Whose Is the Memory in Your Computer?

Dear Young Comrades,

For a year, everyone who has tried to buy a computer, a phone or a games console has hit the same wall: memory prices. This week one of the people who built that wall explained, in his own words, why it is standing there. Sam Altman, the chief executive of OpenAI, said on John Collison's Cheeky Pint podcast that it is "quite terrible" for part of the world's limited DRAM supply to sit in everyone's own computer, used 5 per cent of the time. The remark reached millions of people through a post on X by Charlie Intel, and it produced a justified wave of anger.

In this essay I will first give a short summary, then go into the detail: what exactly Altman said, what the RAM crisis is and where it came from, what the remark means in class terms, and what we can do.

---

Summary: Six Theses

• What was said: Altman argues that the memory in personal computers sits "idle", and that in a time of scarcity memory should be put to use "with maximum utilisation in the cloud".
• What is left out: Personal computers did not create the scarcity. AI data centres did. The preliminary agreements OpenAI made with Samsung and SK hynix for the Stargate project can run to 900,000 DRAM wafers a month. That is about 40 per cent of global output.
• The logic: The discourse of "idle capacity" is not new. It is the digital-age counterpart of the language of the English landlords who enclosed the peasant's land because it was "used inefficiently", and of Locke, who counted America as "empty land".
• The real target: The personal computer is the last instrument of computation left in the hands of the ordinary person. Local artificial intelligence, open models that run on your own machine, is a small but real area of independence against subscription rent. Altman's remark is an ideological attack aimed at that area.
• The cost: Memory prices doubled in a year. Phone and computer shipments are falling, and the hardest blow falls on entry-level devices, that is, on students, young workers and poor households. In Türkiye the exchange rate and the taxes make this burden heavier still.
• Our position: The question is not "where the memory sits". It is in whose hands, under whose control, and to whose benefit. Our answer: a commons of computation, local and open models, public infrastructure, and organised labour in this field.

Now let us go into the detail.

---

What Exactly Did Altman Say?

The remark comes at about the 29th minute of the Cheeky Pint episode published on 29 October 2026. As it reached the press, Altman says this: it is "quite terrible" that part of the world's limited DRAM supply sits in everyone's own computer, used 5 per cent of the time; if you are trying to cope with an environment in which supply is this constrained, you want "maximum utilisation in the cloud"; "and for some reason many people's intuition is the exact opposite" (Shattered; Android Headlines; TweakTown).

Let us underline two points:

1. "Five per cent" is not a measurement. It is Altman's own estimate. No independent study supports it. And by what is "use" measured? Does an open browser tab, a compiler running in the background, or a video edit you will continue tomorrow morning hold memory "in vain"?
2. The remark is presented as a technical observation, but it is in fact a proposal about property. It is the polite form of "let your memory sit with us, and you reach it by paying rent".

The reaction came quickly. Gamers, video editors, programmers, people who run a model on their own machine: they all said the same thing. We buy a desktop computer so that we can do our own work without paying a monthly subscription (Android Headlines; Windows Central). This intuition is not "wrong", as Altman supposes. On the contrary, it is a very correct intuition in class terms.

---

What Is the RAM Crisis, and Where Did It Come From?

First, the basic information, for young comrades.

RAM (DRAM) is the working memory of the computer. It holds the data the processor is working on at that moment. It is in the phone, the laptop, the console, the server, everywhere. The number of companies in the world that make these chips does not exceed the fingers of one hand: Samsung, SK hynix and Micron hold the overwhelming majority of the market. So from the very start we are faced with a monopoly structure.

HBM (high-bandwidth memory) is the special memory stacked beside AI accelerators (Nvidia and the others). It is made in the same factories, from the same wafers. The difference is this: per bit, HBM consumes about three times the wafer capacity of ordinary DDR5, and its yield is lower (Wccftech). For the producers the profit margin is much higher. The result: capacity is shifted to artificial intelligence, line by line and factory by factory.

