Global News Summary as of 18 September 2026

The week to 18 September made the global policy regime considerably clearer: central banks are tightening because growth has held up better than expected while the energy shock and AI investment boom are keeping inflation and demand for capital high. The Federal Reserve raised rates by 25 basis points to 3.75%–4.00% and simultaneously raised its 2026 GDP forecast to 2.3%, lowered its unemployment forecast to 4.1% and lifted its inflation projections, a distinctly non-recessionary combination. US retail sales rose a powerful 1.2% in August, although manufacturing output fell 0.3%, while the ten-year Treasury finished Friday around 5.0%. The Bank of England held Bank Rate at 3.75%, but three of nine members wanted an immediate increase; UK inflation rose to 3.1%, even as employment data softened. Importantly, the Bank also ended market sales of long-dated gilts and paused gilt auctions for six months, easing some pressure on the long end. Eurozone inflation was confirmed at 3.2%, industrial production remained virtually flat and job vacancies softened, leaving Europe with modest growth but little room for easy monetary policy. China became even more visibly two-speed: industrial production rose 5.2%, high-tech manufacturing surged 16.7% and industrial-robot output 34.6%, while retail sales grew only 0.4% and fixed-asset investment fell 7.2%. Japan joined the tightening cycle on Friday, raising its policy rate to 1.25%, the highest in 31 years, with the BOJ explicitly identifying global AI demand as both a source of growth and a contributor to semiconductor and producer-price inflation. Australia’s central bank made essentially the same point: AI is supporting global growth but also raising prices for scarce technology and driving domestic data-centre and renewable-energy investment. New Zealand grew 0.2% in the second quarter, beating the RBNZ’s flat projection, Singapore produced extraordinary 46.2% NODX growth, driven by a 131.8% surge in electronics exports on AI demand, and Switzerland raised its 2026 growth forecast sharply from 0.9% to 1.7% while retaining exceptionally low inflation and sovereign yields. Brent settled Friday at about US$103.9, lower on the week after briefly approaching US$110 but still high enough to keep inflation risk alive. The central investment message is increasingly difficult to ignore: a synchronised recession is not the base case, but the global economy is trying to finance governments, defence, AI, grids and energy infrastructure while central banks are withdrawing cheap money. Capital is expensive because demand for it is very strong.

USA

The Federal Reserve provided the week’s most important macro signal. On 16 September, the FOMC voted unanimously to raise the federal-funds target range by 25 basis points to 3.75%–4.00%. Its statement described economic activity as expanding at a solid pace, domestic spending as resilient, productivity growth as strong and capital investment as robust. Job gains were still keeping pace with the workforce and unemployment had changed little. In other words, the Fed did not tighten into a collapsing economy; it tightened because the economy remains capable of absorbing higher rates while inflation is still too high.

The new economic projections strengthened that message. The Fed raised its median 2026 real GDP forecast to 2.3% from 2.2% and its 2027 forecast to 2.4% from 2.3%. At the same time, the projected unemployment rate for both years was cut to 4.1% from 4.3%. Yet inflation was revised higher: headline PCE inflation is now expected at 3.7% in 2026, versus 3.6% previously, and core PCE at 3.4%, versus 3.3%. The median appropriate federal-funds rate is now 4.1% at the end of both 2026 and 2027, compared with June projections of 3.8% and 3.6%. That implies the median policymaker expects rates to remain substantially higher for considerably longer than markets assumed only a few months ago.

The data explain why. US retail and food-service sales increased 1.2% in August to US$773.9 billion, and were 6.0% above a year earlier. This is remarkable resilience given high fuel prices and higher borrowing costs. Consumers are clearly not behaving as though a recession has already arrived.

Industrial activity was more subdued. August industrial production was unchanged after July’s 0.2% increase, while manufacturing output fell 0.3%. Mining edged up 0.1% and utilities rose 1.8%. Total industrial production was still 1.4% higher year on year, while capacity utilisation remained at 76.3%, more than three percentage points below its long-run average. The US therefore has strong consumption and investment alongside a softer conventional manufacturing sector, rather than a uniformly booming economy.

That distinction matters for inflation. Domestic demand is still healthy enough for companies to pass through some higher costs, while manufacturing has enough spare capacity to prevent the economy from looking conventionally overheated. The Fed is therefore dealing with an awkward mixture: supply and energy inflation, exceptionally strong AI-related capital expenditure, resilient consumer demand and only moderate conventional industrial utilisation.

