CMR is the leading provider of funding and management support for small to medium-sized businesses and entrepreneurs 

Authored by Madge Waggy via A lot will happen in 2026!,

My grandmother kept a cupboard she called the just-in-case, and for most of my childhood I read it as a harmless eccentricity rather than what it was - a wound made edible. She had been a girl during the Depression, and somewhere in her twenties she had learned to keep tinned beans, sacks of flour, jars sealed with wax and paper, all of it rotated with the humourless discipline of a quartermaster provisioning for a war that had already ended but might, at any hour, resume. Nothing in that cupboard was ever eaten first. It sat at the back of the shelf like a private insurance policy against a world that had once taken everything from her family and could, she believed, do so again at a moment's notice.

For years the cupboard looked like a relic. The country around it grew fat. Food in America became cheap in a way that no other civilisation in history had ever managed, cheap enough that a family could throw away a third of what it bought and call it a lifestyle, cheap enough that the word hunger retreated into the vocabulary of charity drives and faraway places. We built an entire national identity on the assumption that the shelves would always be full, that the price would always be bearable, that the great problem of food would remain a problem of excess rather than of want. To doubt that assumption in the middle of the nineties was to be faintly embarrassing, like a man who keeps a shotgun in a suburb.

Then the erosion began, and it began so quietly that it registered less as a crisis than as a mood. A receipt that used to total forty dollars now totals sixty and buys less. A bag of chips feels lighter in the hand. A cut of beef that was once a weekly habit becomes a monthly occasion, and nobody announces the change because there is nothing to announce - the change simply happens, and the new state of things becomes the baseline against which the next change will be measured. This is the particular cruelty of slow decline: it is always, at any given moment, almost bearable, and so it is never quite urgent enough to resist.

What follows is an attempt to take the temperature of that decline, using what can be measured and admitting plainly where measurement fails. The argument is not that America is about to re-enter 1932, with soup kitchens on every corner and men in cloth caps hiding their faces in a breadline. The argument is subtler and, in some ways, more disturbing - that the country is sliding back toward the conditions that produced the Depression diet, that the slide is visible now to anyone who bothers to look at a scale, a price tag, or an empty shelf, and that the most likely future is not collapse but a long, grinding narrowing in which abundance survives as an appearance while it dies as a reality.

?? What the Register Confesses

Begin where every household begins, at the point of sale, because that is where the first symptoms surface and where the official figures are quietly outstripped by what people actually pay. The Bureau of Labor Statistics reported that grocery prices rose by roughly 2.9 percent over the twelve months to September 2026, with overall food inflation running near 2.5 percent, and a government forecast for the coming year of about three percent - numbers that arrive in the newspapers as reassurance. The reassurance is largely an artefact of averaging. A three-percent basket is composed of items that fell, items that held, and items that detonated, and the shopper does not buy an average; the shopper buys the specific things a household needs, and the things a household needs have a habit of being the things that detonated.

Beef is the clearest case, and it deserves the attention it gets because it is the load-bearing protein of the working table. Ground beef, the cheapest way a family can put red meat on a plate, climbed roughly 18.9 percent year over year, averaging near $6.75 a pound, while a pound of steak approached $12.80 by May 2026 - prices that would have read as a butcher's fantasy a decade earlier. The Department of Agriculture expects beef and veal to rise another five to nine percent through the year, depending on which forecast you trust, and the reasons are structural rather than temporary, which is to say the reasons will not politely go away when the news cycle turns. Coffee, meanwhile, has risen about 55 percent since February 2024, a fact that lands harder when you remember that almost no coffee is grown domestically and the entire supply depends on weather and politics thousands of miles away. Lettuce added 39 percent over a comparable stretch, sirloin steak 21 percent, orange juice 15 percent, and the small consolations - sweets, biscuits, the cheap sugar that a household reaches for when everything savoury has become expensive - are forecast to climb close to seven percent in 2026 alone.

• Ground beef: up roughly 18.9 percent year over year, averaging near $6.75 a pound.
• Coffee: up about 55 percent since February 2024, with a 50 percent tariff on Brazilian beans layered on top.
• Lettuce: up 39 percent; sirloin steak up 21 percent; orange juice up 15 percent.
• Sugar and confectionery: forecast to rise nearly 7 percent in 2026.
• Imported pasta: pushed upward by a tariff that has reached 107 percent.
• Cocoa: more than doubled since 2024 as West African harvests fail.

Underneath the headline items sits the tariff layer, which the shopper rarely sees and invariably pays. A levy on Brazilian coffee, a punitive rate on European pasta, duties scattered across imported cheese, olive oil, seafood and out-of-season produce - each one is small enough to be invisible on its own, and together they form a quiet surcharge on the entire act of eating. Add to this the slower pressures that no policy can reverse quickly: drought driving up the cost of feed, ranchers reducing herds because they cannot afford to keep them, and a rebuilding cycle measured in years rather than months. None of it is a cliff. All of it is a slope, and the slope tilts the same way every month.

???	Item	12-month change	Note
??	Ground beef	˜ +18.9%	averages near $6.75/lb
?	Coffee	˜ +55%	since Feb 2024; plus 50% tariff
??	Lettuce	˜ +39%
??	Sirloin steak	˜ +21%	near $12.80/lb
??	Orange juice	˜ +15%
??	Cocoa	> +100%	since 2024
??	Sugar & sweets	˜ +7%	forecast, 2026
??	Imported pasta	tariff 107%	cost passed to shelf

Eggs deserve a paragraph of their own, because they are the one category that fell and because their fall is instructive. After the grotesque spike that briefly made a carton of eggs a luxury item and a national talking point, prices came down by roughly a quarter from the prior year by mid-2026, and the relief was real. It was also narrow. A family that saved two dollars on eggs did not bank the two dollars; it spent them on beef, which had risen by more than the eggs had fallen. The single visible victory in the grocery aisle was therefore not a victory at all but a transfer, and it flattered the whole basket into looking calmer than it was. This is the rhythm of the present: not a uniform grind but a series of lurches, one category collapsing while another detonates, letting a household breathe in one aisle while it bleeds in the next, and always leaving the shopper to guess which is which.

?? The Vanishing Gram

A price rise announces itself on a tag. The other kind of rise hides inside the package, and it is arguably more insidious because it is unmeasured by any index that watches the sticker rather than the contents. A bag of crisps that weighed 300 grams now weighs 250 and costs the same. A carton of ice cream quietly sheds a scoop. A loaf of bread loses a slice, a bottle of oil loses an ounce, a block of chocolate shrinks along its edges while the wrapper stays the same size. Each of these is a price increase dressed as a continuity, and every one of them is invisible on a receipt, which is precisely why the practice has spread until it is nearly universal.

The mechanism works because the human eye is calibrated to price, not to weight. We remember that a thing cost two dollars; we do not remember that it weighed 300 grams, because we never looked. The manufacturer exploits the gap between those two memories, and does so legally, and the consumer absorbs the loss without ever feeling robbed. Over a year, the accumulated effect of shrinkflation across a full grocery basket can add several percentage points to the true cost of eating - points that never appear in the inflation statistics, because the statistics measure the price of the item and not the amount of it. The number that the government reports is, in this narrow and important sense, a number about labels rather than about food.

