2026-08-12 Observatory briefing¶

Excess Savings Are Driving the New China Shock¶

The framing of China’s export surge as an “absorption problem” is a useful inversion of the usual panic. The first shock was about labour arbitrage; this one is about capital that cannot find a home. Gros’s point is that the collapse in property investment has not been matched by a corresponding fall in saving. The surplus has to go somewhere, and the only outlet large enough is the rest of the world’s markets.

What is left unsaid is that this is not a policy choice but a structural trap. The Chinese state cannot simply stimulate its way out of this without reflating the very property bubble whose deflation created the surplus. Nor can it allow the surplus to be absorbed domestically through wage growth, because that would require a redistribution of income away from capital and toward households — a shift that would undermine the export competitiveness that the surplus itself sustains. The savings-investment gap is thus not a malfunction but the system’s equilibrium.

The political consequence is that the “China shock” narrative now serves a dual purpose. For Western protectionists, it justifies tariffs. For Chinese planners, it justifies continued state direction of investment into ever more advanced sectors, since only technological upgrading can raise the marginal efficiency of capital enough to absorb the glut. The two responses reinforce each other: each round of Western barriers pushes China further up the value chain, which in turn makes the next round of export growth more threatening. The surplus is not a temporary imbalance to be corrected; it is the permanent form of a system that cannot consume what it produces.

Dude, Where’s My Recession?¶

Source: Project Syndicate

The puzzle Eichengreen sets out is real enough: the US economy has absorbed a sequence of shocks — some genuinely unprecedented — and still grown at roughly its estimated potential rate of 2%. But the framing of the question, "where's my recession?", smuggles in an assumption worth interrogating. The expectation that major negative shocks must produce a contraction belongs to a particular theory of how capitalism works, one in which the economy is a self-correcting mechanism that only falters when something external breaks it. What the last several years have shown instead is that the system can absorb enormous damage precisely because the damage is being converted into financial claims rather than resolved.

The 2.1% and 1.5% growth figures are not evidence of health; they are evidence that the burden of adjustment has been displaced. Debt-financed speculation, which Eichengreen names as a risk, is not a side effect of growth but its condition. When productive investment opportunities are scarce relative to accumulated capital, money flows into asset markets, and the resulting price inflation generates the appearance of wealth that sustains consumption. The economy grows on paper while the underlying fragility compounds. Eichengreen's own warning — that resilience should not be confused with the real thing — concedes this, though he stops short of drawing the conclusion his evidence points toward: that the absence of recession is itself the crisis in a different form.

The geopolitical tensions he mentions are not external shocks either. They are the expression of inter-imperialist rivalry intensifying precisely because the major powers are competing for the same shrinking pool of real growth. The US economy's ability to shrug off bad news is, in this light, less a testament to its strength than to the scale of the fictitious capital cushioning it from the consequences of its own contradictions. The recession may not arrive on schedule, but the question was never quite the right one.

The Return of the Power Trust¶

Source: Project Syndicate

The holding-company structure that collapsed in the 1930s was not a failure of regulation but the logical endpoint of financialised ownership meeting a capital-intensive industry. Vaheesan’s warning that the pattern is repeating deserves to be taken literally: the current wave of utility mega-mergers is not a return to the same form, but a second iteration of the same underlying dynamic, where the industry’s physical assets become collateral for financial engineering rather than infrastructure for public need.

The AI boom provides the cover. Data centres’ insatiable demand legitimises a construction spree that would otherwise face scrutiny, allowing utilities to load ratepayers with the cost of new capacity while the debt-fuelled acquisitions that consolidate the market extract value at the top. The contradiction is concrete: households are asked to subsidise the fixed capital that makes AI’s profitability possible, while the ownership structures that capture those subsidies grow more opaque and more leveraged. Ratepayer and shareholder interest have never been identical, but the current model makes them actively antagonistic — every megawatt built for a hyperscaler is a megawatt whose cost is socialised and whose return is privatised.

