2026-08-27 ATS briefing

In the AI Gold Rush, the Cloud Wins

Ref: A1 Item-ID: article-4e1b8aaa29 Source: Project Syndicate

Amazon, Microsoft and Google have built something OpenAI and Anthropic cannot buy their way out of: the physical substrate on which every AI product must run. Google's negative free cash flow spooked investors in July, but the cloud division's results were extraordinary, and the distinction matters. A company that owns data centres and networking infrastructure can burn cash on speculative AI ventures and still hold the ground beneath the entire sector. Meta and Nvidia, for all their market power, do not.

The cloud oligopoly's position is structural in a way that even Nvidia's chip dominance is not. Nvidia sells shovels, but the cloud providers own the mine, the railway and the toll road. When the AI bubble deflates, as Rikap suggests it might, the speculative layer of model-makers will be winnowed out, but the demand for computing capacity will not simply vanish. Enterprises that rushed into AI will still need somewhere to run whatever models survive, and the three hyperscalers will be there to collect rent on that necessity.

This is overaccumulation with a safety net. The capital sunk into data centres is enormous and, if AI revenues disappoint, much of it will be written down. But the same infrastructure doubles as the general-purpose cloud business that already generates steady profits. The speculative bet is hedged by the boring one. That is why the cloud giants can afford to keep spending through the downturn while their AI rivals cannot.

The realignment this portends is sharp. If the bubble bursts, the model-makers become dependent tenants of the very companies that funded them, and the cloud oligopoly tightens its grip on the next wave of computing. For those watching from the left, the concentration of this infrastructure in three US firms is not a market outcome but a strategic asset, one that will shape not just the AI industry but the balance of power in the digital economy for the next decade.

Export disruptions increasingly hurting tanker demand

Ref: A2 Item-ID: article-71ec87eb7e Source: Hellenic Shipping News

The US-Iran MoU signed on 17 June has done little more than give BIMCO a pair of scenarios to model. With negotiations stalled, the shipping analyst's two outlooks — Hormuz reopening in Q4 2026, or staying closed through 2027 — bracket a market that has already absorbed a 5.7% fall in year-to-date oil and heavy-product exports and an 11.2% drop in clean products. Yet product tanker tonne mile demand has ticked up, because LR2s have grabbed crude and heavy-product volumes that would otherwise have moved on other vessel classes. The disruption is reshuffling the fleet even as it shrinks the total cargo.

The oddity is that freight rates have risen sharply despite weaker cargo demand. Dirty tanker rates spiked after war broke out and stayed high, because stranded ships, delays and reduced fleet productivity tightened effective supply more than the lost cargoes loosened it. War-risk premiums did the rest. Clean rates rose less, reflecting fleet growth and a thinner Persian Gulf dependence. This is the classic wartime market: scarcity of shipping services, not abundance of cargo, sets the price.

The real pressure is building below the waterline. Global oil and product stocks have fallen by over 500 million barrels as releases compensated for lost production, and OECD cover could drop to 70 days by late 2027. Diesel is the tightest. Stock releases are a buffer, not a strategy; once they run down, the adjustment lands on prices, growth and ultimately demand. BIMCO's own logic concedes that a prolonged closure would eventually erode the very rate strength it now enjoys. The market is borrowing against its own future, and the bill comes due whenever the strait reopens and fleet productivity normalises alongside continued fleet growth.

The Economist’s Embarrassing Proxy War Over AI Policy

Ref: A3 Item-ID: article-bc0d130745 Source: Project Syndicate

The Economist’s unnamed economists, interviewed over drinks, dismiss Acemoglu’s Nobel-winning work on institutions and prosperity as glib, and his AI caution as pessimism built on shaky assumptions. The magazine’s resort to anonymous sniping rather than argument suggests it cannot answer him on the merits. Acemoglu’s position is contestable: technological change is a set of choices shaped by power, and AI’s gains will flow to capital unless public policy steers them toward shared productivity. But it is not unserious.

There is a material reason for the bad faith. Acemoglu’s work threatens the convenient fiction that AI’s trajectory is inevitable and the only question is adoption speed. That fiction props up the current investment wave, where enormous sums of fictitious capital chase promises of future productivity that may never materialise. If he is right that near-term gains are modest and distributional effects malign without intervention, much of today’s AI firm valuations rest on sand. The Economist, cheerleading that wave, has a stake in not conceding the point.

The episode says something about mainstream economics itself. When a Nobel laureate meets character assassination instead of rebuttal, orthodoxy feels cornered. Acemoglu is no revolutionary; his prescriptions stay within managed capitalism’s bounds. But the ferocity of the response shows how narrow those bounds have become, and how much intellectual energy is spent defending an investment boom whose own premises remain unexamined.

Don’t Mess With Markets

Ref: A4 Item-ID: article-c9746cef06 Source: Project Syndicate

Bessent’s defence of his market interventions is that prices are not sending the right signals about fundamentals. Roach’s retort is blunt: they are, and the Treasury Secretary knows it. The interesting question is why a man who spent decades running a hedge fund would pretend otherwise. The answer is not incompetence but the structure of the Trump administration itself, where every official statement is calibrated to please the principal rather than describe reality. Bessent is a courtier performing loyalty.

