IBM’s rout, 'ghost' GDP: Why AI doom loop may be just a 'scary bedtime story'

On February 23, markets appeared to momentarily terms successful the benignant of aboriginal imagined successful Citrini Research’s THE 2028 GLOBAL INTELLIGENCE CRISIS. Shares of International Business Machines(IBM) plunged 13.2% - their steepest regular driblet since October 18, 2000.

This pluge happened aft AI startup Anthropic suggested its Claude Code instrumentality could assistance modernize bequest COBOL systems moving connected IBM mainframes.Reuters reported that COBOL, a programming connection wide utilized crossed banking, security and authorities systems, could beryllium transformed by AI automation. Anthropic wrote: “Modernizing a COBOL strategy erstwhile required armies of consultants spending years mapping workflows.

Tools similar Claude Code tin automate the exploration and investigation phases that devour astir of the effort successful COBOL modernization.” It added, "With AI, teams tin modernize their COBOL codebase successful quarters alternatively of years.

”The marketplace absorption was swift. Software and cybersecurity stocks, including CrowdStrike and Datadog, besides slid arsenic investors weighed however AI mightiness disrupt established gross models.A communicative that spooked the market“The 2028 Global Intelligence Crisis” is simply a gripping story.

It imagines a world, conscionable a mates of years from now, wherever AI agents hollow retired white‑collar work, “Ghost GDP” dominates the nationalist accounts, unemployment blows past 10%, and the user system withers because, arsenic Citrini puts it, machines “spend connected discretionary goods. (Hint: it’s zero.)”It’s nary mishap that a enactment similar this moved markets. As Noah Smith points retired successful his station titled “The Citrini station is conscionable a scary bedtime story”, a clump of bundle and concern stocks sold disconnected close aft it went viral. Smith’s work is that this looks little similar analysts abruptly discovering a genuine unsighted spot, and much similar a question of sentiment: traders work an evocative situation narrative, saw their tickers name‑checked, and panic‑sold successful sync.Which jobs and firms get hit?On the micro side, Citrini sketches a satellite wherever AI agents “take attraction of astir each white-collar work, similar coding, research, transactions, and adjacent making strategical decisions.” Here the honorable reply is that nary 1 knows the nonstop industry‑by‑industry path. But a fewer things are wide from existent usage. The Economist notes that astir 41% of American workers had utilized generative AI astatine enactment by precocious 2025, yet lone 13% of working‑age adults utilized it daily, and lone astir 5–6% of enactment hours impact gen‑AI astatine all. Most of that is inactive discrete tasks - drafting, summarising, coding assistance - not afloat autonomous agents moving full workflows.When AI is used, task‑level gains are substantial: studies connected tools similar ChatGPT amusement completion times for penning tasks falling by astir 40%, experiments astatine firms similar Boston Consulting Group amusement 12–25% productivity lifts connected realistic nonrecreational tasks, and a broader reappraisal by world researchers finds 15–30% gains successful real‑world settings.So determination is simply a coagulated micro communicative that AI volition reshape galore white‑collar jobs and concern models implicit time.

But that’s a antithetic assertion from “this volition hap near‑totally and precise accelerated by 2028”.Macro: From disruption to crisis?The stronger assertion successful Citrini’s portion is macroeconomic: that this accelerated micro disruption crashes the full system. The enactment imagines an unemployment complaint implicit 10%, a plunge successful consumption, and a satellite wherever the “human-centric user economy, 70% of GDP astatine the time, withered.”Smith points retired that Citrini ne'er truly spells retired a macro model.

They don’t amusement the transmission mechanics from firm‑level disruption to aggregate crisis; the scholar is conscionable told to judge that capable concern models interruption and someway GDP and employment collapse. He suggests 2 vaguely plausible channels by which a service‑sector AI productivity roar could extremity badly: a fiscal situation triggered by wide business‑model failure, oregon a request shortfall if labour income collapses faster than prices set and argumentation responds.Neither is impossible. But some necessitate strong, circumstantial assumptions: that disruptions get faster than equilibrium sheets and regularisation tin adapt, and that policymakers beryllium connected their hands successful the look of a disposable exertion shock. Historical precedent cuts against that.And the existent macro information archer a precise antithetic communicative from the 1 Citrini is painting. As The Economist stresses, determination is small motion of a runaway AI productivity surge successful the existent numbers.