What makes the practice feel like a betrayal rather than a mere accounting trick is the way it implicates the household's competence. A careful shopper who compares prices, who notices that brand X is cheaper than brand Y, is still being quietly fleeced by brand X shrinking its box faster than brand Y shrinks its own. The tools of thrift that a Depression-era household wielded with precision - a sharp eye, a good memory, a mental ledger of what things ought to cost - are being systematically defeated by a category of increase that the eye cannot see and the memory cannot hold. To defend yourself you need a scale, and almost nobody owns a scale, and that asymmetry is the whole point.

??? When the Shelf Forgets

Prices tell you what food costs. Absence tells you something colder, which is what food actually is. Anyone who lived through the spring of 2020, when a run on toilet paper briefly made an entire civilisation look deranged, learned a lesson no statistician bothers to print: abundance is a habit rather than a guarantee, and it can be withdrawn from a shelf in a single weekend without any warning that a withdrawal is coming. That lesson was supposed to be forgotten once the crisis passed. Instead it has been quietly rehearsed ever since, in smaller and more local forms, as shortages arrive and depart and arrive again.

Consider the recent litany, each entry trivial on its own and collectively telling. Baby formula vanished for months, exposing the fact that an entire category of infant nutrition depended on a handful of plants. Sriracha disappeared when a chilli harvest failed, and the internet mourned. Cream cheese, canned pet food, popcorn, mustard of all things - each absence carried its own tidy explanation, a shuttered factory here, a bad harvest there, a labour force thinned until the trucks still ran and the shelves still emptied. No single one of these was a crisis. Together they sketched the outline of a system that is less a machine than a nervous organism prone to flinching, and the flinching is becoming more frequent.

A detail from the present deserves more attention than it received, because it marks a shift in where the fragility actually lives. For most of the post-war period, the weak point in the food supply was production - a bad harvest, a disease, a drought. Now the weak point has migrated to labour and logistics, to the warehouses that are half-staffed, the drivers who are aging out, the last mile that grows thinner every year. The consequence is a strange and unsettling paradox: a store can sit full at the distribution centre and bare at eye level, its backroom stocked while its shelves gape, because the people who would have moved the goods from the truck to the aisle are simply not there. The failure is human now rather than agricultural, and a human failure is harder to fix with a policy because it involves bodies rather than acres.

?? The Herd at the Edge

Follow the beef, because beef is the clearest single instrument for reading the whole apparatus, and what it registers is a supply that has been quietly contracting for years while demand has stayed stubborn. The American cattle herd sits at a 75-year low of roughly 86.2 million head, the residue of drought, expensive feed, and the perfectly rational decision by ranchers to sell animals rather than feed them at a loss. A herd, unlike a factory, cannot be switched back on. You cannot simply hire more cows. Rebuilding takes eighteen to thirty months at the very least, and most forecasts push any real relief past 2028, which means the price of a steak this evening is being decided in pastures right now by animals not yet born.

The arithmetic of a shrinking herd is brutal in its simplicity, and it explains why beef has become the emblem of the whole squeeze. When feed costs rise, ranchers cull. When ranchers cull, supply rises briefly and prices dip, which tempts everyone to think the problem is over. Then the culling catches up, supply tightens, and prices climb past where they started, and the ranchers who want to rebuild must do so into a market where the cost of a breeding animal has itself risen. The cycle runs on a clock of years, and no government, however energetic, can compress that clock. Cattle are not software. You cannot ship a patch.

Set beside the historical record, the parallel stops being merely poetic and starts to feel like a warning. In 1930, New York City ran eighty-three breadlines and Philadelphia ran eighty; the mobster Al Capone, of all people, opened a soup kitchen in Chicago and fed thousands because the official machinery was overwhelmed. Peanut butter, at ten cents a pound, became the poor household's protein precisely because meat had priced itself beyond reach, and a generation learned to build meals around legumes and grains not from conviction but from arithmetic. One small fact from the breadline photographs of that era has stayed with me since I first read it: there are almost no women in them. Women, who held roughly a quarter of the workforce and were the first to be laid off, hid their hunger at home while the men queued in public, so that the visible face of the Depression was a man's face and the invisible suffering was a woman's. Something of that shape is already faintly visible now - the roast shrinking, legumes creeping onto plates, the restaurant demoted from routine to occasion - and it is arriving first in the households with the least room to absorb it.

?? The Credibility Gap

An awkward fact hangs over all of this, and it is the gap between what the statistics say and what people feel. More than eighty percent of shoppers report that food prices have risen by more than the official figures suggest, and while it is tempting to dismiss that as ignorance or alarmism, the gap has a rational explanation that deserves a hearing. The statistician measures a fixed basket of goods and weights it according to the average household. The shopper buys the specific basket that their family actually eats, and that basket is not average. It skews toward the staples that have risen fastest, toward the cheap proteins that have detonated, toward the items that a stretched budget cannot substitute away from. The statistician is measuring the temperature of a lake; the swimmer is standing in the one spot that is boiling.

There is a second and less flattering reason for the gap, which is that the official basket adjusts for substitution in a way that flatters the result. When beef becomes too expensive and a family switches to chicken, the statistician records that the family is buying chicken, and chicken is cheaper, so the index improves. The family, however, experiences the switch as a loss - they wanted beef and settled for chicken, and the settling is the whole point. The index measures what they now buy; it does not measure what they have given up. This is why an economy can post a modest inflation number while a population feels that it is eating worse, because it is, and the statistic has no field in which to record the disappointment.

For anyone trying to read the future, the implication is that the official numbers should be treated as a floor rather than a ceiling, and the lived experience of the checkout as the truer signal. When a statistic and a sensation diverge this widely for this long, the statistic is usually the one lagging. The body knows the price of food before the spreadsheet does, and the body has been telling a darker story than the spreadsheet for several years now. It would be prudent to believe the body.

?? The Anatomy of Want

Strip away the sepia romance that clings to the 1930s, and look honestly at what the Depression actually put on the plate. Money collapsed, unemployment reached a quarter of the workforce, and the diet narrowed to whatever could be stretched: beans and cornbread, potato soup, dandelion greens pulled from a yard, organ meats that the butcher had once thrown to the dogs, breads baked without egg or yeast because both had become luxuries. Malnutrition spread alongside the unemployment rolls, and for millions the only square meal of the week came from a soup kitchen or a relief station, eaten standing up and swallowed with a measure of shame. The era did not produce a cuisine of pleasure; it produced a discipline of scarcity, a body of practical knowledge for making the inedible edible and the edible last.

The specific dishes of that period read today like a museum of ingenuity born of desperation. Hoover stew, a thin vegetable and macaroni concoction named after a president blamed for the hunger. Mock apple pie, built from Ritz crackers when apples were unaffordable, and convincing enough to fool a child. Milk toast, cream of wheat, chipped beef on toast - a dish the soldiers nicknamed SOS and the mothers served without apology. Cornmeal mush, fried in the pan the next morning when the mush had set. The empty pockets turned inside out became known as Hoover flags, a joke that carried the weight of a whole generation's humiliation. Every one of those dishes was a small negotiation between what a family wanted and what a family could afford, and the negotiations were won by the second party.