What his intervention actually targets is the cost of US borrowing. A Treasury Secretary trying to talk down the dollar and manage the long end of the yield curve is attempting to suppress that cost while the fiscal position deteriorates. This is the classic move of a government trying to have it both ways: maintain the dollar’s reserve-currency privileges while refusing the discipline those privileges normally impose. The bond market registers the consequences of policy choices, and when a Treasury Secretary calls that registration unfair, he is objecting to the fact that markets eventually price in the gap between political promises and fiscal arithmetic.

Roach’s framing treats market intervention as a fool’s game because markets are too large and too dispersed to be managed. That is true, but the more immediate danger is the attempt itself. Even if it fails, it corrodes the institutional norms that make dollar-denominated assets attractive in the first place. Credibility is a stock of capital, and Bessent is spending it to buy a few months of favourable headlines. The irony is that the intervention is most likely to work when it is least needed, and to fail when it matters most. For those watching from outside the US, the relevant question is what happens when the world’s largest debtor starts treating its own currency as a policy instrument rather than a public good.

India Is Sticking With America—For Now

Ref: A5 Item-ID: article-f915e588c2 Source: Foreign Affairs

Modi's restraint is the puzzle at the centre of this piece, and Raja Mohan's answer is that it is a strategy of damage limitation, not submission. The "America plus" formulation captures it well: New Delhi keeps Washington as its primary partner while thickening ties with the Quad and other U.S. allies, hedging against the volatility Trump represents. The logic is brutally materialist. For all the talk of a multipolar world, India has no alternative patron. China is the threat, not the answer, and Russia cannot balance it. The United States remains the only power that can absorb Indian exports, transfer technology, and provide a counterweight to Beijing. So Modi swallows the tariffs, the snub over Pakistan, the flirtation with Xi, and the collapse of the Quad summits.

The interesting tension here is between the structural basis of the relationship and its political fragility. The Indian elite knows it needs Washington, but the domestic legitimacy of that choice depends on the relationship appearing reciprocal. Trump's coercion has already shifted Indian public opinion, and the opposition has seized on it. The farmers' protests against the trade deal show how quickly a strategic accommodation can be reframed as national humiliation. Modi's political capital is finite, and he is spending it on a partnership that the U.S. president keeps devaluing.

What the article does not quite say is that this is a problem for Washington too. Trump treats alliances as transactional, but the Indian relationship was built on a shared strategic logic that his behaviour is eroding. If Modi falls, the next government will not necessarily be more pro-China, but it will be far less willing to absorb American coercion. The United States is burning the very credibility that makes its alliances worth having. For now, India sticks with America because it has no better option. That is a thin foundation for a partnership, and Trump is doing his best to make it thinner.

The Impossible Middle East

Ref: A6 Item-ID: article-7a09ec7bcf Source: Foreign Affairs

Six months of US-Israeli war against Iran have turned the Gulf states into a firing range, with drones and missiles hitting military bases, desalination plants and oil refineries while shipping through Hormuz idles. The authors, three Bloomberg economists, argue the region faces "no good options," and their framing stays within the horizon of statecraft. The war has damaged infrastructure, but it has also exposed the Gulf's dependence on a security guarantor that just launched two wars against their neighbour, prioritising Israel's desire to debilitate Iran over Arab stability.

The Gulf response is a diversification drive with a difference. Saudi Arabia's mutual defence pact with Pakistan and Turkey, weapons purchases from Europe and South Korea, and continued economic courtship of China all signal a shift in intent. Earlier hedging was coercive, aimed at making Washington act favourably. Now it is about fulfilling Gulf needs directly, which may be enough to erode US influence even without a full break. There is no alternative protector: China will not offer military backing, Russia is consumed by its own war, and Europe cannot coordinate. The Gulf's resentment that China and Russia are helping Iran rebuild its defence industry is a reminder that the multipolar order is not a neutral space for these states.

The sharpest material point is the emerging logic of talking to Tehran. If Washington cannot keep them safe and may even add to their insecurity, then engagement with Iran becomes necessary, however unpalatable. Years of containment have failed, and Iran's direct targeting of the Gulf makes dialogue both harder and more urgent. This is the realignment the war is forcing: not a clean pivot away from America, but a messy, layered set of arrangements where Gulf states seek protection from multiple directions, including from the power currently shooting at them. For the region's ruling classes, survival now means managing contradictory dependencies rather than choosing sides.

Democracies Must Get Serious About Industrial Policy

Ref: A7 Item-ID: article-275953e645 Source: Project Syndicate

Woods’s generational observation lands on a specific failure: Western leaders were trained to referee markets, and that training now reads as abdication. The state that sets rules and hopes for the best is losing to states that pick industries, fund them, and hold the course. Her remedy is not invention but recall. The US built the internet, the semiconductor industry, and postwar aviation through defence procurement and national laboratories, not through spot markets. The current generation inherited the results of that industrial policy and mistook them for the natural output of market freedom. That misreading is now baked into the institutions they run.

The collision is between two time horizons. Markets price the next quarter; industrial policy requires a twenty-year view. Democratic electoral cycles sit closer to the former, which is why Woods stresses discipline alongside institutions and expertise. A state that cannot hold a course through two or three election cycles will fund a battery plant in year one and a hydrogen hub in year four, and end up with neither. Authoritarian competitors do not carry that handicap, and their advantage flows directly from it.

For the left, the uncomfortable fact is that industrial policy carries no inherent politics. It can build green energy or it can build drones. The question is not whether the state should direct the economy but in whose interest and toward what end. Woods argues for competence; the class question is who controls that competence once it exists.