also spent fifty years using public process to block the apartment down the street, protecting land prices that cheap money inflated and that residents count as their savings. Opposition to a memory farm that comes from the same reflex, nothing new near my property, ever, isn't organic resistance to an imposition. It's the windfall defending itself, and a social libertarian should say so.

The tests from the previous section sort the two. Opposition that names a real cost the project isn't paying for, such as the grid upgrade spread across every Silicon Valley Power customer, the water drawn from a basin the valley spent decades cleaning, or the rate discount no ordinary business gets, should be taken seriously, and taken most seriously when it comes from people who know the technology from the inside. Opposition that would object just as loudly if the project paid full rent for everything it used deserves a hearing and not a veto. The valley's own heritage doesn't settle whether a memory farm belongs here. It settles that the answer has to come from the prices, the public meetings, and the people who know these machines best, not from either the industry's sense of entitlement or the neighborhood's reflex.

Responders, not printers

None of us prints the money, and none of us is building the memory farms. We live inside the system that does both: we're paid in its dollars, we borrow at its rates, our retirement accounts hold shares in the companies doing the building, and most of us use the AI services the buildings exist to run. A town council can't change what the Federal Reserve does or what a technology company decides to spend. So the fair question isn't what we'd do if we ran the system. It's what ordinary people should be expected to do as responders within it, and still hold on to liberty when a memory farm is proposed down the road.

Don't pretend to purity. Owning an index fund that holds these companies, or using the services they sell, isn't hypocrisy. Liberty doesn't ask anyone to stand outside the economy they live in. It asks something harder: not to mistake the system's distortions for anyone's virtue, and not to use political power to protect them, whichever side of the deal we happen to be on.

Don't sell the commons for the tax base. A data center carries a huge assessed value and pays a large property tax bill while employing very few people. That makes it tempting for a town to court one with rate discounts, tax breaks, and fast permits, and to treat the revenue as found money. That trade spends the shared things, grid capacity, water, quiet, land that could hold something else, to buy a revenue stream. It's the same windfall logic as cheap money, at the scale of one city. Responders at the council meeting can ask whether the deal would make sense if the project paid full price for everything it used.

Let the community collect the rents. The remedy this collection of essays keeps returning to applies directly. Tax land value, charge for water as a natural rent, and have large loads pay the full cost of the grid capacity they claim. That's how a community takes back the land-and-resource channel through which cheap money becomes private windfall. It's also how a memory farm proves it belongs: if it still pencils out after paying its rents, it's paying its way.

Don't let the windfall inflate the town. Data-center tax revenue swells a city's budget without anyone voting for more government, and the bureaucracy built to spend it outlives the boom that funded it. When a program is funded by one large taxpayer, the responder's job is to ask whether it would exist on the town's ordinary means. When an incentive is promised as temporary, the job is to write in the date it ends.

Show up where the costs are decided. The utility commission decides who pays for the grid. In Santa Clara that's the city council, which owns Silicon Valley Power. The water district decides who gets the water. The planning commission decides what a parcel may become. Those are ordinary public meetings, and they're where a memory farm's costs get assigned to its builder or spread across everyone else. Organic opposition, and organic consent, happen there.

Keep the exits open. Households and neighborhoods that want to generate and store their own power, form their own cooperatives, or hold their savings in forms the printing press can't quietly dilute should be legally free to do so. Separation of Currency and State argued the monetary half of that. The energy half matters just as much now that data centers are bidding up the price of every kilowatt. A responder doesn't have to make those choices to defend the freedom to make them.

We can't stop the printing, and we shouldn't try to ban the buildings. We can refuse to let borrowed money and backroom incentives be the only things that decide what the valley becomes.

The valley's third crop

The Santa Clara Valley has replanted itself twice. The first time, orchards gave way to electronics, mostly one parcel at a time, with costs that showed up decades later in the groundwater. The second replanting is happening now, at a pace and scale the first never had, and much of it is arriving on terms the people who live there never set.

A social libertarian has no quarrel with servers, or with memory chips, or with the companies that need them. The quarrel is with how they arrive: financed by money no one here created, subsidized, preempted, and kept secret until it's too late to object. The answer isn't to ban the crop. It's to make it pay full rent for the soil, the water, and the power it uses, to decide each of its costs at the level where people bear them, and to look at the whole elephant before saying yes. A valley that does that may still end up full of memory farms. If it does, it will be because the farms paid their way and the people agreed. That's the difference between a valley that grows and one that gets installed.

---

A note on how this piece was written: the subject, the questions it asks, and the positions it takes are mine. Much of the research, the examples, and the sentences were drafted by an AI model working from that direction, and then edited by hand. I'd rather say that plainly than have a reader guess at it.