What is returning now is not the recipes but the discipline, and it is returning with a modern accent. The Depression had no freezer aisle, no vacuum sealer, no warehouse club, no discount app; it had a garden, a jar, and a stove that burned whatever was cheap. The modern household inherits the instinct without the toolkit, which is why the substitution shows up first in behaviour rather than in dishes - store brands replacing name brands, bulk buying when a staple is on sale, the meatless dinner that began as a cost decision and quietly settled into habit, the takeaway demoted from routine convenience to occasional treat. Beneath all of it runs the realisation that our great-grandparents reached slowly and we are reaching at speed: that a full shelf is a proposition rather than a fact, and that propositions can be withdrawn.

?? The Shame of the Line

Hunger has an economics and it has a psychology, and the psychology is the part that history books skip. During the Depression, accepting charity carried a stigma that the modern welfare state has largely dissolved; people hid their faces from photographers when they queued, because the breadline was a public confession of failure in a culture that had promised that hard work would always suffice. That shame shaped how a generation ate long after the money returned. Families developed elaborate codes to disguise substitution - the meat stretched with breadcrumbs so no one would notice, the guest served first and best, the mother eating least and claiming she was not hungry. The scarcity was real, but the performance of plenty was a kind of armour.

Something of that performative instinct has returned, though in subtler form. The modern version is less about hiding from a photographer and more about hiding from the group chat - the holiday meal assembled with care so the guests do not see the corners cut, the birthday cake that cost less than it looks, the dinner party where the host spends the week eating plainly so that the evening can look abundant. Scarcity, when it arrives in a wealthy country, does not announce itself with the drama of a breadline; it arrives as a quiet series of choices made at the margins, and the household works hard to keep the margins out of sight.

There is a darker implication buried in this, which is that the full scale of the narrowing may be invisible precisely because people are skilled at disguising it. If everyone is performing plenty, no one can easily tell how many are actually struggling, and the problem stays underdiagnosed by the institutions that rely on visible distress to justify action. The Depression announced itself through breadlines because the shame was eventually too heavy to hide. The present squeeze announces itself through nothing at all - through a lower cut of meat, a cheaper coffee, a smaller bar - and a population trained to perform comfort may hide the strain from itself as efficiently as it hides it from the neighbours.

?? The Substitution Kitchen

Look at what people are actually cooking, rather than what the recipe sites promote, and the substitution is easy to trace. Meat is the first thing to be rationed by price, and so the kitchen turns toward the ingredients that once carried the Depression: dried beans and lentils, which are cheap, filling, and nutritionally excellent; root vegetables, which store well and forgive neglect; eggs, rice, oats and cornmeal; the cheap cuts that reward long cooking; the canned tomato that turns a handful of odds and ends into a sauce. None of this is deprivation in the strict sense. A pot of beans with a smoked bone and a heel of bread is a decent dinner by any standard, and a skilled cook can feed a family well on a budget that a careless one would blow on a single restaurant meal.

The catch is time. Cheap food is almost always slow food, and the household that saves money on ingredients frequently pays for the saving in hours, which is a form of cost that the ledger of a busy family cannot easily absorb. Soaking beans overnight, simmering a cheap cut for three hours, baking bread from scratch - each of these trades the scarcest resource in a modern household, which is time, for the resource that has become tightest in the wallet. This is the cruel trade that the Depression household did not have to make, because its time was less monetised and its rhythms slower; the modern family must find the hours somewhere, usually by sacrificing rest.

The generation now learning these skills is doing so without a teacher, which is its own kind of loss. Depression-era families inherited cooking knowledge along an unbroken chain from a mother or grandmother who had always cooked this way, and the chain carried not only recipes but technique, judgement, the small calibrations that no written recipe captures. Two generations of convenience eating broke that chain, and what is being rebuilt now is being rebuilt from videos and guesswork, imperfectly, by people whose parents never soaked a bean and whose grandparents were too distant to ask. The skills are returning; the continuity is not.

?? Ground Under the Fingernails

The first real defence against all of this is the oldest one, and it costs almost nothing but attention: a patch of ground, a windowsill, a few pots, planted with something that will feed you. A garden converts sunlight and patience into calories that no supermarket can reprice, and it does so at a return that no savings account currently matches. It need not be grand - a strip of beans along a fence, a few tomato plants, a pot of herbs by a kitchen window will outproduce the beginner's expectation and cost less than the same harvest would at the till by the following autumn. The point is not self-sufficiency, which is a fantasy, but resilience, which is achievable.

• Start small and stagger the planting, so that a single failure does not wipe out the season.
• Favour calorie-dense and fast-cycling crops - courgette, beans, leafy greens, potatoes - over decorative ones.
• Learn to save seed, which converts one purchase into an indefinite supply.
• Compost everything, turning kitchen waste into free soil and cutting the bin at the same time.
• Keep a rough ledger of what the garden spares you at the till, because the numbers are more motivating than the theory.
• Join a community plot if you have no yard; it solves the space problem and supplies the thing a solitary gardener most lacks, which is company.

High-yield crops reward effort with a disproportionate harvest, and the surplus, unlike money, does not lose value to inflation; a glut of courgettes in August is a glut of courgettes whenever the prices spike. The garden also delivers a benefit that is harder to quantify and easy to underestimate: it restores a sense of agency. When everything about the food system feels distant and uncontrollable - the weather in Brazil, the tariff in Brussels, the herd in Texas - the garden is small enough to be understood and immediate enough to be influenced, and that psychological return may be worth as much as the produce.

?? The Long Keeping

Preservation is the second defence and the more serious of the two, because a garden gluts and gluts rot, and the difference between a household that eats from its summer all winter and one that does not comes down to the willingness to spend a hot afternoon at a stove. Canning once felt like a grandmother's eccentricity, the sort of thing one admired and never attempted; it is now a quiet form of self-insurance against a shelf that may or may not hold what you need at a price you can bear. The methods differ sharply in difficulty, and the honest ranking below is worth more than any glossy guide, because pairing the wrong method with the wrong food is how a pantry turns poisonous.

???	Method	Best For	Difficulty	Shelf Life
??	Water-bath canning	High-acid: fruit, tomatoes, pickles, sauces	Moderate	12-18 months
??	Freezing	Most vegetables, fruit, meat, leftovers	Easy	3-12 months
??	Dehydrating	Herbs, fruit leather, jerky, some vegetables	Easy	6-12 months
??	Fermenting	Cabbage, cucumbers, carrots, hot sauce	Moderate	2-6 months
??	Salting & curing	Meat, fish, preserved lemons	Harder	Months to years

Fermentation deserves a particular nod, because it is the oldest and cheapest of the methods and the one most likely to survive a power cut. A crock of sauerkraut bubbling on a counter needs no electricity, no freezer space, and no skill beyond patience and salt, and it improves as it sits. The pressure canner, by contrast, is the method that frightens beginners and the one that most rewards the courage to learn it, since it is the only safe way to preserve low-acid vegetables and meat at room temperature. A jar of preserved food is more than a memory of the past; it is a small refusal to let the price of things dictate every meal, and the refusal compounds with every shelf filled.

??? The Short Chain

Beyond the fence and the jar lies the deeper question of where food actually comes from, and the honest answer for most households is a chain so long and so leveraged that its fragility is built in rather than accidental. Shortening that distance is the patient work of a lifetime: buying from farmers directly, joining a cooperative, subscribing to a community-supported agriculture share, treating the local grower less as a quaint alternative than as a load-bearing wall in the structure of a household's food security. The shorter the chain, the fewer the joints at which it can snap, and the more of the money stays in the place where the food is eaten.

