2026-08-18 Observatory briefing¶

A Second ADNOC Strike in a Week as Koh-e-Mubarak Hardens Into an Evasion Hub¶

The August 14 strike on two ADNOC-linked tankers in the Strait of Hormuz has been absorbed into a familiar rhythm: the UAE names Iran, cites UN Security Council Resolution 2817, and neither UKMTO nor CENTCOM confirms a thing. The vessels were AIS-dark since July, so independent verification is structurally impossible. This is the second such incident in a week, and the response cycle is already standardised.

Koh-e-Mubarak anchorage has hardened into something more consequential than a transit chokepoint. Twenty vessels cluster there, eight OFAC-designated, with one broadcasting a false identity while holding 1.84 million barrels of crude that appears as ballast. The anchorage is not a waypoint but a permanent infrastructure node for the shadow fleet, with bulk carriers rotating in and out between August 11 and 14. The Iranian state's oil exports have been reorganised around these evasion hubs, and the UAE's condemnation of Iranian UAVs sits awkwardly alongside the fact that the evasion economy operates in its own territorial waters.

Kharg Island's terminals sit empty again, the VLCC that loaded between August 12 and 14 already gone. The LPG terminal has been vacant for twenty days. Iranian export capacity is being exercised through dark transits and ship-to-ship transfers rather than visible terminal throughput, which makes the official statistics of Iranian production increasingly fictional.

The Bab el-Mandeb picture sharpens the regional split. Four Pakistan-flagged tankers have now transited the Houthi blockade unimpeded toward Yanbu, while twelve Saudi-flagged vessels reroute via the Cape of Good Hope, adding up to 6,000 nautical miles per voyage. Yanbu's King Fahd port has been fully dark for six weeks, its vessel count of 26 running above the established band. The Houthis are enforcing a selective blockade that distinguishes between state sponsors and adversaries, and the rerouting costs are being borne asymmetrically. Saudi Arabia pays in voyage days and fuel; Pakistan's state tankers sail through. The Gulf's inter-state rivalries are being fought through maritime logistics, with the shadow fleet as the common currency.

Ebola is Back-and the IMF's Relief Fund Is Empty¶

Source: Project Syndicate

The IMF’s Catastrophe Containment and Relief Trust holds $120 million while the Democratic Republic of the Congo alone owes the Fund nearly $300 million in debt service next year. The arithmetic is brutal enough on its own, but the institutional logic behind it is worse. The CCRT was created in 2015 by retrofitting a post-earthquake trust, itself a retrofit of a debt-relief initiative from the 1990s. Each layer of the mechanism was improvised after the disaster it was meant to address had already arrived. The Fund’s firefighting capacity is built entirely from the ashes of previous fires.

Zucker-Marques proposes selling 10% of the IMF’s gold holdings, valued on the books at $45 an ounce while trading near $4,000, to seed a permanent endowment. The numbers work: a $35.8 billion corpus yielding 3% would generate over $1 billion annually, enough to fund the CCRT and the entire concessional lending architecture without further donor supplication. The proposal is technically sound and politically plausible, which is precisely why it will struggle.

The obstacle is not technical but constitutional. An 85% supermajority is required for gold sales, which hands effective veto power to the US Congress. The author frames this as an opportunity for American leadership, but the deeper point is structural. The IMF’s shareholders have designed a system where emergency relief depends on their periodic goodwill, and they have designed it that way on purpose. Donor pledges arrive with political conditions and delays attached. The $800 million raised during COVID arrived within days, but only because the pandemic threatened rich countries too. Ebola in the DRC does not carry the same urgency for Washington.

The gold sale would convert a passive, non-interest-bearing asset into a permanent income stream for the world’s poorest countries. It would also remove the leverage that donor governments currently exercise over the Fund’s emergency response. That is the real reason it will face resistance, dressed up as procedural caution. The IMF’s balance sheet is not the constraint. The distribution of power within it is.