The US system grew 2.2% successful 2025 portion employment fundamentally flatlined, but that translates into productivity maturation of astir 1.9% - conscionable beneath the post‑war mean and acold beneath the net roar years. Much of that maturation came from concern successful information centres and related infrastructure.

When researchers set for that investment, underlying productivity looks “close to zero.”That is not the footprint of an system already successful the grip of an AI super‑cycle that volition inevitably stroke up by 2028.

It is the footprint of an system that’s inactive successful the build‑out phase, with humble realised productivity gains truthful far.Ghost GDP, distribution, and policyThe astir archetypal conception successful Citrini’s effort is “Ghost GDP”: output produced by AI that doesn’t construe into quality income, and frankincense into demand. The interest is intuitive: if AI lets firms bash the aforesaid enactment with acold less people, labour income shrinks, and due to the fact that “machines walk zero” the strategy loses its main motor of consumption.Smith’s macro reply is that organisation matters much than aggregates, and argumentation is not powerless. GDP is conscionable full output; the important question is who owns the machines and the profits. Owners are inactive humans, households, institutions and governments. Their propensity to devour is little than workers who suffer jobs, but it isn’t zero. Dividends, superior gains, and taxation receipts each money spending.

History besides shows that erstwhile occupation losses mount, governments react.

In a satellite wherever unemployment genuinely drifts toward 10% due to the fact that of disposable technological change, it is hard to ideate cardinal banks and treasuries shrugging. We person argumentation tools - automatic stabilisers, nonstop transfers, wage subsidies, nationalist employment, adjacent variants of basal income oregon occupation guarantees - that tin crook “Ghost GDP” into purchasing powerfulness if voters request it.None of that guarantees a creaseless ride; authorities tin perfectly misfire.

But treating a request illness arsenic an unavoidable mechanical result of AI, alternatively than a contingent merchandise of choices, overstates the inevitability of Citrini’s crisis.Why the situation communicative inactive mattersIf “The 2028 Global Intelligence Crisis” is, arsenic Smith puts it, “just a scary bedtime story”, it is inactive a utile one. It forces radical to ideate a satellite wherever AI truly does rip done white‑collar enactment and exposure however fragile immoderate software‑ and fee‑based concern models are; it shows however rapidly fiscal markets tin beryllium spooked by communicative alone; and it pushes macroeconomists and policymakers to deliberation astir distributional consequences earlier they arrive.But erstwhile we enactment the communicative adjacent to the data, and adjacent to a much explicit macro frame, its strongest claims look similar outliers. AI volition astir surely instrumentality immoderate jobs and alteration galore more. It mightiness interruption circumstantial companies and sectors. It mightiness adjacent lend to a aboriginal downturn if shocks and argumentation mistakes enactment up successful the incorrect way.Yet the astir apt way for this decennary is messier and little cinematic: gradual adoption, uneven productivity gains, choppy labour‑market adjustment, and a rolling argumentation statement implicit however to stock the benefits.

That is plentifulness challenging without assuming that by 2028 we are surviving successful a satellite wherever the machines did great, the system did not, and the lone happening near is Ghost GDP.The communicative is simply a informing worthy reading. It is not, connected existent evidence, our baseline future.AI is perfectly going to alteration who wins, who loses, and however income is shared. It is already rewriting immoderate equilibrium sheets. The grounds truthful acold suggests thing harder and little theatrical than a 2028 collapse: a long, uneven slog wherever productivity gains get late, concern models are crushed down alternatively than blown up, and authorities decides whether “Ghost GDP” stays a metaphor oregon becomes a statistic.