The State of the World.

1. Russia is declining faster than the headlines suggest, and Beijing has noticed

Ukraine's drone campaign against Russian refining infrastructure has done what more than four years of sanctions could not: made the war's costs visible to ordinary Russians. Between a fifth and a third of Russia's refining capacity is offline, depending on the estimate. Formal fuel rationing covers 53 of Russia's 83 regions and shortages are reported in nearly all of them; Crimea has declared a state of emergency and banned public fuel sales. The state is importing gasoline and has turned to Kazakhstan to cover the shortfall. The government has cut its 2026 growth forecast to around 0.4 per cent, the budget deficit is running well ahead of plan, and polling shows the most pessimistic economic mood in two decades, with trust in the president at 69 per cent, its lowest since the full-scale invasion began.

Andrew Latham calls this a repair war, and it's the right frame. Strikes land faster than refineries can be patched up, and the specialised kit needed to fix them is exactly what sanctions make hard to buy. The question stops being how much damage Russia can absorb and becomes whether it can rebuild faster than Ukraine can destroy. Right now it can't.

So what does a struggling nuclear power do next? Our view: what Russian states have always done under supply stress. Protect the centre. Fuel goes to Moscow and St Petersburg, to the army, to the industries the war runs on, and the shortfall gets pushed out to the regions least able to make themselves heard. It's the rational short-term move and it will probably work. Empty pumps in the capital are a political event; empty pumps two thousand kilometres away are an administrative one.

But the cost compounds. Muscovites already live in something close to a separate country, insulated by wealth and proximity to power from how the regions actually live. Ration fuel by geography and you turn a resented asymmetry into official policy. It's worth remembering that the Soviet Union didn't collapse from the centre; it came apart at the edges, where the bargain between state and governed had already broken. Today's regions don't have the exit ramps the Soviet republics had, so we're not predicting secession. We're predicting something slower: erosion of the deal that holds the whole arrangement together, acquiescence in exchange for stability and cheap fuel. Putin consolidates now and weakens the state later. He'll take that trade, because the alternative is worse for him today.

Then there's China. Watch what Beijing did this quarter rather than what it said: it again withheld final agreement on the Power of Siberia 2 pipeline at the May summit, holding out for better prices and refusing to deepen its own dependence. Beijing isn't distancing itself from Moscow. It's repricing Moscow. Russia is turning into a supplier of discounted raw materials and a captive market for Chinese goods, which is not an alliance of equals; it's a client relationship forming in real time, and parts of the Russian elite know it.

The catch, from Beijing's side: it wants the client, not the collapse. A cheap, compliant, resource-rich neighbour is useful. An unstable one on a shared border, with the energy supply interrupted and nobody quite sure who controls the nuclear arsenal, is not. Our house view is that if Putin's position starts to falter, Beijing will de-personalise the relationship, building ties designed to outlast any individual leader rather than betting on a successor. The tell will be engagement that runs through ministries and commercial contracts instead of summits.

Why should Australia care? Not because of Russia. Because of what the episode shows about China: it extracts advantage from a partner's weakness patiently, commercially and without sentiment, and it hedges against that partner's instability while professing friendship the whole time. Worth remembering in our own dealings.

2. The Gulf: a ceasefire in name only, and a floor under oil

The war that began with the American and Israeli strikes on Iran in late February has settled into a pattern: ceasefire, memorandum, violation, retaliation, talks resume. As this briefing goes out, the latest round is still unfolding. Iranian attacks on three commercial vessels in the Strait of Hormuz drew more than 80 US strikes in response, including on more than 60 Revolutionary Guard small boats, along with the reimposition of US oil sanctions; Iran answered overnight with missile and drone launches against US military sites in Kuwait and Bahrain. On our assessment it is the most intensive exchange since the June memorandum was signed, and each side accuses the other of breaching an agreement that neither has abandoned.

Our read: the pattern is the equilibrium. It isn't a phase on the way to peace, and it isn't the slide into a wider war either. Washington wants the strait open and Iran contained without committing ground forces. Tehran's new leadership needs to look defiant without inviting the strike that ends the regime. Both sides can live with a cycle of limited strikes more or less indefinitely.

We expect it to persist, though not statically. Each round narrows what either side will try next, and over time the pattern should grind toward a rough settling point: fewer novel escalations, more ritual ones, an accommodation neither government will ever name. Don't price it as peace. This is a conflict becoming a condition, and the practical upshot for markets is that Gulf risk premium is now built into the oil price. Waiting for it to pass is not a strategy.