Nine shipbreaking workers killed in Bangladesh¶

Source: Hellenic Shipping News

The Hong Kong Convention was meant to close the gap between the Basel Convention's prohibitions and the reality of global shipbreaking. Instead, it has produced a certification regime that launders the practice. Ferdous Steel's yard was HKC-compliant when nine workers, including a 17-year-old, died from toxic gas exposure inside the MT RASI on 14 August. The convention entered into force in June 2025; by August 2026, Bangladeshi yards had recorded 84 accidents, 15 dead and 81 injured. Certification is not a safeguard against these deaths. It is the administrative mechanism that permits them to continue.

The legal architecture deserves scrutiny. The HKC sets a weaker baseline than Basel, which was designed to stop hazardous waste being dumped where it cannot be managed safely. A ship is waste the moment its owner decides to scrap it, and the RASI was laden with the residues of decades of operation. The convention's weakness is not an oversight but a function of its purpose: to keep the global fleet's obsolete tonnage moving to the cheapest labour and weakest enforcement on earth, while giving owners and flag states a paper trail of compliance.

H-Line Shipping sold the vessel to cash buyer GMS, which beached it at Chattogram. The Platform had warned H-Line in March 2025 that it was preparing to retire the vessel on South Asian shores. The company proceeded anyway. The Bangladeshi regulator had already taken legal action against Ferdous Steel a month before the deaths. None of this stopped the sequence. The division of responsibility between owner, cash buyer and yard is precisely what makes accountability evaporate: each party can claim it merely facilitated a transaction.

The European Commission now considers approving beach-dismantling facilities. The bodies in Chattogram are the cost of treating certification as a substitute for contained, safe recycling infrastructure. The question is whether regulators in Brussels will read the casualty figures as a warning or as a price worth paying.

Zhu Rongji's Legacy of Demographic Decline¶

Source: Project Syndicate

Zhu Rongji’s death has produced the usual hagiography, and Yi Fuxian’s intervention cuts against it by tracing China’s demographic collapse to the very reforms that made the man a hero. The 1994 tax-sharing system and WTO accession are normally treated as pure growth accelerants. Yi argues they operated as a structural vice, squeezing local governments into land-finance dependency and pushing rural labour into coastal factories where the one-child policy bit hardest.

The analytical weight falls on the tax reform. Before 1994, local governments retained a larger share of revenue and had some incentive to keep people on the land. After it, Beijing centralised tax receipts while leaving expenditure responsibilities local, forcing counties and provinces to monetise land sales and urban expansion. That is not a policy failure; it is a fiscal architecture that made population movement a revenue strategy. The demographic outcome was collateral damage, but collateral damage with a logic.

WTO accession accelerated the same process, deepening the export-manufacturing model that absorbed migrant labour and deferred family formation. Yi’s point is that China’s growth and its fertility collapse are not parallel stories. They are the same story told twice, once as an economic miracle and once as a demographic ledger.

The piece stops short of drawing the political conclusion, but it is there for the taking. A ruling party that built its legitimacy on growth now faces a population structure that growth itself produced. The contradiction is not between reform and tradition; it is between the fiscal and industrial forms of the last three decades and the biological reproduction those forms made impossible. For the party, the problem is not that Zhu’s reforms failed. It is that they worked exactly as designed.

Russia warns UK over supplying drones to Ukraine¶

Source: BBC News

The Russian Embassy's statement lands with the rhetorical force of a state that has run out of patience and is now calculating costs. London is named an "accomplice and co-perpetrator" of "bloody crimes," with the explicit warning that deeper involvement means a higher price. The MoD's reply, that Britain stands "shoulder to shoulder" with Ukraine, is the standard vocabulary of a war fought by proxy, where the actual combatants are the Ukrainian military and the Russian economy's industrial base.

The material detail here is the target set. Ukrainian drones have disabled seven of the ten largest Wildberries logistics centres, hitting refineries and warehouses. This is not a campaign of symbolic strikes on military parades; it is a systematic assault on the circulatory system of Russian commerce and fuel supply. The £35bn in damage is a figure that does real analytical work: it represents the cost of defending a war economy against a weapon system that costs a fraction of that to produce and deploy. Russia's air defences destroyed 180 of 620 drones in one night, a ratio that exposes the arithmetic of attrition. Each interception burns a missile that may cost more than the drone it destroys, and the drones keep coming in waves of eight hundred.