What is genuinely new is the speed. The 1920s trusts took decades to overreach; the current consolidation is compressing that cycle, with private equity and infrastructure funds treating regulated utilities as yield vehicles. The political question — whether electricity remains a public necessity or becomes a speculative asset class — is being answered by default, merger by merger, while the public debate fixates on data centres’ consumption rather than the ownership that determines who pays.

China’s Hunger Games¶

Source: Foreign Affairs

Beijing’s food strategy is a study in the dual character of state power under conditions of globalised accumulation. The authors correctly identify the central irony: a state that has made “absolute security” in staple grains a matter of regime survival is simultaneously the world’s largest agricultural importer, with a bill above $200 billion. This is not hypocrisy but the logic of a developmental state that has chosen to externalise the ecological and labour costs of feeding 1.4 billion people while retaining domestic control over the political flashpoints of food distribution.

The historical memory is doing real work here. The 1959–61 famine and the 1989 price protests are not abstractions for the CCP; they are the material basis for Xi’s obsession with self-reliance. Yet the response to that memory has been to deepen integration into global markets rather than withdraw from them. The ChemChina-Syngenta acquisition is the clearest expression of this: a $43 billion bet that the path to food security runs through owning the intellectual property of agricultural inputs, not through peasant smallholding. This is the state acting as a capitalist actor, seeking to capture the rents of biotechnological innovation while managing the political risks of domestic scarcity.

The leverage strategy works because China’s import dependence is asymmetrical. Suspending US soybeans or Canadian pork hurts the exporters far more than it hurts Beijing, which can source from Brazil or Argentina. But the fragility is real: 85 percent of soybeans from foreign suppliers is not a position of strength, it is a hostage situation that Beijing has learned to weaponise. The question the authors leave hanging is whether the pivot to GM crops and export capacity will resolve the underlying contradiction or merely shift it—turning China from a vulnerable importer into a competitor in the very markets it once dominated as a buyer.

Zelenskyy says Ukraine has sent proposals to US to end war with Russia¶

Source: Al Jazeera

The substance of Zelenskyy’s announcement is less a peace plan than a plea for a shift in the balance of coercion. Kyiv’s proposals are aimed at Washington, not Moscow; the request for air defences and pressure on the Kremlin is an admission that Ukraine’s leverage is exhausted and that the only variable left in the war is the intensity of US involvement. That the proposals were handed to the Americans rather than tabled at a negotiating table confirms the war’s trajectory: it is no longer a bilateral conflict but a managed proxy arrangement whose tempo is set by the patron.

The timing is instructive. Talks have stalled since the US-Israel war on Iran, a reminder that Washington’s strategic bandwidth is finite and that Ukraine’s fate is subordinate to broader inter-imperialist priorities. The release of Robert Gilman, framed as a humanitarian gesture, is the kind of transactional signal that keeps the diplomatic channel open without conceding anything material. Russia, for its part, continues to demand the cession of four regions and a settlement on its own terms — a position that suggests Moscow believes time is on its side.

Zelenskyy’s warning about a post-election mobilisation is the most revealing element. If accurate, it points to a Russian state preparing for a war of attrition it intends to win through sheer demographic and industrial weight. The "pseudo-election" framing is telling: a ruling class that must manufacture consent for a war it cannot afford to lose, while the Ukrainian state, stripped of its own productive base, increasingly depends on Western air defences to protect what remains. Neither side is negotiating in good faith; both are calculating how much more destruction the other can absorb.

Thai Airways second-quarter profit shrinks on fuel cost pressures¶

Source: FlightGlobal

The headline frames fuel as the culprit, but the underlying mechanics are worth closer attention. Thai Airways cut capacity in the April-June quarter and still grew passenger revenue year on year — a combination that points to pricing power rather than volume. That is the classic position of a carrier operating in a constrained market: fewer seats, higher yields, and costs that rise faster than the ability to pass them through.