What would break the equilibrium? For the record, our nomination is a tolled strait. If Tehran ever formalises passage fees for Hormuz, harassment becomes administration: a claim of sovereignty over a global chokepoint that Washington could not ignore, and that others would study closely. Read the strait item in What we are watching with that in mind.

The Strait of Hormuz disruption is the single most important economic fact of 2026 for energy-importing Asia. It's why Japanese and Australian leaders signed energy security agreements in May, why fuel is driving inflation across the region, and why “economic security” has migrated from think-tank papers into treaty language.

3. Australia: not in recession, but squeezed from both ends

Is Australia in recession? Not on the data. Output is still growing at around 2.5 per cent annually, though the Reserve Bank expects that to slow to 1.3 per cent through 2026, from 2.6 per cent over 2025. The squeeze in two numbers: 2.5 per cent behind us, 1.3 per cent ahead. The March quarter, at 0.3 per cent, was the softest since mid-2024, and unemployment, 4.4 per cent in May, has drifted about a third of a percentage point higher over the year. In some ways this is harder to manage than a clean downturn. It's a supply-shock squeeze.

The Iran war pushed fuel and raw material costs sharply higher. Headline inflation is forecast by the Reserve Bank to peak near 5 per cent around mid-year, though the monthly indicator has eased in recent prints and the peak may prove lower. The Reserve Bank, determined not to repeat 2022, when inflation expectations nearly slipped away from it, has raised the cash rate three times this year, to 4.35 per cent, before pausing in June with pointedly hawkish language. At least one major bank still expects more hikes before year-end. So households are wearing higher petrol prices and higher mortgage rates at the same time, for reasons that have nothing to do with anything they did. Sentiment has deteriorated accordingly. The recession risk is real. But if it comes, it will be a recession made in Canberra and Martin Place, in response to a shock made in the Gulf.

And then there's China. Growth is running in the high-4s and should slow toward the mid-4s over the year, but the composition matters more than the headline. The property downturn is still going, with development investment down 11.2 per cent in the first quarter. Consumption is weak. Beijing is choosing supply-side, technology-led growth over broad stimulus (a modest consumption package aside). A China that grows by making chips rather than building apartment towers needs less of what Australia has traditionally sold it.

What this means in practice:

For consumers, the squeeze is the story. Cost of living stopped being a political slogan some time ago; it's a balance-sheet fact for a large share of households, and it's arriving alongside a housing market that is slowing and, in places, going backwards. The exposed cohort is anyone who borrowed at the bottom of the rate cycle. They're exposed twice over: to the mortgage, and to the price of the asset securing it.

For businesses, the response is already visible in behaviour. Firms are shifting resources into lower-price subsidiaries and value ranges to follow the consumer down. Hiring has cooled hard: employment roughly flat over the past two months, job ads down for the third month this year, applications per role surging, and redundancy announcements a steady drumbeat across the corporate sector. That's what firms do when they think the squeeze will last. Energy costs and freight uncertainty are the near-term planning variables, and the cost of capital will stay higher for longer than the 2025 consensus assumed. The cheap-money assumption is the single most dangerous line item in any Australian plan written before 2026; businesses whose models depend on it should stress-test it now.

For exporters, the iron ore and construction-linked channel to China is softening structurally, even as education, services and critical minerals hold up or grow. Diversification stopped being a slogan this year; Japan and India are where the policy energy and the money are actually going.

For investors, the AI boom is still the big offsetting tailwind to everything above, here as everywhere. It's also the question we keep coming back to without being able to answer: supercycle or bubble? We don't claim to know. What we'd point out is that the two risks in this briefing are correlated in a way that's easy to miss. If capital rotates away from an economy it reads as stagnating, and the global equity story carrying portfolios turns out to have been priced for perfection, Australian investors face a domestic downturn and an offshore correction in the same moment, with over-leveraged housing underneath both. That isn't a forecast. It's the scenario worth having an answer to. Your China exposure and your AI exposure are not two risks; on a bad day they are one.

