2026-08-21 Observatory briefing

Global grain shipments fall 8% as attacks in the Black Sea escalate

Source: Hellenic Shipping News

The 8% fall in global grain shipments is the arithmetic of a war economy colliding with a harvest calendar. BIMCO's numbers separate the wheat from the chaff: Black Sea ports move only 3% of dry bulk seaborne exports, but 14% of seaborne grain. That asymmetry is the material basis for the disruption's outsized effect. Ukraine's Operation MoLoChKa, launched 6 July, targeted Russian-linked shipping in the Sea of Azov before expanding; Russia's response has been to hammer Ukrainian port infrastructure and merchant vessels, prompting shipowners to suspend calls. The rejection of Ukraine's proposal to halt attacks on shipping closes off the diplomatic exit.

What stands out is the geography of substitution. Russia's redirect towards the Caspian, Baltic and Far East runs into railway bottlenecks already strained by coal exports. Ukraine's Danube and Romanian alternatives cannot absorb the volumes that flowed through its Black Sea ports. The storage deficit of 11 million tonnes by November, 13% of the expected harvest, is the point where military strategy meets agricultural waste. Grain that cannot be moved rots; the Minister of Agrarian Policy's estimate converts a logistics problem into a direct loss of use-value.

The deeper tension is that both belligerents remain dependent on the very trade they are disrupting. Russia needs export revenue to fund the war; Ukraine needs foreign exchange to survive it. Each side's attacks on the other's shipping capacity degrade the common infrastructure of grain circulation, and the world market absorbs the cost through price volatility in Africa, the Middle East and Europe. The Baltic Dry Index's rebound elsewhere does little for importers who now face thinner supplies and higher freight risk premiums. For shipping capital, the war is a rerouting problem; for grain-dependent states, it is a subsistence question. Those two registers have not yet converged into a unified crisis, but the storage deficit is where they threaten to meet.

Why US Bond Markets Are Trembling

Source: Project Syndicate

Jim O'Neill's column performs a familiar act of reassurance: the US bond market's fragility is real, but it is not a sign of American decline. He points to the correlation of weakness across global markets on bad days for Treasuries as evidence that the US retains its anchor role. This framing deserves scrutiny. The fact that other markets fall in sympathy with US bonds does not prove the US is a safe haven; it proves the depth of its integration as the centre of the system. A tremor at the epicentre will always register on the periphery. That is not stability, it is contagion.

The timing is telling. O'Neill notes the bond market's fragility coincides with softening US inflation and weakening high-frequency data. This is the material basis of the market's anxiety. The bond market is not trembling because of a loss of faith in American institutions, but because the Federal Reserve's policy space is narrowing. If inflation cools while growth falters, the Fed faces a choice between cutting rates into a potential recession or holding firm and risking a sharper downturn. The bond market is pricing in the probability of policy error, not geopolitical decline.

The deeper issue is that the US economy's "exceptionalism" has been propped up by enormous fiscal deficits and a dollar that the rest of the world must hold. If the bond market begins to question the sustainability of that debt pile, the anchor role becomes a liability. O'Neill's dismissal of the decline thesis is too quick. The correlation he cites is precisely what a crisis of confidence in the centre would look like: a simultaneous repricing of risk everywhere, with no safe harbour left.

Why Criminals Love the Chinese Economy

Source: Foreign Affairs

Beijing’s enforcement record against its own criminal economy is real enough on paper: 57,000 telecom fraud suspects arrested in joint Myanmar operations since 2023, passports cancelled for offshore gambling operators in the Philippines, several compound bosses executed. Yet the illicit economy has not shrunk. It has grown, and the reason is that the same infrastructure built to manage China’s legal economic contradictions now serves as the circulatory system for its illegal ones.

The $50,000 annual foreign currency quota, designed to keep capital trapped inside the country, created a parallel banking network from the 1990s onwards. Shell companies in Hong Kong, disguised trading firms, Macau junket operators extending credit in Hong Kong dollars and collecting debts in renminbi: all of this was built to move private wealth out of a system that forbade it. That network now launders money for Mexican cartels, converts cryptocurrency for North Korean hackers, and settles payments for sanctioned oil from Iran. The same channels that let a Chinese billionaire buy a Lisbon passport let a scam compound in Myanmar pay its suppliers.