The Treasury market reflected that tension. The ten-year yield briefly reached 5.04% on 15 September, its highest level since 2007, before ending Friday at approximately 5.0%. The two-year finished around 4.76%. The level matters more than the daily movements. A risk-free ten-year yield around 5% materially raises discount rates for property, corporate investment, infrastructure and equities, even if economic growth itself remains healthy.

AI remains a crucial part of this story even without a blockbuster earnings release this week. The Fed’s description of capital investment as “robust” comes after a year in which data centres, semiconductors, computing and power infrastructure have become an unusually large part of US investment growth. The investment boom is productivity-enhancing, but it is also occurring in an economy with strong consumer spending and elevated energy prices. This makes AI simultaneously disinflationary in the long run through productivity and potentially inflationary in the short run through demand for capital, electricity, chips, construction and specialised labour.

Brent crude settled at approximately US$103.9 on Friday, after approaching US$110 earlier in the week. It was lower over the week, the first respite after the recent surge, but remains high enough to keep gasoline, freight and industrial costs elevated. The importance is no longer simply whether Brent rises another US$5 next week. The question is how long US$100-plus oil persists and how much of it reaches wages and prices.

Equities were remarkably resilient. The S&P 500 was essentially flat over the week, while the Nasdaq managed a modest gain and the Dow fell more materially. Technology’s relative resilience suggests investors still believe AI earnings can outrun higher discount rates, but the hurdle has plainly risen.

For investors, the US remains an inflation-and-cost-of-capital problem, not a growth-collapse problem. The Fed itself now expects stronger growth, lower unemployment and higher rates at the same time. That makes cash and short- to medium-duration Treasuries highly competitive, while long duration still requires a stronger conviction that oil, fiscal borrowing and inflation will normalise. Within equities, the focus should remain on companies able to convert heavy AI investment into durable revenue, free cash flow and high returns on incremental capital.

United Kingdom

Britain moved in the opposite direction from the US in one important respect: the labour market is softening even while inflation is rising. The unemployment rate remained 4.9% in May to July, 0.2 percentage points higher than a year earlier. Provisional PAYE data showed the number of payrolled employees falling by 26,000 in August and by 145,000 from a year earlier, to 30.2 million. Vacancies fell to 702,000, the lowest outside the pandemic period since 2014.

Wage growth is also becoming less threatening. Regular pay rose 3.5% year on year and total pay 3.9%, while private-sector regular pay growth was only 2.9%. Real regular earnings nevertheless increased 0.6% after CPIH inflation, so workers are still receiving a modest improvement in purchasing power. The labour market is therefore cooling rather than collapsing.

Inflation is moving the wrong way. August CPI rose 3.1% year on year, up from 2.9% in July, and 0.5% during the month. Transport, particularly motor fuels, was the largest upward contributor. Core CPI was 2.6%, unchanged, while services inflation remained 3.4%. This matters because much of the latest headline deterioration is still energy-related rather than evidence of a fresh domestic wage-price spiral.

Consumers nevertheless remained surprisingly active. Retail-sales volumes increased 0.5% in August, reversing July’s 0.5% decline, and were 0.9% higher over the latest three months. That sits comfortably with the unexpectedly strong July GDP figure and argues against an imminent UK recession despite weaker hiring.

The Bank of England therefore had a genuine dilemma on 17 September. It voted 6–3 to hold Bank Rate at 3.75%, with Greene, Mann and Pill preferring an immediate 25-basis-point increase to 4%. The Bank said activity had been slightly stronger than expected but acknowledged soft labour-market conditions. It also warned that inflation risks had tilted further upwards as the Middle East energy shock persisted.

The more interesting decision for bond investors was quantitative tightening. The Bank laid out a multi-year plan to run down its remaining gilt portfolio by 2034, but paused gilt auctions for six months and ended market sales of long-dated gilts. Following the announcement, the ten-year gilt yield fell about eight to nine basis points to roughly 5.24%, while the thirty-year yield dropped towards 5.74% after having touched approximately 5.96% earlier in the week. The Bank has effectively reduced one source of supply pressure at precisely the part of the curve that had been under greatest strain.

That does not make long gilts risk-free. The UK still has 3.1% inflation, a Bank of England with three votes for another hike and substantial government refinancing requirements. But it does mean the technical supply-demand picture has improved. For the first time in several weeks, the long end received a genuine institutional tailwind rather than another source of forced supply.