Let us look at the chronology:

• 1 October 2025: Samsung and SK hynix signed preliminary agreements for OpenAI's Stargate data-centre venture. OpenAI's demand could run to 900,000 DRAM wafers a month. That is about 40 per cent of global DRAM output (Tom's Hardware; TrendForce).
• The last quarter of 2025: Lenovo and Dell warned of price rises, and a "DRAM supercycle" began to be talked about.
• The first quarter of 2026: According to TrendForce, DRAM contract prices rose 90–95 per cent from one quarter to the next. Increases of up to 110 per cent were seen in consumer RAM prices, and of up to 147 per cent in SSD prices (Tech Insider; Wccftech).
• July 2026: According to Bloomberg, DRAM spot prices rose by about 700 per cent in a year. Kwak Noh-Jung, the chief executive of SK hynix, said the shortage would probably continue past 2030. According to IDC, data centres will absorb about 70 per cent of the memory chips produced in the world in 2026 (Tech Insider).
• The last quarter of 2026: The rise slowed, but it did not stop. TrendForce forecasts a further increase of 10–15 per cent in ordinary DRAM and of 15–20 per cent in NAND (Shattered).

The counterpart of this in daily life:

• IDC expects smartphone shipments to shrink by 12.9 per cent in 2026, and the PC market by 11.3 per cent. The heaviest blow will fall on entry-level laptops.
• According to HP, memory's share of the material cost of a laptop rose in one quarter from 15–18 per cent to about 35 per cent.
• Some producers, without changing the model name, are cutting the RAM or the storage in mid-range devices (Tech Insider).

In Türkiye the picture is harsher still. When the exchange rate, VAT and the special consumption tax pile on top of one another, headlines saying "buying a computer has now become a luxury" fall into the mainstream press (Habertürk). The price that meets the student who has to work while studying, the young programmer who works on their own laptop in a first job, and the worker's family looking for a tablet for a child's distance education, is the price of the Stargate contract.

Now let us set Altman's remark beside this picture. The head of the company that is the single largest source of the scarcity points to the memory on the user's desk as the party responsible for the scarcity. The person who started the fire scolds the neighbourhood: "why are you keeping the water in your house?"

---

The Long History of the "Idle Capacity" Discourse

Altman's language should sound very familiar to us. Capitalism has always justified its own birth with the same excuse: "You are using this resource inefficiently."

The English enclosures. From the sixteenth century to the nineteenth, the pastures, forests and fields that peasants used in common were declared "inefficient" and enclosed. The excuse was always "improvement". In the chapters of Capital on "so-called primitive accumulation", Marx shows the real meaning of this process: the means of production are taken from the producer's hands, the producer is turned into a labourer who is "free" but propertyless, and from then on they have to secure a living in return for access to means that belong to someone else.

Locke and land that "goes to waste". In the Second Treatise of Government, John Locke defines the indigenous land in America as land that "goes to waste". Land that is not worked, and not tied to the market, is empty, and therefore it can be seized. For centuries this argument was the philosophical passport of colonialism.

The sharing economy. In the recent past the same language came back as "open your idle car to Uber, open your idle room to Airbnb". The result: housing rents rose, the driver who owns the car was tied to the platform's algorithm, and the profit stayed with the platform.

Altman's remark about "memory used 5 per cent of the time" is the latest link in this chain. For something to be counted "idle", it has to be measured from the point of view of its exchange value. Your computer counts as idle for capital, because your using it makes no one pay rent. From the point of view of use value, that memory is valuable because it is ready in your hands at the moment you need it. The knife in your kitchen is also used 5 per cent of the day. No one describes this as "a terrible use of the world's steel supply". They do not need to, because no one is yet aiming to sell the cutting of a knife by subscription.

---

The Real Target: The Personal Computer and Local Artificial Intelligence

The timing of this remark is not a coincidence. In the last two years, open-weight models (derivatives of Llama, Mistral, Qwen, DeepSeek and others) have become small enough, and good enough, to run on personal computers. Today, on a machine with 32 or 64 GB of memory, it is possible to run a language model that does the work without sending a single piece of data to the cloud and without paying a monthly subscription. In PCMag's reading, Altman's remark in effect means "running artificial intelligence locally is a waste of memory" (Shattered).

Let us set out plainly what this means in class terms:

• The personal computer is the last general-purpose means of production left in the hands of the ordinary person. A tech worker, a graphic designer, a student, a union volunteer produces with it. When this instrument turns into a subscription terminal, the ownership of the means of production is once again torn away from the producer.
• The cloud is the enclosure of computation. "Maximum utilisation" in the cloud means that the capacity of your memory is pooled in someone else's safe and rented back to you hour by hour and token by token. This is not a question of efficiency. It is a question of rent.
• The cloud is also surveillance. A model that runs locally shows your data to no one. A model that runs in the cloud passes every question you ask, every document, every draft, through the company's server. The walls of the digital panopticon are built of memory chips.
• Monopoly feeds on scarcity. For the memory producers this crisis is a period of record profit. For the AI companies, scarcity is a competitive advantage that makes local solutions, which could be rivals, more expensive. Scarcity is not a fault in this structure. It is part of how it works.

