Vol. III, No. 56
Covering 27 January - 9 February 2025
Monday, February 10, 2025
Late edition · A Relentless publication
The Relentless Dispatch
All the fortnight that mattered, in technology and in the world, read next to what we were building at the time.
TECHNOLOGY

A cheap Chinese model humbles Silicon Valley, and $600 billion vanishes in a day

DeepSeek, an AI model built by a small Chinese lab for a fraction of the cost, matched the American giants and was given away free. Nvidia lost nearly $600 billion in a single day, and the whole "spend hundreds of billions" thesis wobbled.

The AI industry's confidence took its sharpest blow yet this fortnight, from an unexpected direction. A Chinese startup called DeepSeek released a model, R1, that performed comparably to the best American systems on many tasks, that it claimed to have trained for a tiny fraction of the hundreds of millions or billions the American labs spend, and that it gave away free and open for anyone to run. The market's reaction was violent: on January 27, Nvidia, the company whose expensive chips the entire American AI build-out assumes are indispensable, lost about $589 billion in value in a single day, the largest one-day loss for any company in history, as investors confronted a terrifying possibility, that the frontier of AI might not require the staggering spending, the vast data centres, the endless chips, that the whole $500 billion Stargate thesis rests on.

The panic may prove overdone; DeepSeek's cost claims are disputed, it likely used more and better chips than it admits, and its achievement rests partly on learning from the very American models it undercuts. But the shock was real and the questions it raised will not go away. If a small Chinese lab can approach the frontier for a fraction of the cost, then the assumption underpinning the entire American AI economy, that dominance requires enormous, ever-growing spending on compute, and that this justifies half-trillion-dollar infrastructure bets and trillion-dollar valuations, is suddenly in doubt. And it came from China, despite years of American export controls designed precisely to prevent Chinese labs from reaching the frontier, raising the possibility that the controls have failed, or worse, spurred exactly the efficient, home-grown innovation they were meant to stifle. In a fortnight, the story of AI shifted from American inevitability to genuine contest, and the most expensive assumptions in technology were, for the first time, seriously questioned.

WASHINGTON

The blitz accelerates: an agency dismantled, tariffs launched, and a plan to "own" Gaza

The new administration's assault on the shape of the government accelerated. An effort led by Elon Musk moved to dismantle USAID, the agency that administers American foreign aid, freezing its funding and recalling its staff, effectively shuttering overnight a body that had been a pillar of American soft power for six decades. Trump launched his trade war, imposing then partly pausing steep tariffs on Canada and Mexico and levying new ones on China, sending markets lurching. And in a statement that stunned the world even by the standards of the fortnight, Trump declared on February 4 that the United States would "take over" the Gaza Strip, relocate its Palestinian population, and develop it into a "Riviera of the Middle East," a proposal condemned across the region and the world as a fantasy of ethnic cleansing.

IN BRIEF

Disaster over the Potomac; a game decided

The deadliest American aviation disaster in a generation struck on January 29, when an American Airlines regional jet collided with an Army helicopter over the Potomac River on approach to Washington's national airport, killing all 67 people aboard the two aircraft, in congested airspace whose risks had been flagged for years. And the Philadelphia Eagles won the Super Bowl on February 9, thumping the Kansas City Chiefs and denying them a historic third straight title.

The Column

The wall that trained its own defeat

A cheap Chinese model humbled Silicon Valley this fortnight, despite years of American controls built to prevent exactly this. The controls did not merely fail. There is a case that they helped cause the thing they were meant to stop, and the mechanism is an old one in economics.

By Daniel Reeve

A Chinese laboratory released an artificial-intelligence model this fortnight that matched the best American systems at a fraction of the cost, wiping some six hundred billion dollars from the value of American chipmakers in a single day and humbling an industry that had believed its lead unassailable. What makes the episode more than a market story is that it happened despite, and this is the uncomfortable part, in some measure because of, years of American export controls designed to deny China exactly the advanced chips such a model was supposed to require. The controls did not merely fail to prevent the breakthrough. There is a serious case that they helped produce it, and the mechanism is one economists have understood for the better part of a century.