Permaculture and urban farming push production back toward the city that consumes it, and seed networks keep genetic variety from collapsing into a handful of vulnerable commercial strains, which is a form of insurance that pays off only in the years when it suddenly matters. Cutting waste is the cheapest yield of all, since the food you bought and threw away was the most expensive food in the house, purchased and then discarded, paid for twice and eaten once. A household that halves its waste has, in effect, given itself a raise without earning a penny more, and that is a return no market can offer.

?? Distant Weather

No account of the present squeeze is complete without the global dimension, because the price of food in a Midwestern town is now set partly by rainfall in Brazil, drought in West Africa, and politics in capitals the shopper will never visit. Coffee is the clearest example: almost none of it is grown domestically, the supply depends on harvests in two countries that have both suffered badly, and a tariff on one of them has been layered on top of an already tight market. Cocoa is worse, with disease and drought in West Africa having more than doubled the price of the raw material in two years, which means the chocolate bar is a commodity long before it is a treat. The shopper experiences all of this as an inexplicable increase; the reality is a chain of cause that runs across an ocean and a hemisphere.

Climate is the slow variable that no policy can negotiate with, and it is reorganising the map of what grows where. Regions that were reliable are becoming erratic; the crop failures that used to happen once a decade are happening more often; and the buffer stocks that once absorbed a bad year have been drawn down across the world. The planet is, in effect, running a thinner pantry than it used to, and a thinner pantry turns every local shock into a global one. The household that plants a garden and fills a jar is, in a small way, insulating itself from weather it will never see.

?? What Could Break

It is worth asking, coldly, what the failure modes actually are, because the answer is more varied and more frightening than the single image of a breadline. The mild scenario is what is already happening: a long, uneven narrowing, punctuated by shortages that stay local and prices that rise faster than wages, producing a population that eats worse without ever quite being hungry. The middling scenario involves a genuine shock - a war that closes a grain corridor, a disease that guts a staple crop, an energy crisis that makes fertiliser unaffordable - and the narrowing accelerates from years into months. The severe scenario, the one that keeps the doomsayers in business, is a combination of shocks that overwhelms the buffer, and in that world the breadline stops being a metaphor and becomes a queue.

What makes the severe scenario more plausible than it used to be is the thinness of the buffers. Just-in-time logistics, which made food cheap, also removed the slack that used to absorb a shock, and a system optimised for efficiency is by definition a system with little room for error. The warehouse that holds two days of stock instead of two weeks is a marvel of cost control until the trucks stop, and then it is two days from empty. A country that has spent forty years removing redundancy in the name of price has, without quite intending it, spent forty years removing the very thing that would protect it in a crisis.

??? No Tidy Ending

Where this all ends cannot be known, and the honest position is to refuse the tidy arc that turns every worry into a lesson and every lesson into a moral. Prices may plateau; some already have, eggs have fallen, and a forecast of three percent for the year sounds almost gentle, until you remember that gentleness is measured against a 2020 baseline that groceries have already left twenty-five to thirty percent behind, and that the cattle herd will not recover before 2028 whatever any politician does this year. The shortages will stay small and local until, one week, they do not. The jars will hold until they are eaten. The herd will be rebuilt by ranchers watching a sky that has stopped being predictable, and the price of a pound of beef will be set by weather in places most shoppers will never name.

What lingers is a question that admits no comfortable answer. Are we living through a temporary squeeze that will ease as the shocks recede, or the early, gentle slope of a longer descent that our grandparents already mapped and we merely forgot? And will the difference even be visible from inside a supermarket, where the shelves still look full enough, most weeks, to let us keep believing that plenty is the default and scarcity the interruption, when the truth may be the other way around? My grandmother's cupboard, which I once found charming and faintly ridiculous, now looks less like an eccentricity and more like a quiet piece of wisdom passed down from a woman who had already learned, at great cost, that the good times are not obliged to last.

Two DeepSeek headlines landed within hours of each other on Monday, and read side by side they tell you pretty much everything about where the US-China AI race is heading: two separate stacks, built by two separate camps, with less and less overlap every month.

Let's start in Hangzhou. According to The Information, DeepSeek has gone well beyond being a Huawei customer and is now effectively a co-designer of Huawei's next-gen silicon. Huawei engineers have reportedly been stationed at DeepSeek's offices for an extended period, folding the AI lab's requests and suggestions directly into chip design decisions. In return, Huawei plans to hand DeepSeek most of the supply of its Ascend 960DT training chip when it launches next year. And on the software side, both companies have been using the open-source programming language TileLang as an alternative to Nvidia's CUDA, with DeepSeek's models now largely handled by TileLang running on Huawei chips.

None of this should come as a surprise. Back in September, when DeepSeek said it aimed to deploy over 160,000 Huawei AI chips, we wrote that DeepSeek's 160,000 Huawei Order Is A Real Threat To Nvidia In 2027. A few weeks later, on Sept 30, DeepSeek open-sourced an Ascend port of TileLang along with Ascend versions of its DeepGEMM and DeepEP libraries, explicitly crediting Huawei and touting a jointly optimized 128-chip Ascend 950 "supernode." Meanwhile Huawei pulled the launch of the Ascend 960DT forward to the first quarter of 2027, three quarters ahead of schedule.

What's new in Monday's report is the depth of the co-design. Nvidia spent two decades building its moat by tuning hardware and CUDA in lockstep with its biggest customers; China is now attempting the same trick with the lab and the chipmaker sitting, quite literally, in the same building. TileLang is still early and nobody serious thinks Ascend matches Blackwell (let alone Rubin) chip-for-chip, but the trajectory matters more than the snapshot. Which brings us back to a question we asked in August, and which nobody in Washington seems to have answered yet:

With open model token prices rising, one can guess what is going on at Anthropic and ChatGPT. What happens when China floods the world with cheap Ascend chips to power the Chinese open models https://t.co/NGn3T78UOo

— zerohedge (@zerohedge) August 13, 2026

Meanwhile, In Washington...

The US response arrived the same morning, and it was, well, a procurement notice. The Department of Homeland Security published an order directing federal civilian agencies to remove DeepSeek products, and any third-party services built on the Chinese model, from their IT systems, and banned agencies from acquiring them going forward.

A few details worth flagging:

• The order (Exclusion and Removal Order #2026-01) was issued under the 2018 Federal Acquisition Supply Chain Security Act and was actually signed back on July 24; it only became public Monday via a Federal Register filing. So agencies have presumably been quietly scrubbing DeepSeek for three months already.
• It covers civilian executive branch agencies such as Treasury, State and Justice. The Pentagon and the intel community are carved out, mostly because the Pentagon already banned DeepSeek and High-Flyer AI via the FY2026 NDAA.
• DHS argues DeepSeek's models pose "multiple threat vectors" and that government use may result in the transfer of sensitive data to the PRC, pointing to China-based servers, PII collection at sign-up and possible user fingerprinting.
• The order doesn't lay out the underlying evidence and stops well short of an Entity List designation, so purely commercial use of DeepSeek is untouched... for now.

Step One Toward Banning All Chinese Open-Source AI?