Industrial overcapacity plays its part too. When domestic production outruns domestic demand, the surplus finds foreign buyers regardless of sanctions, and the financial plumbing to move that money already exists. Beijing’s selective enforcement follows the same logic. Cracking down on scam compounds along the Myanmar border serves diplomatic interests; disrupting the oil trade with Iran would not. The state cannot dismantle the illicit economy without dismantling the mechanisms that keep its legal economy liquid, and it will not do that. Washington’s Scam Center Strike Force treats this as a law enforcement problem, but the target is a structural feature of how Chinese capital accumulates and moves.

More than 100 dead after goldmine collapses in Central African Republic

Source: The Guardian

The landslide at Zamboye buried its victims under sand in seconds, but the conditions that put them there were decades in the making. Artisanal goldmining in western CAR persists because the formal economy offers nothing else; the youth council member's phrase about young people coming "to save their families from despair" is the closest the reporting comes to naming the structural compulsion. These are not rogue operations flouting sensible regulation. The criminalisation of small-scale mining does not deter it, it simply strips the work of any safety oversight while leaving the demand for the gold intact.

The death toll of 100, with at least 40 killed in other mining incidents in CAR this year alone, is the routine cost of a sector that global capital extracts from without absorbing any of its liabilities. The gold leaves, the tunnels remain, and the risk stays with the miners. The heavy machinery now being used to retrieve bodies could crush anyone still alive beneath the collapse, a grim metaphor for how the logic of extraction treats the people caught in it: they are recoverable only as resources, not as lives.

The cross-border dimension matters. Cameroonian citizens were among the dead, and the governor of Cameroon's Eastern region is coordinating with CAR authorities. This is not inter-imperialist rivalry; it is the shared periphery absorbing the externalities of a commodity chain whose profits accrue elsewhere. The investigation into the collapse will likely find a landslide, or tunnel failure, and stop there. The deeper cause, that unemployment is the only recruiter these mines answer to, will not be on the charge sheet.

El Niño weather system set to be 'strongest in living memory', warns Met Office

Source: BBC News

The Met Office’s forecast of an El Niño exceeding 3C above average sea-surface temperatures in the central Pacific is not a weather story but a statement about the physical limits of the current economic order. The anomaly, which would be the strongest since records began in 1950 and possibly the most powerful in a millennium, is being supercharged by a reservoir of warm water sitting 100 metres deep at 8C above normal. That heat is the accumulated waste product of two centuries of fossil-fuel combustion, and it is now being redistributed through the atmosphere in a way that no trade negotiation or emissions pledge can capture.

The timing is the sharpest detail. The Met Office expects the temperature anomaly to peak later this year, with 2027 "very likely" to replace 2024 as the hottest year on record. Each successive record is not an anomaly but the new baseline against which agricultural yields, insurance premiums and labour productivity are being priced. The UN Food and Agriculture Organization’s warning that tens of millions face crisis-level acute food insecurity is the clearest indicator of where the burden falls: not on the financial centres that priced climate risk into derivatives, but on the agrarian peripheries whose harvests are the first casualty of suppressed monsoons and shifting rainfall.

The suppression of Atlantic hurricane activity and the below-normal South Asian monsoon are already visible. These are not isolated meteorological events but the material conditions that will shape next year’s food prices, migration pressures and inter-state tensions over water and grain. For the UK, the prospect of a wet, stormy autumn is a minor inconvenience in comparison, yet it signals that even the temperate core of the world-system is no longer insulated from the volatility it helped generate. The question is not whether the system can adapt, but which populations will be asked to absorb the cost of its failure to do so.

AirAsia drops Sydney as 'cost pressures' bite

Source: FlightGlobal

The Sydney suspension is one node in a broader retreat: AirAsia X is cutting its widebody fleet as part of a network "recalibration". The language of deliberate strategy papers over what is, in substance, a defensive contraction. A low-cost long-haul model that depended on filling A330s with discretionary traffic is now hitting the ceiling of what that traffic will bear, and the airline is pulling back to denser, shorter Asian routes where its cost base still functions.

The widebody cut matters more than the single route. AirAsia X's entire proposition rested on using cheap, older aircraft to undercut full-service carriers on trunk routes. That arithmetic collapses when fuel and maintenance costs rise and yield stays flat. Sydney is the first casualty, not the last. Every dropped frequency and parked aircraft represents capital that cannot be redeployed at the same rate of return, and the group's response is to shrink the scale of its own operations rather than confront the structural squeeze on its margins.