For investors, Britain remains a mixture of softening employment, resilient spending and stubborn inflation. Banks and companies with strong cash flows remain preferable to highly leveraged domestic businesses. Gilts are becoming increasingly interesting for income, particularly at the short and medium end, while the Bank’s change in QT policy modestly improves the case for longer maturities without removing the underlying inflation and fiscal risks.

EU and Eurozone

The final August inflation figures were slightly better than the earlier flash estimate but still uncomfortable. Euro-area inflation was confirmed at 3.2%, up from 2.9% in July but below the preliminary 3.3%. Energy contributed 1.29 percentage points to the annual rate and services another 1.43 points. Core inflation excluding energy, food, alcohol and tobacco was 2.4%, while inflation excluding energy alone was 2.1%. The picture remains one of a substantial imported-energy shock layered on top of much more moderate underlying inflation.

Growth data were far less dramatic. Euro-area industrial production fell 0.1% month on month in July and was unchanged from a year earlier. Capital-goods output rose 0.5% during the month, while energy production increased 0.9%, but non-durable consumer-goods production fell 1.6%. Germany’s industrial production declined 1.5%. Europe’s industrial economy is therefore stable at best rather than booming.

The labour market is also slowly loosening. The euro-area job-vacancy rate declined to 2.1% in the second quarter, from 2.3% in the first quarter and 2.2% a year earlier. Germany’s vacancy rate fell from 2.7% to 2.4%. This is exactly the sort of gradual softening the ECB would ordinarily welcome because it reduces pressure for wage-driven inflation without implying mass unemployment.

Following the ECB’s rate increase the previous week, European bond markets remained under pressure. Germany’s ten-year Bund traded around 3.5% on Friday and closed near 3.52%, slightly higher than the previous Friday. The policy-sensitive two-year Bund finished around 3.26%–3.27%, reflecting expectations that monetary tightening is not yet finished.

The larger structural issue has not changed. Europe has modest conventional growth but enormous capital requirements for defence, grids, renewable and nuclear energy, AI infrastructure and industrial re-shoring. Those projects increasingly compete with government borrowing for the same pool of capital. A Bund around 3.5% is not a crisis, but it changes the economics of every long-duration infrastructure and property project built on the assumption of cheap European money.

AI remains part of the solution. Europe’s capital-goods production is holding up better than consumer-goods production, while recent large investments in AI and energy infrastructure underline that digital infrastructure is becoming a genuine industrial cycle. But the region does not yet have the same explosive technology-export momentum visible in China, Japan or Singapore.

For investors, Europe remains slow-growth rather than recessionary, with inflation still too high for easy monetary policy. Banks benefit from a positive-rate environment so long as credit losses remain contained. Defence, power infrastructure and selected industrial capital goods retain structural support. Property and highly leveraged long-duration businesses remain the weak link.

China

China’s August activity data made the economy’s internal split almost impossible to miss. Industrial production accelerated to 5.2% year on year, from 4.5% in July. Equipment manufacturing increased 12.1%, while high-tech manufacturing surged 16.7%. Industrial-robot output jumped 34.6%, lithium-ion batteries 57.2% and 3D-printing equipment 29.9%. The technology and advanced-manufacturing economy is not merely holding up; it is expanding rapidly.

The consumer economy is another matter. August retail sales increased only 0.4% year on year, slowing from 0.6% in July. During the first eight months, conventional retail sales of consumer goods increased only 1.1%, although broader goods-and-services retail sales rose 2.5% and services consumption 4.9%. Online retail expanded 4.6%. Chinese consumers are spending, but hardly at a pace capable of replacing exports and industrial investment as the principal growth engine.

Fixed investment was weaker still. National fixed-asset investment fell 7.2% year on year during January to August, worsening from a 6.7% decline through July. Non-governmental investment fell 10.1%. Yet the composition again reveals the same divergence: investment in intellectual-property products increased 9.2%, information-transmission infrastructure 28.4%, and equipment purchases 9.3%. Capital is retreating from large parts of the conventional economy while continuing to flow into technology and strategic infrastructure.

This explains why descriptions of China as either “booming” or “in recession” are both unsatisfactory. Property, broad fixed investment and household spending are weak. High-tech manufacturing, advanced equipment and exports are exceptionally strong. It is a deliberate economic reallocation towards semiconductors, automation, batteries, computing and industrial technology, but one that has not yet generated equally strong household income and confidence.