Probably not step one... more like step four or five, and the important part is hiding in the fine print. The order reaches beyond DeepSeek's own app to third-party products "developed using DeepSeek implementations," which in plain English means any vendor tool built on a fine-tune or distillation of DeepSeek's open weights. And because federal acquisition rules bind prime contractors and subcontractors to these exclusion orders, the provenance of a model's weights has just become a procurement criterion for anyone selling software to Uncle Sam. That is a much bigger deal than banning a chatbot app nobody at the State Department was supposed to be using in the first place.

Look at the sequence so far: NASA, the Navy and roughly 17 states blocked DeepSeek on government devices in early 2025; the FY2026 NDAA booted it from the Pentagon and its contractors; now civilian agencies. And the pipeline behind it is already full:

• The Senate Armed Services Committee's version of the FY2027 defense bill would extend the Pentagon ban beyond DeepSeek to AI from ten other Chinese companies and their derivative models.
• In July, after Moonshot's Kimi K3 dropped (and sparked what we called another "DeepSeek moment" for US AI stocks), Axios reported the administration was reviving plans for de facto bans on Chinese open models, with Kimi and Alibaba's Qwen in the crosshairs via procurement rules, Entity List threats and security advisories. Scott Bessent added a sanctions threat over distillation a day later.
• And in August, The Information reported the US was mulling a self-regulatory organization for AI companies, which we flagged at the time for what it likely was:

There it is: just as @DavidSacks predicted, the push to ban open source AI begins, first with the arrival of SROs *US MULLS ORDER FOR SELF-REGULATORY ORG FOR AI COS: INFORMATION

— zerohedge (@zerohedge) August 27, 2026

If all of this feels familiar, it should: it's the TikTok playbook. TikTok was banned on federal devices in December 2022; the divest-or-ban law followed in April 2024. Government devices first, everyone else later.

That said, an outright ban on Chinese open weights would be a much harder lift than TikTok. The weights are already sitting on Hugging Face and on millions of hard drives, a weights-level ban runs into obvious First Amendment problems, and a lot of US startups (and, reportedly, at least one very large software company in Redmond) have built on Chinese open models precisely because they are cheap and nearly as good. As we laid out in Answering The "Trillion Dollar Question" back in June, the cost gap is enormous. So the more likely path is the de facto one: procurement bans that widen to every Chinese lab, then to contractors, then "advisories" that let enterprise compliance departments do the rest. No law required.

Decoupling, Courtesy Of Both Sides

The irony is that Beijing may get there first. Just last week China's commerce ministry floated export controls that would, for the first time, cover AI model weights and source code above certain performance thresholds (models trained with more than 10²⁶ FLOPs, over 2 trillion parameters, or ranking in the global top 10), even if it promises a fast-track license for open-source releases. Chinese AI commentators immediately took it to mean the next Kimi, GLM and DeepSeek flagships may never be open-sourced at all. So the free lunch of frontier-grade Chinese open weights could end from both directions at once.

And every step Washington takes to wall off DeepSeek pushes it deeper into Huawei's arms, which is exactly what Monday's Information report describes. The US is cutting DeepSeek out of its government while Beijing cuts DeepSeek off from Nvidia, and the result is two fully separate AI stacks: CUDA, Nvidia and closed US frontier models on one side; TileLang, Ascend and (maybe, for a while longer) open Chinese models on the other. Jensen Huang has been warning for years that export controls would hand Huawei the Chinese market on a silver platter. Turns out they also handed it a very motivated design partner.

A CBS News town hall Saturday night saw Vice President JD Vance several times put in a very tough spot. It was a rare moment that a top White House official like him was put directly before voters as they fired off questions.

A one point he was somewhat blindsided by a grieving military husband's necessary and very blunt question over how the admin as well as military leadership has handled the Iran war. His wife, 39-year old Sgt. First Class Nicole Amor was among the first American troops killed in Operation Epic Fury.

The surviving husband, Joey Amor, called out "fundamental failures" in Trump's war on Iran and expressed that he holds decision-makers responsible for his wife's death. WATCH:

The husband of a U.S. soldier killed in Trump's Iran war confronts JD Vance over alleged military leadership failures that led to the deaths of his wife and five other service members. JOEY AMOR: “There's a failure in the leadership.” VANCE: “This is literally the first time I've heard of this.” https://t.co/ofolOyh58a

— MeidasTouch (@MeidasTouch) October 18, 2026

"Though the Iranian drone was the weapon that killed my wife, there were a lot of mistakes and a lot of fundamental failures that led up to that," Joey told Vance during the town hall.

"My wife has told me things before she passed. There are other soldiers that know things that are being quiet about certain aspects of this," he continued.

"I want to know how this administration is going to start holding those leaders accountable for the actions and the deaths of the soldiers that we that we honor and that we love."

Vance's response was widely criticized as appearing robotic and distant - and many also critiqued that the biggest surprise was that he as number two in the administration would express bewilderment and 'surprise' over any of this (now seven months in to a war that initially the White House kept insisting would just take 'days' or 'weeks').

Six American soldiers of the 103rd Sustainment Command were killed that day by Iranian drone strike on Kuwait:

Vance at one point in the response strangely asserted that it was "literally the first time I've heard of this."

"When you talk about failures that were, that we were implicated in.....this literally [is] the first time I’ve heard of this," Vance stated.

"Now, it may be that some of the things that you’re talking about are things where our tactics have already improved," Vance continued in the response. "There were issues with where troops were located that we realized were a huge mistake and that we could do better on, and we have changed that."

He did list out a couple of specifics that "we could do better on" - referencing decision-makers and the administration, and added:

Now, I'm not going to get into the details because, God forbid, I don't want the Iranians to learn something else that they're going to use to target somebody else's wife or somebody else's mother.

There have long been widespread complaints among military analysts and regional watchers that the Pentagon seemed to inadequately account for Iran's response after the opening days and weeks of bombing Iranian cities and installations.

Vance while addressing another question actually seemed to suggest his view that the US attack of Iran was a mistake - though without openly saying so.

"fundamentally, we are where we are..."

JD Vance on Iran: I see what Abdul El-Sayed says about the war in Iran. I see him criticizing decisions that the president has made, the military has made. But fundamentally, we are where we are. https://t.co/bHRUjrePNX

— Clash Report (@clashreport) October 18, 2026

Some reports have said that regionally deployed troops and their family fear the service members in the Middle East are "sitting ducks" and that the White House doesn't seem to have a strategy or end goal.

Back in January, MSCI fired a warning shot at Jakarta. On Friday it fired the real thing.

Indonesian markets extended their selloff on Monday, with stocks deepening their worst rout in almost 30 years and losses spreading to bonds and the rupiah after MSCI warned the country may face a two-step downgrade: straight from Emerging Market to "Standalone" status, skipping Frontier altogether.

The Jakarta Composite (JCI) closed down 3.6% on Monday after crashing 11.6% on Friday, its biggest one-day drop since January 1998. The combined two-session drop of almost 15% is steeper than any two-day stretch during the 1998 Asian financial crisis, and the index hit a six-year low intraday.

Regular readers will recall that we first flagged this exact risk in January, when Indonesian stocks were halted for 30 minutes after crashing on MSCI "investability" concerns. Back then MSCI merely froze index additions and dangled the threat of a downgrade to frontier status if Jakarta did not clean up its free-float data. The market was halted again the next day...