The UK's position is that of a state supplying the means of escalation while insisting it is not a party to the conflict. That distinction is becoming untenable. When British-built drones strike Russian territory, the fiction of non-belligerency dissolves into a practical question of how far the supply chain extends before Moscow decides the price of retaliation is worth paying. The warning about consequences is vague, but the trajectory is not.

Vietjet to boost Philippines operations in latest network expansion¶

Source: FlightGlobal

Vietjet’s announcement of three new Vietnam–Philippines routes lands in a regional market already convulsing. AirAsia has just postponed London and Bahrain flights while absorbing a quarterly loss on fuel costs, and Turkish Airlines still finds its capacity capped by the fallout from the Iran conflict. The Vietnamese carrier’s expansion is therefore not a signal of sectoral health but a targeted grab for yield in a fragmented corner of Southeast Asia where legacy flag carriers have retreated and the LCC model remains the only viable form of accumulation.

The Philippines is a peculiar prize. Its domestic market is dominated by Cebu Pacific, but international point-to-point traffic from provincial Vietnamese cities bypasses the Manila hub entirely. Vietjet’s strategy has always been to manufacture demand where infrastructure constraints make full-service economics impossible. By linking secondary cities directly, it converts what would otherwise be connecting passengers into origin-destination traffic, capturing ancillary revenue that hub carriers surrender to airport fees and transfer costs.

What bears watching is the aircraft utilisation maths. Vietjet’s A321 fleet is among the highest-density in the region, and each new route must clear a break-even load factor in the mid-80s. The carrier has historically subsidised growth through sale-and-leaseback gains and related-party transactions, a financial structure that decouples reported profitability from operational reality. Should fuel prices stay elevated, these routes will need to perform quickly or face the same capacity reset AirAsia is now executing.

The inter-imperialist dimension is muted but present. Vietnam’s manufacturing surge has created a business travel corridor with the Philippines that neither country’s flag carrier has seriously contested. Vietjet is filling that vacuum with Chinese-leased aircraft and Vietnamese labour costs, undercutting any regional competitor that tries to follow. For aviation analysts tracking the broader crisis, the relevant question is not whether Vietjet can fill these planes, but whether the underlying trade growth justifies the capacity before the next fuel spike tests the entire LCC edifice.

The Mid-Size Florida Hub Betting $1.5 Billion That Delta Air Lines Will Turn It Into An International Gateway¶

Source: Simple Flying

Tampa International Airport's $1.528 billion Airside D terminal is a wager that Delta's 20-year, six-gate anchor lease will transform a mid-size Florida station into something more than a spoke. The steel is already going up: 6,200 tons, 7,300 pieces, 96,000 bolts, with completion slated for late 2028. Delta's commitment runs to roughly 2049, matching the terminal's useful life, and includes a purpose-built Sky Club, a facility the airline typically reserves for its largest operations.

The airport is approaching its existing 25 million passenger ceiling, having handled about 24 million in 2024. Airside D raises capacity to 35 million by 2037. But the decisive element is the Federal Inspection Services area, which gives TPA the customs processing capacity to receive widebody international arrivals. Delta's year-round Amsterdam service on the A330-900neo is the proof of concept; the FIS facility is the infrastructure that lets the airline expand European flying from Tampa without competing for processing capacity in the existing international facility.

The lease does not make Tampa a hub. Delta is not building a connecting operation comparable to Atlanta's 1,000 daily departures. The distinction between a focus city and a hub matters for how the airline allocates scheduling priority and capital. A 20-year lease with dedicated gates and a lounge signals a different tier of commitment than shared-gate operations on short-term agreements.

The public authority is betting $1.5 billion of borrowed money on Delta's continued appetite for Florida leisure demand from its Northeastern and Midwestern hubs. That demand has been reliable, but a 20-year lease locks both parties into a relationship that outlasts any single route's profitability. The airport gets its growth; Delta gets priority access to a market it believes will sustain premium traffic. Whether the widebody international expansion materialises depends on demand holding up through the lease's full term, a horizon that spans at least two business cycles.