4. Japan: the ally question, answered quickly and quietly

If the quarter had a winner in Australian strategic policy, it was Tokyo. In the space of three months: contracts signed on 18 April for the first three of eleven Mogami-class frigates, Japan's largest defence export contract to date; Prime Minister Takaichi's visit to Canberra, which produced a joint declaration on economic security cooperation, agreements on critical minerals (up to A$1.3 billion) and energy, and her description of the relationship as a quasi-alliance; and, most consequentially, constitutional revision formally on Japan's national agenda, with the government targeting a Diet proposal within the year and a national referendum likely in 2027 or 2028. The Lower House has already passed the enabling referendum procedures legislation.

The Lowy Institute's 2026 poll has Japan as Australia's most trusted nation for a sixth consecutive year, at 89 per cent, a record in two decades of polling. Calling Japan our most dependable partner used to be a forecast; it now looks like the settled mainstream view. And the fit is real: Japan's technology and capital, our resources and geography, and a shared exposure to both Chinese coercion and American unpredictability.

The demographics objection comes up every time Japan does, usually as a substitute for analysis rather than a piece of it. The constraint is real: a shrinking, ageing population limits the domestic market and, eventually, the pool of people to crew the ships Japan is now exporting. But it's a long-run constraint being applied to short-run decisions. Japan is still one of the world's largest economies. Its capital markets are deep, its industrial base sits at the technological frontier, and its government is spending decisively on defence and economic security. Over the horizon that matters to a board deciding this year or next, none of that is in question.

Our view: demography changes the investment case, it doesn't kill it. Don't go to Japan looking for a growth market. Go looking for a partner economy: defence co-production, energy and critical minerals offtake, automation and ageing-economy services where Japan is innovating because it has no choice, and a deep pool of capital looking for exactly the assets Australia has. Treat Japan as a 1980s consumer story and you'll be disappointed. Treat it as a strategic complement and you'll find the most aligned major economy in Asia.

Our call, for the record: the Diet proposal passes within the government's one-year window. The referendum is the genuine coin-toss, and we will not pretend otherwise.

5. The new export-control frontier: AI access as alliance policy

For nineteen days in June, the question of whether Washington would restrict allied access to frontier AI stopped being hypothetical. On 12 June the US Commerce Department, invoking export-control authorities, directed Anthropic to suspend access to its two most capable models, Claude Fable 5 and Claude Mythos 5, by any foreign national anywhere in the world, including foreign nationals employed inside the United States. Unable to verify user nationality in real time, the company disabled both models globally, without warning, for every customer. Access was restored in stages: the restricted model, Mythos 5, to a set of approved US organisations from 26 June, and Fable 5 globally from 1 July, following a fortnight of joint analysis with the US government and a safeguards determination by the Commerce Secretary. Mythos 5 remains limited to approved partners, with international access still under negotiation. During the outage, demand reportedly moved toward Chinese open-source alternatives.

Three lessons for the region. First, this was done with a company-specific directive, not legislation. No new rulemaking, no notice period, applicable to any model. The legal basis was never published, and whether remote use of a model even counts as an “export” is untested. But that's the point: companies comply at short notice whether or not the authority would hold up in court. Frontier releases in the United States now look less like product launches and more like negotiated deployments.

Second, the restriction ran on nationality, the same deemed-export logic that governs defence and semiconductor technology. If that template hardens, a company outside the United States with access to a controlled model could find itself obliged to restrict use by employees who hold the wrong passport, potentially including dual citizens, depending on how nationality gets defined. Australian defence suppliers already live with nationality-aware access controls under export-control law. The new part would be running that discipline on a general-purpose business tool.

Third, allied access is a policy variable, not an entitlement. Our base case isn't a cut-off; it's the tiered release, and Australians know that pattern in their bones. Once a developer has been ordered to shut off foreign nationals overnight, it has every commercial reason to design the next launch so that can never happen again. That means US and pre-cleared customers first, everyone else as approvals land. We've waited for everything here: films, streaming catalogues, technology. The original iPhone never went on sale in Australia at all; we came in at the iPhone 3G, a year later. A consumer can shrug at that. Can a business afford the same wait on frontier AI? The delay is a capability gap against your American competitor, measured in months and renewed with every model generation.

So if your business runs on US frontier models, treat access the way you'd treat any other supply-chain exposure: contractual protections, a fallback option, and an honest map of where AI actually sits in your workflows. A lasting exclusion of close allies is a tail risk. Another short-notice outage, or a built-in head start for US-domiciled competitors, is not. The strategic chokepoints of this decade include compute and model access, not just shipping lanes, and this quarter was the first in which both kinds were squeezed at once.

What we are watching