Indonesian stocks were halted for a second day, as investors panicked over potential downgrade risks and transparency concerns flagged by MSCI. The Jakarta Composite Index tumbled as much as 8%, prompting a 30-minute suspension.

— zerohedge (@zerohedge) January 29, 2026

... and the head of the stock exchange did what heads of stock exchanges do when the market they run gets halted two days in a row:

Indonesia Stock Exchange CEO resigns after market rout smart, after two days when the country's entire market was halted

— zerohedge (@zerohedge) January 30, 2026

Nine months, a wave of resignations and a pile of failed reforms later, MSCI has concluded that Jakarta's problems are now too deep for frontier status. The damage since January: the JCI is down about 45% from its January 20 peak and some 42% year-to-date, the steepest fall among major indexes anywhere on the planet, with about $350 billion in market value wiped out since January, according to Reuters (some $60 billion of that on Friday alone). Although to be fair, NVDA loses they by breakfast and gains twice that by dinner. 

"Foreign capital is staging a full-blown exit from Indonesia's equity market, and now the bond market is feeling the heat," SEB's head of Asia strategy Eugenia Victorino told Bloomberg. "A sliding rupiah is eating real yields. Until the currency finds its footing, global money is content to watch this one safely from the sidelines." The locals aren't stepping in either: "For now, even the locals are not buying. There's too much uncertainty," added ACG Management's Hui Ling Loh.

"Frontier Would Not Properly Recognize The Depth Of The Market Problems"

In its announcement, released a month ahead of its scheduled November review, MSCI said that despite the reforms Jakarta announced, "significant concerns remain around market dysfunction, including continued suspected coordinated trading behavior and lack of evidence of meaningful enforcement." And then came the line that broke the market: feedback from investors suggested that a reclassification to Frontier "would not properly recognize the depth of the market problems."

Translation: Indonesia is too broken even for the frontier bucket, which is traditionally the pinnacle of the banana republic yardstick.

MSCI's original gripes, per Reuters, were i) opaque shareholding structures that obscured how many shares actually trade freely, ii) coordinated trading behavior that undermined "proper price formation," and iii) weak enforcement. In plain English, some Indonesian stock prices were suspected of being artificially propped up by a handful of tightly connected holders. Jakarta's response since January included doubling the minimum free float to 15% from 7.5%, tighter disclosure for anyone holding 1% or more, and a list of stocks with "high shareholder concentration." Not enough, apparently.

The timeline: market feedback is due by April 30, 2027, MSCI will announce its decision by June 30, 2027, and implementation could begin as early as the August 2027 quarterly review. Unless Jakarta can show "clear evidence of material improvements," MSCI "expects to proceed."

Jakarta's response came with all the urgency you'd expect from officials who were, by their own admission, caught completely off guard. "We did not expect there would be such announcement (from MSCI). Such an unexpected statement caused the market to react in panic," said IDX chief Jeffrey Hendrik ahead of a virtual meeting with MSCI on Monday, after which he said he "could not disclose details of the meeting." Which is the kind of non-answer that tends to precede actual answers, or even more non-answers. Within hours of Friday's announcement, regulator OJK slapped 2.1 trillion rupiah ($116 million) in fines on entities tied to manipulation in three stocks, and the police were called in to probe "alleged violations or criminal acts in the capital market." Funny how fast enforcement happens when the index provider stops asking nicely.

The exchange also announced it would exclude stocks with highly concentrated ownership from the JCI and, more to the point, cut the daily down limit to 7% from 15%, meaning sell orders below that level are simply rejected. Because nothing says "orderly market" like making it impossible to sell. Per Goldman's Asia desk, Monday's turnover hit IDR10 trillion by midday, with many stocks quickly hitting the new 7% floor and foreigners dumping another ~IDR1 trillion, led by the banks (BBCA and BBRI among the worst performers).

Corporate Indonesia is also doing its part. Three firms linked to billionaire Prajogo Pangestu announced a combined 2.5 trillion rupiah ($138 million) of buybacks, while Sinar Mas Group's PT Dian Swastatika Sentosa proposed a 1 trillion rupiah buyback of its own. Bloomberg notes the buybacks come days after regulators fined some of these same companies over manipulative market practices, which is probably not the kind of "price support" MSCI had in mind.

For what it's worth, even Goldman wasn't expecting this. As the bank's London FX desk noted on Friday morning, there had been discussion that Indonesia could face frontier status at the next review: "We had put this as somewhere around 15 per cent chance of happening." Standalone, it seems, was not even on the menu.

Goldman: "The Burden Of Proof Has Shifted"

In a note published Sunday titled "MSCI Market Classification: Indonesia Downgrade Risk Crystallizes" (available to pro subs), Goldman's index strategist Alvin So wrote that the proposal "materially raises the risk of removal from MSCI composite indexes, although it does not trigger immediate passive selling." Then the punchline:

The burden of proof has shifted: avoiding reclassification will require clear evidence that the announced reforms have been consistently implemented and materially improved market functioning.

And here is the number every EM PM is now staring at:

We estimate that full MSCI removal could trigger ~US$5.6bn of passive outflows, equivalent to 18 days of ADV for MSCI Indonesia IMI constituents, with a further US$2.7bn at risk if FTSE Russell subsequently took similar action. Active funds benchmarked to these indexes hold an additional US$8bn+, although any selling would be discretionary rather than benchmark-driven.

Add it up and you get roughly $16 billion of Indonesian equities held by funds that benchmark to MSCI or FTSE, of which about $8 billion is forced, index-driven selling if both providers pull the plug. FTSE has not gone that far yet (it still classifies Indonesia as Secondary Emerging and hasn't put it on its Watch List, though it too froze positive index changes), so Goldman treats that as a separate downside scenario.

The bank also points out that Standalone is meaningfully worse than Frontier: Frontier status would at least move Indonesia into another composite benchmark with some dedicated demand, while Standalone "would remove it from the major composite indexes without an equivalent destination benchmark." In other words, there is no lifeboat. And roughly $1.4 trillion in funds tracks the MSCI EM index.

Of course, foreigners didn't wait for MSCI to make it official. Per Goldman, foreign investors have already pulled a net $5.1 billion from Indonesian equities in 2026 (on course for the largest annual outflow on record), and the selling accelerated again into Friday's close:

The value of foreign holdings has collapsed from ~$250 billion to ~$130 billion, Indonesia's weight in MSCI EM has shrunk from 1.16% at end-2025 to just 0.39%, and the number of MSCI Standard Index constituents has been cut in half, from 18 to 9, with financials now 72% of what's left. Put differently, by the time MSCI actually hits the button next summer, there may not be all that much left to kick out.

One more thing: active funds are still long. As of September, EM and Asia ex-Japan funds were still ~20bp and ~70bp overweight Indonesia respectively, per Goldman, which means the discretionary selling hasn't really started yet.