Earlier, writing on Meta's 145 billion dollar investment and the AI bubble, we asked "who gained while it inflated, and who pays if it bursts". The RAM crisis is one of the first concrete answers to that question: even while the bubble is inflating, the person who pays the cost is the student and the worker who cannot renew their computer. As we also showed in the essay on the price of the fund heist, the rapid accumulation of capital is always built on a resource drawn from somewhere. This time the thing drawn off is millions of people's access to the device on their desk.

---

Two Languages, Two Worlds

Altman's language	Its class meaning
"The world's limited DRAM supply"	What limits the supply is the production decisions of three monopolies, and the Stargate contracts that tie up as much as 40 per cent of global output
"It is used 5 per cent of the time"	Use value that produces no exchange value is counted as "waste"
"Maximum utilisation in the cloud"	The enclosure of computation, and its conversion into rent
"An environment of constrained supply"	Scarcity is not a fault of monopoly. It is a mechanism of profit
"People's intuition is the opposite"	The intuition of the masses is correct: the instrument should stay in our hands
"Efficiency"	Efficiency for whom? For the rent flowing into the safe
Personal computer = idle capacity	Personal computer = the last means of production left in hand
Local artificial intelligence = a waste of memory	Local artificial intelligence = computation without surveillance and without rent

---

What Would Someone Who Really Wanted Efficiency Do?

Inside Altman's remark, read the other way round, there is a true kernel: a scarce resource has to be used rationally by society. What would be done if social efficiency were really the aim?

• Needs would be ranked first. The device needs of schools, hospitals, municipalities and students would be put ahead of an AI company's model-training race, whose end is uncertain.
• Capacity would be used for a plan, not for a race. The same model being trained separately by five monopolies that do not know of one another is a "terrible" waste of resources in the real sense. We will not hear this from Altman's mouth.
• A common computing infrastructure would be built. Computing centres used in common by universities, public institutions and cooperatives, and open to democratic control. This is exactly the direction proposed by what we have written on public artificial intelligence and by the Socialist AI Manifesto.
• The memory chip would be treated as a strategic public resource. Production decisions would be taken according to social need, not by the boards of three companies.

So the question is not the duality of "cloud or personal computer". The question, whether in the cloud or on the desk, is in whose hands the power of computation is, under whose control, and to whose benefit. A common cloud, under democratic control, working for need and not for profit, would serve us all. Today's cloud is not a commons. It is a property ringed with a fence.

---

Concrete Tasks

• Protect the instrument in your hands. Repair your computer, upgrade its memory, make use of second-hand devices. Do not throw away devices that are called "obsolete". Pass them to someone who needs them. The struggle for the right to repair is here a class struggle directly.
• Spread local and open software. Learn, and teach, open models that run on your own machine, and free software. At school, in the union, in the association, build a culture of computation without rent and without surveillance.
• Build common spaces of computation. Associations, unions, student collectives and cooperatives can set up shared servers and pools of shared hardware. A knowledge commons is incomplete without a hardware commons.
• See the price as a political question. The price of a computer and a phone is not only a "market" news item. It is a question of the right to education, the right to communication and the right to work. Device support, tax relief and public device programmes should be demanded for students and workers.
• Organise as a tech worker. It is also workers who build the data centres that created this crisis, who train the models, and who keep the servers standing. As long as capital sits alone at the table of decisions, the memory, the model and the future are theirs.
• Expose the discourse. Tell the people around you about the property relation behind the words "idle capacity", "efficiency" and "maximum utilisation". Share this essay, discuss it, criticise it, develop it.

---

Dear Young Comrades,

Sam Altman's remark is not a slip of the tongue. It is a moment of candour. In the world the AI monopolies imagine, there will be no means of production left in your hands. There will be only a screen, a subscription and a bill for use. Your memory will be in their data centre, your data on their server, and your habits of thinking in their model.

But the intuition of the masses, that "exact opposite" intuition Altman complains about, tells us something: people want to keep the instrument in their hands. When this intuition turns into an organised consciousness it becomes a demand, and when the demand is organised it becomes a force.

Those who enclosed the land also said "efficiency". History also wrote how those fences were pulled down. This time it is we who will rebuild the commons.

The machine is new, the chain is old. Knowledge belongs to everyone; so does memory.

---