Fuel is the obvious pressure point, but it is not exogenous weather. Jet fuel prices track the crude complex, and crude tracks the expectations embedded in financial markets as much as physical supply. For a flag carrier still emerging from its restructuring, the exposure is doubly acute: the balance sheet is thinner, hedging strategies are more conservative, and the margin for error in fare setting is narrower. The profit is real but fragile — a function of disciplined capacity rather than structural health.

What is not in the text is the competitive context. Thai sits in a region where low-cost carriers have been aggressively adding seats on the same trunk routes. If Thai is holding yields while shrinking capacity, it suggests either a favourable demand environment or a strategic retreat from the price-sensitive end of the market. The latter would be rational for a full-service carrier, but it also cedes ground to competitors who can undercut on cost. That is the quiet contradiction: the same fuel prices that squeeze Thai's margins are the ones that make its low-cost rivals relatively more resilient, because their cost base is lower to begin with.

The second-quarter result is a snapshot of a carrier managing decline carefully rather than growing confidently. Whether that is a prelude to a stronger position or a slow slide into irrelevance depends on factors the earnings release will not disclose — fleet plans, route rights, and the state's appetite for further capital injections.

Single-Pilot Airliners Might Be Closer Than You Think: Here's Why That's Controversial¶

Source: Simple Flying

The push for single-pilot airliners is a straightforward labour-cost story dressed in technological clothing. Airlines face a ceiling on fuel efficiency gains, but crew salaries have no such ceiling — they are recurring, unionised, and resistant to the productivity improvements that automation has delivered elsewhere in the operation. Removing one pilot from a long-haul roster is not a marginal saving; it is a structural reduction in the highest variable cost an airline controls.

The industry's framing of "extended Minimum Crew Operations" as a modest step — one pilot rests while the other monitors — obscures the direction of travel. The EASA's own language gives it away: the pilot becomes a "systems administrator rather than a physical pilot." That is not a description of a job evolving; it is a description of a job being prepared for elimination. The rest period is the thin end of the wedge, a trial run for the single-pilot operations that would follow once the public and regulators have been acclimatised.

The February 2026 US law mandating two rested pilots is a rare instance of the state intervening against capital's immediate interests in commercial aviation. But it would be a mistake to read this as a principled defence of working conditions. The law reflects a risk calculus, not solidarity: a single pilot incapacitation event — 287 recorded in European monitoring between 2019 and 2024, roughly one every six days — could produce a mass-casualty disaster that would trigger regulatory chaos and liability costs far exceeding the wage savings. The state is protecting the industry from its own cost-cutting logic.

Cathay Pacific's Project Connect, which aimed to cut A350 crews from four to two, shows the capital already committed to this trajectory. The US law delays the plan; it does not bury it. The technology is developed, the procedures are drafted, and the cost pressure remains. Regulatory barriers of this kind tend to hold only until the next accident-free year makes them look like obstructionism, or until a competitor jurisdiction — likely in Asia — moves first and forces the rest to follow.

The AI Growth Paradox¶

Source: Foreign Affairs

The debt question has been reframed as a technological one, and that is precisely where the danger lies. Rogoff’s argument is not that AI-driven growth is impossible, but that the political arithmetic of the US state cannot absorb the good news. The federal government adds $2 trillion a year to its obligations; the ten-year inflation-indexed Treasury yield has climbed from near zero to over 2.4 percent. Should AI productivity gains materialise, the bonanza will be spent before it arrives — tax cuts and spending hikes will get out in front of the boom, as they always do when the state anticipates a windfall. The contradiction is not between growth and debt, but between the temporal logic of capital accumulation and the electoral cycle of the bourgeois state.