The economist John Hicks, formalising an older intuition, described what is called induced innovation: the principle that scarcity of a particular input does not merely constrain producers but actively induces them to innovate their way around it, to develop precisely the techniques that economise on the thing they have been denied. When a resource becomes expensive or unavailable, the rational response is not to give up but to invent a way of needing less of it, and necessity, in this precise and non-sentimental sense, really is the mother of invention: the constraint reshapes the direction of innovation, pointing it straight at the bottleneck. American controls denied China the most advanced chips, making abundant computation scarce for Chinese labs, and induced-innovation theory predicts exactly what followed: that the denial would push Chinese engineers to innovate on efficiency, to develop methods that achieve more with less silicon, to solve, under the pressure of scarcity, the very problem that American abundance had let American labs ignore. The wall did not stop the thing it was built to stop. It aimed Chinese ingenuity directly at getting around it, and Chinese ingenuity did.

The counterargument deserves real weight, and the controls' defenders make it: that the controls did impose real costs, did slow China down, did buy time, and that without them the Chinese breakthrough might have come sooner and larger; that a policy which merely delays a rival is not a failure just because the rival eventually adapts. This has force. Controls are not useless simply because they are not permanent, and the counterfactual, a China with free access to the best chips, might indeed be worse. But the induced-innovation point survives, because it identifies a cost of the controls that their defenders rarely count: that by making the scarce resource the axis of competition, the controls did not merely delay the rival but redirected its effort toward efficiency, and efficiency, once achieved under pressure, is a permanent capability that does not depend on the chips at all. American abundance had made American labs profligate, solving problems by throwing more computation at them because they could; Chinese scarcity made Chinese labs frugal, and the frugal solution, once found, is the more dangerous one, because it is cheaper, more portable, and no longer constrained by the bottleneck the wall was defending.

And that is the deeper lesson, which reaches past this one episode into the whole logic of denial as strategy. A wall built to deny a rival a resource does not merely hold the rival back; it tells the rival, with perfect precision, exactly where to concentrate its ingenuity, and a determined adversary pointed at a specific bottleneck by the very act of denial will, often enough, innovate straight through it, emerging with a capability it would never have developed had it simply been given what it wanted. The controls made advanced chips the prize and the constraint at once, and in doing so induced the Chinese to build the thing that needs fewer of them, which is a worse outcome for American advantage than if the chips had never been denied. This is not an argument that all controls are futile; it is a caution that denial is a strategy with a boomerang built in, that scarcity induces the innovation that routes around it, and that a wall, at the scale of a determined nation's ingenuity, is not only a barrier but an instruction, a map of exactly where to dig. Silicon Valley built the wall to keep its lead. The wall taught its rival to need less of the thing the wall was defending, and this fortnight the rival showed what it had learned.

Field Notes
A Relentless build, told plainly

Making the systems that will not talk to each other talk to each other

This fortnight's front page is about a model that learned to do more with less. This is about the least glamorous and most valuable work in enterprise software: getting a company's disconnected systems to share the truth, so the business stops arguing with itself. No client is named.

By Alosh Paul

The single most valuable thing I have repeatedly done, across many years and many organisations, is also the one that sounds the most boring when I describe it: I made systems that could not talk to each other talk to each other. A company does not run on one system. It runs on many, a sales system, a finance system, a system that came with an acquisition, a system a department bought without telling anyone, a decades-old system nobody understands but nobody dares replace, and each of them holds a piece of the truth about the business, and none of them agrees with the others, because they were never built to. The customer is one person, but they exist, differently, in eight systems, with eight slightly different addresses and balances and histories, and the business, made of these disconnected pieces, quite literally does not know its own customer, because the knowledge is scattered across systems that do not speak.