The AI link is unusually clear. High-tech manufacturing growing at nearly 17%, industrial robots above 34% and information-infrastructure investment above 28% show that the AI and automation cycle is now materially affecting China’s physical economy. The question for investors is increasingly whether those returns can remain high enough to justify continued capital formation if trade restrictions intensify.

There is also an ESG dimension. The 57.2% increase in lithium-ion battery output and continued investment in electricity and information infrastructure reflect the convergence of computing, storage, transport electrification and grid investment. China is building an energy-and-digital industrial system rather than treating the green transition and technology policy as separate themes.

For investors, China remains a market for sector selection rather than broad beta. Semiconductors, automation, industrial software, robotics, batteries and grid infrastructure have real demand. Property and weakly capitalised consumer businesses remain far more dependent on policy support and household confidence. The greatest risk to advanced manufacturing is increasingly geopolitical and trade-related rather than a lack of domestic industrial capacity.

Japan

Japan supplied the week’s other major central-bank decision. On 18 September, the Bank of Japan raised its policy rate from 1.0% to 1.25%, its highest level in 31 years, by a 7–2 vote. The new rate becomes effective on 24 September. Two members preferred to leave policy unchanged because they believed the economy and inflation were not yet strong enough to justify another increase.

The most interesting part of the BOJ statement was its treatment of AI. The Bank said global AI demand is supporting Japanese exports, industrial production, corporate profits and investment. But it also said the expansion in AI demand is contributing to high producer-price inflation through higher semiconductor and related technology prices. AI is therefore explicitly appearing in central-bank monetary analysis as both a growth impulse and an inflation impulse.

The BOJ expects Japan to continue growing moderately despite the Middle East energy shock. Exports and industrial production have begun to rise, corporate profits remain high and fixed investment is trending upwards. Private consumption has remained resilient despite weak consumer sentiment, while labour conditions remain tight. Core CPI excluding fresh food is currently around 1.5%–2.0%, but the Bank expects it to move clearly above 2% in the second half of fiscal 2026 as oil, semiconductor prices and the weaker yen feed through.

The Bank also gave unusually clear forward guidance: if its economic scenario remains on track, it intends to continue raising the policy rate and reducing monetary accommodation. Financial conditions remain accommodative even after the increase, with real short- and medium-term rates low and corporate bond issuance conditions favourable. The era of assuming every BOJ hike is the last one is becoming harder to defend.

Japan’s ten-year JGB yield remained close to 3% on Friday. More surprisingly, the yen weakened after the hike, trading at around ¥157 per dollar intraday and closing near ¥156.9 as investors focused on the two dissenters and the absence of more aggressive forward guidance. The Nikkei gained about 1.4%. Tightening policy while the currency weakens is a reminder that Japan still faces a large interest-rate differential with the US.

The bond implications remain global. At a domestic ten-year yield close to 3%, Japanese insurers, banks and pension funds have a meaningful home-market alternative to US Treasuries and European sovereign debt. Even without dramatic capital repatriation, the removal of Japan as an automatic exporter of low-cost savings raises the equilibrium return required elsewhere.

The BOJ also adjusted its climate-finance facility, moving its loan rate to a floating basis and imposing upper limits while continuing support for private-sector climate projects. That is a small policy change, but a useful indication of where ESG is heading: climate investment must increasingly coexist with normalised interest rates rather than depending indefinitely on ultra-cheap central-bank funding.

For investors, Japanese automation, semiconductor equipment and advanced manufacturing remain structurally attractive. The currency remains difficult, and JGB duration continues to face further rate-normalisation risk. The more important global conclusion is that Japanese capital now has a domestic yield worth considering.

Australia

Australia had no major monthly labour or inflation release this week, but the Reserve Bank’s parliamentary testimony on 18 September provided an unusually useful summary of the economy. Governor Michele Bullock said inflation remains around or slightly above 3½%, unemployment is still only 4.5%, and the labour market remains close to, and slightly tighter than, full employment. The RBA has already increased the cash rate by 75 basis points this year.

The Bank’s problem is that the inflation risks are still worsening even as growth slows. Bullock said oil prices are feeding directly into petrol and indirectly into broader business costs, with companies already passing some of those costs through. The Bank’s August forecasts did not expect inflation to return to the midpoint of the 2%–3% target until late 2027, and it now sees some of the upside risks in that forecast materialising.