"As Much An FX And Rates Story As An Equity One"

The bigger problem for Jakarta is that the selloff refuses to stay inside the equity market. As Goldman's Asia sales trader Peter Sheren wrote on Friday in "Indonesia's Drawdown post the MSCI Reclassification Comments" (available to pro subs), "the MSCI review is increasingly being treated as a country-risk event rather than an equity-specific one." He continued:

Dealers report little willingness to provide USD/IDR liquidity, while persistent USD/IDR demand has pushed spot through 18,000 despite suspected smoothing intervention. The market is now eyeing 18,200 and potentially 18,500 should outflows accelerate. In rates, the key concern is the risk of prolonged portfolio outflows... Overall, the MSCI consultation is becoming as much an FX and rates story as an equity one.

Right on cue, the rupiah slid to 18,030 per dollar on Monday, per Reuters, within spitting distance of the record low of 18,190 hit in June, while the 10Y IndoGB yield jumped to 7.4% from 7.169% on Friday. 

So how much ammo does Bank Indonesia have to defend the currency? Less than the headline suggests. Gross FX reserves stand near $146 billion, but as Goldman's EM strategists Rishi Radia and Lu Sun point out in "Indonesia: We like paying IDR NDS" (also available to pro subs), BI has been quietly leaning on the forwards market:

BI's short forward position has increased to US$33bn, taking gross FX reserves net of the forward book to around US$93bn, a low level last seen in 2014.

That is why Goldman's EM desk thinks this ends up in front-end rates. Radia and Sun note that GIR's combined $10 billion potential outflow (MSCI plus FTSE, in their earlier math) is "substantial relative to Indonesia's BOP," given that combined current and financial account outflows in all of FY25 were only around $6 billion. Their trade: pay 3Y IDR NDS vs receive 3Y SOFR, entry 215bp, target 290bp, stop at 180bp, because "we expect BI to prioritise FX stability, which may put pressure for front-end rates to go higher."

Goldman EM trader Jake Fox was even more blunt about the setup:

The changes in policy making bodies, the constant intervention, regulatory capture of the banking sector into local bonds and the constant tug of war between BI rates, SRBI yields, DNDFs and the FX is unstable... there has not been a compelling argument to me (yet) where local assets give enough of a premium to justify the risk of ownership.

Fox adds that with import coverage down to about three months, Indonesia is moving "towards the convex point," and that since "neither local rates or FX is allowed to adjust... FX implied interest rates have to rise to compensate." In other words, the pressure valve is the forwards market, and it's hissing.

Which brings us to Wednesday's Bank Indonesia meeting. A Reuters poll shows a bare majority (15 of 29) expects BI to hold at 5.75%, but forecasts are flipping fast: Citi now expects a "pre-emptive" 50bp hike and OCBC pulled its December hike forward to this week. Goldman's economists expect two more 25bp hikes, in Q4 2026 and Q1 2027. Higher rates may help the rupiah, but they also hit growth and equity valuations at the worst possible time; so, well, pick your poison. DBS is even more hawkish: a broader correction could push the 10Y toward 7.5% and force BI to hike by another 50 to 100bp, the bank warned.

Bloomberg's Markets Live strategist Mark Cranfield summed it up: "The picture facing investors is a negative feedback loop developing across Indonesian assets, which will likely need local authorities to provide a circuit breaker to create a positive change in sentiment." Judging by the 7% down limit, the authorities took the circuit breaker part a little too literally.

Credit, for now, is the one corner holding up. Goldman's Asia EM credit trader Nitin Pahwa noted that sovereign CDS was only ~3.5bp wider on Friday (at ~97bp) while onshore govies sold off ~30bp, but he wasn't about to fade it: "I just feel the move onshore is ugly, looks like it has legs, and so this small 3-5bp move is just not enough for me to want to fade this as yet." Indonesia's sovereign rating outlooks were already cut to negative by Moody's and Fitch earlier this year, and S&P's sovereign analyst Rain Lin warned on Monday that the "more concerning" risk is "the potential damage to investor confidence and the risk that sentiment weakens more broadly across government bond markets and foreign direct investment."

Welcome To The Club: Argentina, Russia, Venezuela... Indonesia?

Being booted straight from EM to Standalone puts Indonesia in very exclusive company. Goldman's history of MSCI downgrades shows only a handful of EM-to-Standalone moves: Argentina (2019-21), Russia (2022) and Venezuela (2006). The common thread? Capital controls or outright inaccessibility. As Citi analysts noted (via Breakingviews), this would be the first time a country's stock market is expelled primarily for perceived market integrity failures.

Goldman's Sheren drew the obvious comparison with Argentina, whom MSCI moved from EM directly to Standalone in June 2021, but stressed that "the circumstances were dramatically different": Buenos Aires was mired in capital controls and post-restructuring distress, and MSCI was following an S&P downgrade and a FTSE move. The Argentine downgrade "generated an acute, focused reaction in equities and parallel foreign exchange, but a contained reaction in fixed income." Indonesia, which has none of Argentina's capital controls, is already getting the fixed income reaction.

As for what happens next, history offers a modest silver lining: recent MSCI EM downgrades typically underperformed the broader benchmark for two to three months after the initial consultation announcement before stabilizing. Then again, two to three months from now is Christmas, and the decision isn't due until June.

Meanwhile, the one other market MSCI has put on notice, Türkiye (where MSCI froze positive index changes but has not launched a reclassification consultation), shows how much worse Indonesia's case is, as Emre Akcakmak pointed out:

Market performance since the first serious MSCI warning for a downgrade, ???? Indonesia -46% ???? Turkey -21% https://t.co/J3KWDU34aP

— Emre Akcakmak (@akcakmak) October 19, 2026

Cheap Is Not The Same As The Bottom

Inevitably, the "too cheap to ignore" crowd has started circling. And yes, after Friday's crash MSCI Indonesia's 12-month forward P/E dropped to a post-GFC low of 7.3x. But Goldman isn't biting: the multiple "remains above the GFC and 2001 troughs of 5.1x and 3.9x, respectively, suggesting valuations may not yet have reached a floor." The bank remains Underweight, citing "the prolonged index overhang, weak earnings, and macro headwinds": consensus 2026 EPS has been cut 3% YTD, household credit growth has slowed to 5% and retail sales growth is stuck near 1%. (And Goldman's bank analyst Melissa Kuang had already downgraded the Indonesian banks a few weeks ago, cutting BMRI to Sell, which matters when financials are now 72% of the MSCI index.)

To be fair, the bulls have a few talking points: hedge funds "generally continue to view Indonesia as tradable" according to Goldman's desk, domestic institutions have been absorbing foreign selling all year, and Breakingviews argues MSCI "unnecessarily" threw Indonesia into a panic by blindsiding local officials. Maybe. But the real shock came from MSCI looking at nine months of Jakarta's reforms and deciding that frontier status was too generous.

We think the overhang wins. Jakarta now has until April 30 to prove a negative (that coordinated trading has stopped) to an index provider that has already said it "expects to proceed," while the rupiah sits less than 1% from its record low, reserves net of forwards are at 2014 levels, and the central bank must choose between defending the currency and defending growth. Goldman's own history shows these downgrades tend to grind lower for months, and with active funds still overweight, the discretionary selling probably hasn't even started. Which is why for Indonesia, the problem is quickly morphing from "where is the bottom in stocks" into "how much does it cost to defend the rupiah", the same question that marked the start of the last Asian crisis. Let's hope this time it doesn't take a 1998-style answer.