What is notable is the timing. This is not a distressed carrier in crisis; it is a profitable group trimming its weakest exposure. That is the more telling signal. When a low-cost carrier with AirAsia's cost discipline decides it cannot make Sydney work, it suggests the problem is not managerial but structural: the long-haul budget segment, as currently configured, may simply not generate enough surplus to sustain the fleet it was built around. The implications for other carriers on the Kangaroo route are uncomfortable. If AirAsia cannot extract value from that corridor, the airlines still flying it are either absorbing losses or subsidising it with more profitable operations elsewhere.

Icelandair Group agrees price for share in former rival Play's Maltese AOC holder

Source: FlightGlobal

Icelandair Group has agreed a purchase price for the Maltese AOC holder that once served its low-cost rival Play, a move the flag-carrier frames as flexibility to "streamline" its operation. The purchase agreement follows Play's collapse into administration in late 2025, when Icelandair stepped in to acquire the carrier's assets and slots at Keflavik. Buying the Maltese entity, rather than merely wet-leasing capacity, gives Icelandair a second operating certificate inside the EU's regulatory perimeter, a hedge against the bilateral constraints that govern Icelandic traffic rights.

The consolidation is unremarkable in itself. Small-market flag carriers absorb failed discounters as a matter of survival, and Iceland's aviation market is too thin to sustain two full-service models. What deserves attention is the timing. Play's failure was not an idiosyncratic business plan gone wrong; it was the predictable outcome of a capacity glut on North Atlantic routes, where US carriers and Gulf giants dumped seats through 2024-25, compressing yields below unit costs for any operator without a fortress hub. Icelandair's purchase price, undisclosed in the report, will reflect the distressed value of an AOC that cost Play's founders millions to establish. The asset is worth more to Icelandair as a defensive option than it ever was to Play as a going concern.

The Maltese certificate also carries a quiet strategic weight. With it, Icelandair can shift aircraft registration and crewing into a jurisdiction with more permissive labour rules than Iceland's unionised environment. The "streamlining" language points toward exactly that: not operational efficiency, but the reorganisation of the workforce along cheaper lines. For a carrier that has repeatedly cycled through collective agreements with Icelandic pilots, the AOC is a lever against domestic labour costs, held in reserve until the next negotiation round.

US approves sale of four KC-46 tankers to Qatar

Source: FlightGlobal

Four aircraft at $4.5 billion works out to over a billion dollars per tanker, a price that reflects not the airframe but the political architecture bolted onto it. The State Department's boilerplate about interoperability and regional stability masks a more specific function: Qatar hosts the forward headquarters of US Central Command and the region's largest American airbase, and its fleet of F-15QAs, Typhoons and Mirages has grown faster than its ability to sustain them in the air. The KC-46 sale closes that gap while binding Doha's logistics to Boeing's sustainment pipeline for decades.

The deal also resurrects a contract that collapsed once before. Qatar had announced intent to buy the Airbus A330 MRTT, but the order lapsed in 2019, and an Airbus official later attributed the failure to the pandemic-era oil price crash. That explanation deserves scrutiny. The 2020 downturn hit Qatari revenues hard, but the country's sovereign wealth fund continued acquiring assets through the same period. More plausibly, the MRTT purchase stalled because Washington made clear which supplier it preferred for a Gulf state hosting its combat airpower. The KC-46 sale is less a response to Qatari need than a reassertion of American primacy over a customer that had briefly flirted with European equipment.

Boeing's position in this arrangement is ambivalent. The KC-46 has been a commercial disaster, with production halted for years over defective fuselage fittings and the company absorbing billions in charges. This sale offers a lifeline for a programme the US Air Force cannot abandon, since it is the backbone of its own refuelling fleet. The Pentagon is effectively subsidising its own supplier by exporting the same troubled aircraft to a regional ally at a premium price. Qatar pays for the privilege of joining a logistics ecosystem dominated by Washington, and Boeing converts a liability into export revenue. The tanker war in the Gulf is not about refuelling; it is about which imperial power gets to keep the region's air forces on its own equipment standard.

Is AI reducing employment for software coders?

Source: FRED Blog

Crane and Soto's counterfactual modelling at the Fed is careful work, but the framing deserves scrutiny. They isolate an occupational shock specific to coders after November 2022 and attribute it to generative AI's public launch. The method is sound as far as it goes: comparing actual employment against a modelled no-shock scenario lets them separate a coder-specific effect from a broader industry downturn. What the counterfactual cannot capture is why capital adopted these tools so rapidly in the first place.