More unusually, the RBA explicitly identified the global AI boom as an inflation factor. It said AI is strengthening growth across supply-chain economies while pushing up prices for scarce AI-related technologies. Domestically, business-investment growth has accelerated sharply, driven principally by data centres and renewable-energy projects. That is precisely the AI-energy link that has been appearing elsewhere in this report: more computing requires more electricity, generation, transmission and physical infrastructure.

There is a less comfortable side to the story. Household spending growth is moderating, housing prices have fallen in most capitals and new housing loans have declined. Bullock also stressed that productivity growth remains weak, meaning Australia cannot sustain very strong demand growth without generating inflation. The RBA is deliberately trying to create a period of subdued demand to remove capacity pressure without producing a recession.

The next monetary-policy meeting is 29 September, and the central question is whether the 75 basis points of tightening already delivered this year will be enough. The Governor deliberately left the answer open. With inflation still too high, unemployment at 4.5% and AI/data-centre investment adding to demand, another increase remains entirely plausible.

Australia therefore offers perhaps the clearest example of why the AI boom complicates macro policy. New data centres and renewable projects raise potential productivity and productive capacity over time. During construction, however, they consume labour, power equipment, land, financing and imported technology, all while an oil shock is already raising costs.

For investors, Australia remains a modest-growth economy with real inflation and capital-demand pressure, not a recession economy. Banks and resources still have supportive nominal conditions. Property and leveraged infrastructure remain rate-sensitive. The most interesting structural opportunities continue to sit around data centres, power generation, grids, storage and the equipment needed to connect them, provided projects can earn returns above a much higher cost of capital.

New Zealand

New Zealand’s second-quarter GDP release on 17 September showed real GDP growing 0.2% quarter on quarter, beating the RBNZ’s flat projection; several banks had earlier pencilled in a contraction. First-quarter growth was reported at 0.9%. Nine of sixteen industries expanded, with construction rising 2.7%, its strongest quarterly increase since 2023.

That does not make the economy strong. Growth slowed dramatically from the first quarter, and the RBNZ continues to estimate substantial spare capacity. The economy is better described as a fragile recovery than a robust expansion. This matters because New Zealand’s central bank is simultaneously raising rates to contain energy-driven inflation.

The external accounts show why the energy shock matters so much. The seasonally adjusted current-account deficit narrowed by NZ$666 million to NZ$3.8 billion in the June quarter, but the goods deficit widened to NZ$1.5 billion because imports rose 8.2%, led by petroleum and petroleum products. Diesel, petrol and jet-fuel prices were the main drivers. Goods exports rose 6.6%, led by dairy and meat.

For the year to June, however, the current-account deficit widened to NZ$14.6 billion, or 3.2% of GDP, from 3.0% in March. New Zealand’s net international liability position nevertheless improved significantly to NZ$178.3 billion from NZ$191.7 billion, helped by gains in overseas share portfolios.

The GDP result makes the RBNZ’s recent tightening somewhat easier to justify, but only marginally. Inflation remains above target and the Bank raised the OCR to 2.75% earlier this month, yet domestic spare capacity is still significant. If oil continues falling from its recent highs, New Zealand may eventually offer one of the cleaner developed-market duration opportunities because growth is much weaker than in the US or Australia.

For investors, that opportunity is not quite here yet. The central bank is still tightening, and the current account demonstrates the economy’s sensitivity to imported fuel. But the combination of weak domestic demand, significant spare capacity and only 0.2% quarterly GDP growth makes New Zealand one of the first markets where a sustained easing in energy inflation could genuinely change the fixed-income argument.

Singapore

Singapore delivered perhaps the strongest AI-related macro data of the week. On 17 September, Enterprise Singapore reported that non-oil domestic exports increased 46.2% year on year in August, accelerating from 24.1% in July. Electronics NODX surged an extraordinary 131.8%, which Enterprise Singapore explicitly attributed to strong AI-related demand.

The detail is even more striking. Integrated-circuit exports rose 90.9%, disk-media products 290.2%, PCs 237.9% and specialised machinery 57.7%. Non-electronics NODX also returned to growth at 12.0%. Total merchandise trade rose 44.5%, while NODX for the first eight months of 2026 was 22.4% higher than a year earlier.

Geographically, the expansion was broad. NODX to the US increased 91%, China 70.3% and South Korea 87.1%. That matters because it suggests Singapore’s export strength is not simply one customer or one trans-shipment route; it reflects the breadth of the regional AI and semiconductor investment cycle.