Much more in the full Goldman "MSCI Market Classification: Indonesia Downgrade Risk Crystallizes" and the "Indonesia: We like paying IDR NDS" notes, both available to pro subs.

When a foreign government volunteers to bankroll your release from federal custody, it may be time to reconsider the "no ties to a foreign state" defense.

On Monday, a San Francisco federal magistrate ordered Abdulrahman Alfozan, the 35-year-old Saudi national accused of bugging an executive conference room at OpenAI's headquarters, detained without bail pending trial, saying there is a "significant" risk he would flee. The ruling came after Judge Alex Tse grilled the defense over an offer by the Saudi Consulate in Los Angeles to pay Alfozan's full bail, which CNBC reports that the judge called "problematic".

In other words, the Kingdom offered to post bail for the man accused of bugging the room where OpenAI meets government officials, and the judge decided that was not, in fact, reassuring.

But, there's a lot more going on here...

"Problematic"

Alfozan's lawyer, William Pernik, had asked for release on home detention with GPS monitoring and a co-signed bond, backed by a written undertaking from the Saudi Consulate General. In court, the defense said Alfozan would also put up $1 million in assets on top of the consulate's offer. The defense package also came with dozens of support letters and a commitment from the consulate not to issue him travel documents.

Tse wasn't buying it. The judge noted that OpenAI had already fired Alfozan, so there is no job keeping him in the Bay Area, and ordered him back in court next month.

Courtroom accounts had the judge likening Riyadh's open-ended support to a "blank check", which is roughly the opposite of what you want a federal judge saying about your sureties.

“Even if I set it at $10 million, they are going to pay the entire bond," the judge reportedly said, adding that "there is a real threat here that once he leaves, the case is over."

Prosecutors from the Northern District of California had argued that Alfozan's alleged conduct could pose "a danger to the national security of the United States" and that he enjoyed "considerable support" from the Saudi government.

The defense countered that nationality is not evidence of flight and that the planned trip was for family and immigration reasons. According to CBS, Alfozan has been a US permanent resident since September 2020 and had applied for citizenship,

"Saudi Needs You More Then The U.S.!"

The government's case for detention, filed Sunday, is where things got colorful. Prosecutors said Alfozan was in frequent WhatsApp contact with a member of the Saudi Royal Family throughout 2025 and 2026, as recently as Oct. 5, four days before his arrest. The royal (unnamed in the filing) told him about a new Saudi AI program and repeatedly pressed him on when he'd leave OpenAI. The May 2026 exchange is worth quoting from the filing directly:

"Saudi needs you more then [sic] the U.S.!"

Alfozan's reply: "100"

Translation: agreed. According to CBS, Alfozan added he wouldn't leave until his stock vested, which, in fairness, is the most relatable thing anyone in this story has said.

To be clear (and the defense will make this point at every opportunity), the filing does not allege the royal directed Alfozan to plant the device or steal anything from OpenAI. It also doesn't say how the two knew each other. But it does describe internal searches concerning meetings between OpenAI and Saudi officials: reports say Alfozan accessed an internal folder labeled "Saudi Gov Docs" and searched for a meeting tagged "HUMAIN x MCIT @OpenAI", a reference to the PIF-backed AI champion and the Kingdom's communications and IT ministry. Prosecutors also flagged that further charges could follow.

Regular readers will recall that HUMAIN is the vehicle through which Riyadh has been hoovering up US AI chips (May 14, 2025), with Washington signing off on advanced chip exports (Nov 20, 2025) as part of what we described as American AI imperialism's reliance on the Middle East (Jan 6, 2026). Turns out the reliance may run both ways.

Meanwhile, after his arrest derailed a family trip, prosecutors say the Saudi government and a Saudi military official each arranged flights out of the US for his wife, Lujain Alassaf, and their three-month-old son. That trip, recall, was a one-way booking to Saudi Arabia via Istanbul for Oct. 12, three days after the FBI picked him up, as first disclosed last week. Alfozan is also listed as a venture partner at Riyadh's Raed Ventures, whose partners include a member of the royal family; one former CIA official told CNBC the firm is basically an arm of the state.

Toilet Paper Tradecraft

As a reminder, the underlying case is one we covered when it broke last week. According to the DOJ complaint, OpenAI staff found an encrypted audio recorder wrapped in toilet paper and plugged into a USB port on the back of a TV in an executive conference room on Sept. 30; it had been installed around Sept. 28.

OpenAI then added a camera, which on Oct. 5 caught Alfozan entering and checking behind the TV. The same day, security found a second recorder from the same Chinese manufacturer in a bag he'd left in a printer room - along with his wallet and ID.

The complaint also cited AI prompts about building, obfuscating, and remotely accessing a recorder, and says he used an internal productivity tool to map executive meeting rooms and leadership schedules. The room in question routinely hosted meetings between OpenAI executives and government officials, including on days close to when the device was allegedly planted. OpenAI said it has no indication that model weights, source code or customer data were accessed.

The charge itself - using or trying to use a device to intercept oral communications - carries a maximum of five years and a $250,000 fine (which, as prosecutors seem to appreciate, is a much smaller number than what a one-way ticket to Riyadh is worth).

Same Consulate, Same Offer, Different Ending

Why would a judge balk at a foreign consulate guaranteeing bail? Because this particular consulate has done it before.

In 2016, the Saudi Consulate in Los Angeles posted $100,000 in bail for Abdulrahman Sameer Noorah, a Portland Community College student charged with manslaughter in the hit-and-run death of 15-year-old Fallon Smart. His passport was confiscated and he wore a GPS ankle monitor. Two weeks before trial in June 2017, he cut it off and vanished, turning up in Saudi Arabia a week later, reportedly on a Saudi-issued passport under another name.

The Oregonian later tallied more than two dozen Saudi suspects who fled US charges, in several cases after Riyadh posted bail. And since there is no US-Saudi extradition treaty, Noorah has never been back.

Passport surrender, GPS monitoring, a consular guarantee: the exact package Alfozan's lawyers offered on Monday. We suspect Judge Tse has read The Oregonian.

Bottom Line

The timing is not lost on Washington. Late last week, DNI Jay Clayton - who called the OpenAI incident a "serious situation" - warned that the administration and US companies "need to be wary" of AI espionage, singling out China, Russia, Iran and North Korea.

Trump Admin, US Businesses 'Need To Be Wary' Of AI Espionage, Intelligence Chief Warns https://t.co/7pwqpqNiLd

— zerohedge (@zerohedge) October 19, 2026

Notably absent from that list: the close US ally that just offered to bail out the accused. Officially, nobody has alleged state direction, and Alfozan is presumed innocent. But between the royal WhatsApps, the "Saudi Gov Docs" searches, the military-arranged flights for his family, and a consulate eager to cut a check, prosecutors don't need to prove motive to win a detention hearing - and they didn't.

The bigger question is what this does to the Gulf AI trade. Washington has spent two years waving through chip exports to Riyadh and Abu Dhabi on the premise that 'Gulf capital' is a partner in the US AI buildout, not a collector of its secrets. One alleged toilet-paper bug won't unwind those deals, but it hands every export-control hawk in Congress a brand-new exhibit - or, more likely, gets quietly filed under "isolated incident" the next time a HUMAIN megadeal needs a signature.

Alfozan is due back in court next month. We'll be watching whether "further charges" arrive before he does.