Integration is the work of fixing that, and it is far harder than it sounds, because the difficulty is never technical, or not mainly. The technical part, moving data from one system to another, is a solved problem with mature tools. The hard part is that the systems disagree about what things mean. One system thinks a "customer" is an individual; another thinks it is a household; a third thinks it is an account number. One counts revenue when the deal is signed; another when the money arrives. To integrate them, you cannot just move the data; you have to reconcile the meanings, decide, for the whole organisation, what a customer is and when revenue counts, and then make every system defer to that shared truth. That is not an engineering problem. It is a negotiation, conducted in software, between departments that have spent years quietly meaning different things by the same words and never had to reconcile because their systems kept them politely apart.

I raise this in the fortnight of DeepSeek because there is a connection, and it is about where the real value in technology actually lives. Everyone is transfixed by the models, the frontier, the raw capability, and that is real. But most of the value most businesses will ever get from technology is not at the glamorous frontier; it is in the unglamorous middle, in getting their own systems to agree with each other so that the business can see itself clearly and act on the truth rather than on eight contradictory versions of it. A company with a brilliant AI model bolted onto a landscape of systems that disagree about who its customers are has built a genius on top of a swamp. The integration, the patient reconciliation of what the systems mean, is the foundation that makes everything above it worth having, and it is exactly the work that gets deferred, because it is invisible and difficult and nobody gets excited about it, right up until the business discovers it has been making decisions for years on data that never agreed with itself. Make the systems talk. Make them agree what the words mean. It is the least glamorous work in technology, and underneath almost everything, it is the work that actually matters.

The Ledger
AI
DeepSeek's cheap, open model humbled the "spend hundreds of billions" thesis and cost Nvidia a record $589 billion in a day. The claims are disputed, but the question is real: does the frontier require the vast spending the whole American AI economy assumes? And American export controls did not stop it.
Data centers & power
DeepSeek's efficiency, if real, would ease the power question this paper has pressed; if it means the same capability for less compute, the grid gets a reprieve. But history says efficiency gains get eaten by expanded ambition. Watch whether cheaper AI means less power, or just more AI.
Rates
Trump's tariff launch injected fresh inflation risk and market volatility; the Fed holds, watching a trade war whose inflationary effects it cannot easily offset. The path of cuts grows more uncertain.
Real estate
US 30-year mortgage 6.89 percent (February 6). Elevated, range-bound, hostage to an inflation outlook that the tariffs cloud further.
India tech
DeepSeek's frugal achievement resonates with a theme this paper explored through India's Moon landing: that constraint breeds efficiency, and that the assumption only the biggest spenders can reach the frontier may be exactly wrong. India's own AI ambitions, necessarily frugal, should take note.
What we called wrong
Nothing to retract, and one long argument sharpened: this paper questioned the AI-spending thesis as far back as No. 45's bubble column. DeepSeek is the market finally asking, violently, whether the emperor's half-trillion-dollar wardrobe was necessary.
The Back Page

The airspace everyone knew was dangerous

The collision over the Potomac this fortnight, which killed 67 people, happened in an airspace that everyone who flew it knew was dangerous. The approach to Washington's national airport threads commercial jets through a corridor crowded with military helicopters running training routes along the river, in some of the most congested and complex airspace in the country, hard by the restricted zones around the capital, with margins that pilots and controllers had complained about for years. There had been near-misses, warnings, reports, the familiar paper trail of a known risk documented and not resolved, because resolving it, rerouting the helicopters, easing the congestion, spacing the traffic, was expensive and inconvenient and the system had, so far, always just barely worked. And then, on a clear night, it did not, and a regional jet full of people coming home, figure skaters and their families among them, and an Army helicopter on a training run, occupied the same piece of sky at the same instant, and 67 people died. The investigators will spend a year on the proximate causes, the altitudes, the instructions, the exact failures of that exact night. But the deeper cause is the one this paper writes about in every issue, the known risk, flagged and documented and left unaddressed because addressing it was costly and the system kept not-quite-failing, until the night it failed completely. The warnings about that airspace existed. They were, as they always are, filed and not acted on, and 67 people paid the price of the gap between knowing and doing that this paper has spent two years, wearily, chronicling.