Anthropic's annualized revenue surges to $65B¶

Source: TechCrunch

Anthropic's annualised revenue run rate hit $65 billion at the end of July, up from $47 billion in May and $9 billion at the close of last year. Investors expect the company to finish 2026 somewhere between $100 billion and $120 billion, and the firm is reportedly seeking a public valuation of $2 trillion or more, which would make it the largest market debut on record. OpenAI, by contrast, has merely doubled its revenue to $40 billion over the same period, and its growth has captivated investors far less.

The numbers are worth sitting with, because they describe something stranger than a successful startup. A company that was valued at $965 billion in late May, when it raised a $65 billion round, is now preparing to sell shares at more than double that valuation roughly three months later. The revenue run rate itself is a projection, not a result: it assumes the most recent short period extends unchanged for a full year. That the financial press and the investors feeding this cycle treat the metric as solid ground says more about the demand for a story than about the durability of the underlying business.

What is being priced here is not current earnings but the conviction that corporate AI spending will keep compounding without limit. The capital being poured into model makers is itself the primary driver of their revenue, as the same firms that fund Anthropic's compute also buy its output. This circularity is not an anomaly to be corrected; it is the mechanism. The $65 billion round in May was not a bet on a proven enterprise software company but a down payment on the next round's valuation, and the IPO prospectus now being drafted is the same logic extended to public markets.

The rivalry with OpenAI matters less as a contest of technical capability than as a race to exit before the music stops. Both have filed confidential IPO paperwork, and Anthropic is expected to hit the public markets first, possibly this autumn. The first mover gets to define the category for public investors, and the second will face the unenviable task of justifying a similar valuation against an established comparable. Inter-imperialist rivalry has its corporate analogue here: the two firms are burning through capital at a pace that demands continuous external validation, and the public offering is the final, largest infusion of fictitious capital the market has yet been asked to absorb.

For the rest of the economy, the question is what happens when the projection meets the quarter. If Anthropic's growth decelerates even modestly after listing, the correction will not be contained to its share price. The entire edifice of AI valuations, from Nvidia's multiple to the cloud providers' capital expenditure plans, rests on the assumption that this revenue curve bends only upward. A company whose annualised revenue grew sevenfold in seven months is not a business; it is a speculative instrument wearing a business suit. The IPO will determine whether the public markets are willing to keep playing the game.

Google buys Spirit Airlines data for $10m to train AI models¶

Source: FlightGlobal

Spirit Airlines’ corpse is being picked over for its emails. Google’s $10 million winning bid in the bankruptcy auction buys access to a trove of correspondence and financial records, the raw material for training AI models. The airline collapsed in May 2026, and its assets have been sold off piecemeal, but the final act of liquidation turns out to be a data harvest.

The transaction is a small, clean illustration of how value migrates in the current phase of accumulation. Spirit’s physical assets — aircraft, gates, routes — were presumably auctioned to other carriers, which will absorb the capacity and continue the same competitive scramble. But the data, the residue of the airline’s daily operations, has a different buyer and a different purpose. Google is not interested in Spirit’s business; it is interested in the patterns embedded in that business’s paperwork. The emails and financial records are not being read for insight into a failed low-cost carrier. They are being fed into models that will learn to generate or process similar documents, a process that converts the detritus of one company’s collapse into an input for another company’s speculative valuation.

What makes the deal striking is the price. Ten million dollars is trivial for Google, a rounding error in its cash reserves, yet it represents the entire residual value of a company that once operated hundreds of aircraft. The bankruptcy process has already written off the equity, the bondholders have taken their losses, and now the final scrap of value is being sold at a price that reflects not what the data is worth but what the auction market will bear. The asymmetry is the point: the data’s value to Google lies in its scale and messiness, not its commercial significance to anyone in aviation. Spirit’s collapse was a crisis of overcapacity and fare competition in the US domestic market; its afterlife as a training corpus is a reminder that the same forces that destroyed the airline are now feeding the machinery that will reshape how such documents are produced.