The deeper issue is the capital squeeze. The AI build-out is not a productivity story in the first instance; it is an investment story. Massive data-centre construction and chip fabrication absorb capital that would otherwise circulate elsewhere, pushing up real interest rates across the economy. This is overaccumulation in its most literal form: too much capital chasing a narrow set of speculative infrastructure, while the state’s financing costs rise in tandem. Rogoff’s historical point about financial repression is worth holding onto — post-war debt reduction worked because regulated markets forced savers to subsidise the state. That mechanism is gone, and the saver class that absorbed it has been replaced by institutional investors who demand a market return.

The "this time is different" orthodoxy of the 2010s — secular stagnation, permanently low rates — was always a projection of a specific balance of class forces onto eternity. Rates reverted to the mean because the conditions that suppressed them were political, not natural. The same error now repeats in reverse: the AI boom is treated as a permanent productivity shock rather than a concentrated burst of fixed-capital investment with an uncertain payoff. If the productivity gains disappoint, the state is left with higher debt service and a capital stock that cannot reproduce its own financing costs. For global supply chains, the relevant signal is not the productivity headline but the cost of capital embedded in every long-term infrastructure decision. The crash, if it comes, will not be a technology failure. It will be a fiscal one.

Morocco’s Bet on Targeted AI¶

Source: Project Syndicate

Morocco’s pitch is a quiet rebuke to the prevailing logic of the AI race, yet it is a rebuke that operates entirely within the market’s terms. The author’s framing of “sovereignty” — over data, talent, and use-cases — is a strategic retreat from the frontier, not a challenge to it. By ceding the hardware race to the US and China, Rabat avoids the ruinous capital expenditure of building compute clusters that would be obsolete within a cycle. Instead, it positions itself to capture value at the point of application, where the real surplus in a mature AI market will be extracted.

This is a sound calculation for a semi-peripheral state. The US and China are engaged in an inter-imperialist contest over foundational models, a contest that demands the kind of overaccumulation of capital that only states with vast domestic markets or deep financialisation can sustain. Morocco cannot play that game. Its bet is that usefulness — the adaptation of generic models to local logistics, agriculture, and public administration — will generate a more durable form of economic rent than the speculative valuation of frontier labs.

But the strategy’s internal tension is evident. Sovereignty over data is meaningless if the foundational models remain foreign-owned. Morocco will be renting intelligence from the very powers it seeks to insulate itself from, paying in the currency of its own economic activity. The author’s insistence that the country can “embrace innovation without ceding its identity” papers over this dependency. The real question is whether targeted AI can generate enough domestic surplus to fund the next stage of development, or whether it will simply integrate Morocco more deeply into the global division of labour as a service provider for foreign capital. For now, the bet is plausible; the exit strategy is not.

The Token Economy Requires Collective Data Governance¶

Source: Project Syndicate

The argument for data cooperatives rests on a curious symmetry: the same people whose unpaid output constitutes the training corpus are invited to become shareholders in their own expropriation. Benhamou is right that the current arrangement is lopsided — trillions of tokens extracted from collective labour, with the surplus accruing to a handful of firms whose models could not exist without that input. But the cooperative remedy, as sketched here, treats the symptom while leaving the underlying property relation untouched.

A data cooperative would negotiate access to content, presumably extracting licensing fees from AI developers. Yet the value of any individual token is negligible; its worth emerges only in aggregate, at the scale of the entire corpus. This is precisely why the platforms that assemble and process that corpus capture the surplus — they own the means of computation, not the raw material. A cooperative that merely sells access to its members' data is in the position of a peasant collective bargaining with a grain merchant: it may win a marginally better price, but it does not thereby gain control over milling, transport, or the market that sets the terms.

The deeper issue is that the token economy does not simply monetise knowledge; it converts the entirety of human cultural production into a feedstock whose value is realised only through infrastructure that no cooperative could realistically own. Without confronting the concentration of compute and model distribution, collective data governance risks becoming a legitimating device — a way to secure consent for extraction while redistributing a small fraction of the rent. The question is not whether users should organise, but whether organising at the level of data access can alter the balance of power when the decisive assets lie elsewhere.