The timing is convenient for a technology story, yet the underlying dynamic is older. Software firms spent the 2010s absorbing cheap credit and expanding headcount on the assumption of perpetual growth. When interest rates rose in 2022, that overaccumulated labour became a cost to shed. AI tools arrived precisely when managers needed a justification for layoffs that would not spook investors or invite scrutiny. The technology is real, but its deployment as a labour-displacing force was conditioned by the balance sheet, not the other way around.

The concentration figures matter here: over 30% of national coder employment sits in computer systems design, and 40% of that industry's workforce are coders. That clustering means a shock to this one occupation ripples through a narrow but significant segment of professional employment. Yet the Fed researchers treat the occupational category as the natural unit of analysis. A Marxist reading would ask instead about the class position: coders are wage labourers whose skills have been commodified, and the current wave of displacement is capital reasserting control over the labour process after a period when skilled workers held unusual bargaining power. The deceleration is not an accident of technology but a restoration of the normal relation between capital and labour, with AI as the instrument rather than the cause.

AI data startup Micro1 reaches $500M gross run rate amid AI training boom

Source: TechCrunch

Mercor hit $2 billion in gross annualised revenue this summer, Handshake crossed $1 billion, and Micro1 has gone from $100 million to $500 million in eight months. The data-labelling sector is no longer a cottage industry feeding scraps to frontier labs; it is a primary site of accumulation in its own right, with researchers now hypothesising that spending on training data could rival spending on compute. That comparison matters. Compute has been the bottleneck narrative for years, the thing that justified trillion-dollar capital raises and energy deals. If data becomes a comparable cost centre, the value concentrated in firms like Micro1 is not a passing arbitrage but a structural shift in where the surplus from AI development lands.

Micro1 retains 60-70% of gross revenue, and its margins on synthetic, off-the-shelf data run as high as 90%. The labour cost that defined the sector's early phase, paying doctors and lawyers by the task, is being engineered out of the equation. Automated video descriptions and reusable datasets mean the same product can be sold to multiple buyers at near-zero marginal cost. This is the familiar trajectory of platform capital: a service that begins as a marketplace for human expertise transforms into a rentier operation extracting value from proprietary data assets. The pivot from AI recruiting to data labelling, and now to synthetic generation, is the same logic applied at each stage, shedding human intermediation wherever possible.

The controversy over selling datasets to Chinese developers is where the contradictions surface. Ansari's protestation that Micro1 does not sell to "foreign adversaries" sits awkwardly beside the fact that his company's entire business model depends on the free circulation of data as a commodity. The same off-the-shelf products that generate 90% margins are, by definition, indifferent to their ultimate consumer. Nationalist posturing cannot resolve the tension between accumulation and geopolitical competition; it can only draw a line that capital will cross when the price is right.

Brazil launches AI supercomputer push while balancing US and Chinese tech

Source: Al Jazeera

Lula’s government has priced its geopolitical balancing act at 2.3bn reais, and the arithmetic is doing the diplomacy. Huawei and iFlytek get 1.3bn reais to build a Rio supercomputer for large language models; a separate 1bn reais tender for a machine in Rio Grande do Norte is widely expected to go to Nvidia. The administration’s stated principle, that it will not depend on a single company, technology or country, is a fair description of the outcome but a generous one of the intent. Splitting contracts between the two blocs does not neutralise dependency; it diversifies it across two poles, each of which retains leverage over the hardware, software and standards on which Brazilian AI will run.

The choice of Rio Grande do Norte for the second machine, on the strength of its energy potential, hints at the real constraint. Supercomputing is an electricity play as much as a chip play, and Brazil’s comparative advantage in renewables is what makes it an attractive site for either Washington or Beijing to anchor. National sovereignty over data is the stated goal, but sovereignty over the means of processing that data remains split between foreign vendors whose home states are actively competing for technological primacy.

The timing matters. China is Brazil’s largest trading partner; the US remains its largest source of foreign direct investment despite new tariffs. The investment is a hedge against both, but a hedge is not a position. Phased disbursements through the FNDCT mean the projects will stretch into 2027, by which point the US-China rivalry over AI infrastructure will have shifted again. Brazil is buying itself options, not independence, and the tender process will reveal which option the market thinks is worth more.