This is possibly the clearest evidence in the entire report that AI has moved from a stock-market theme into macroeconomic data. Singapore is a small, open economy, so semiconductor investment appears rapidly in exports, manufacturing and trade. The benefit is obvious. The vulnerability is equally clear: if global AI capex ever slows sharply, Singapore will feel it sooner than most developed economies.

Energy remains the constraint. AI hardware and data centres require large quantities of electricity, while Singapore imports almost all its primary energy. Regional power links, cooling efficiency and low-energy computing are therefore not merely ESG considerations; they are strategic economic infrastructure.

For investors, Singapore remains exceptionally well placed in the current cycle. Institutional strength, semiconductor exposure and a large trade and financial-services ecosystem give it direct access to AI growth. The risk is concentration rather than recession. As long as the AI capital cycle remains intact, Singapore remains one of its clearest macro beneficiaries.

Switzerland

Switzerland delivered another substantial upside surprise, this time in its official economic forecast. On 17 September, SECO raised its forecast for 2026 GDP growth from just 0.9% to 1.7%, while leaving 2027 growth at 1.6%. The revision followed exceptionally strong sport-adjusted GDP growth of 1.5% in the second quarter.

There is an important qualification. SECO estimates that almost half of the second-quarter expansion came from the volatile chemicals and pharmaceuticals sector and expects some correction during the second half. The 1.7% forecast therefore does not mean Switzerland will continue growing at anything resembling the second quarter’s pace. It does mean the economy entered the second half from a much stronger position than previously thought.

Inflation remains Switzerland’s remarkable advantage. SECO left its forecast for average annual inflation at only 0.6% for both 2026 and 2027. Investment is being supported by rising capacity utilisation, private consumption continues to expand moderately and the weaker franc has helped exporters. Unemployment is expected to average 3.1% this year and fall to 3.0% in 2027.

Swiss sovereign debt remains in a different universe from the rest of this report. As of 18 September, the SNB policy rate remained 0%, while the SNB’s 10-year spot rate on Confederation bonds was approximately 0.581%. Compare that with roughly 5% for a US Treasury, above 5% for UK gilts and almost 3% for a JGB. Switzerland offers very little income because investors require very little compensation for inflation and fiscal risk.

The Middle East remains the principal external risk. SECO specifically warned that persistent high oil prices and potential natural-gas pressure could weaken global activity and lift Swiss inflation. Even then, Switzerland begins from a far more favourable inflation position than most developed economies.

For investors, Swiss bonds remain primarily a capital-preservation and diversification asset, not an income asset. Swiss equities remain attractive for pharmaceuticals, specialised industrials and globally competitive high-value businesses. Near-term recession risk is low, although the pharmaceutical contribution means the headline GDP figures should not be extrapolated casually.

What this implied for markets

The week’s clearest message was that the global tightening cycle is no longer theoretical. The Federal Reserve raised rates to 3.75%–4.00%. The Bank of Japan raised rates to 1.25%. The Bank of England held at 3.75%, but one-third of its committee wanted an immediate increase. The ECB had increased rates only a week earlier, the RBNZ raised rates earlier this month, and the RBA is openly considering whether its existing tightening is sufficient.

This is occurring because the global economy has not rolled over.

The Fed raised its growth forecast while cutting its unemployment forecast. US retail sales rose 1.2%. UK retail sales increased despite higher fuel costs. Eurozone industry is flat rather than collapsing. China’s high-tech manufacturing is growing 16.7%. Japan’s exports and industrial production are being supported by AI demand. New Zealand managed positive GDP growth. Singapore’s AI-heavy electronics exports are up 131.8%. Switzerland just almost doubled its 2026 growth forecast.

That is not an Armageddon economy.

The more interesting problem is that central banks are tightening into an investment boom rather than trying to rescue a recession. AI, defence, energy security, grids and government deficits all require large amounts of capital. The economic returns from some of that spending may be excellent. But capital cannot simultaneously become more scarce and remain cheap.

The Treasury market is saying exactly that. The US ten-year ended around 5%. UK ten-year gilts remain above 5%, despite rallying after the Bank of England changed its QT programme. Bunds closed around 3.52%. JGBs are close to 3%. Switzerland is the exception precisely because its inflation and fiscal dynamics are exceptional.

Another conclusion is that AI is now unquestionably macroeconomic. China reported high-tech manufacturing growth of 16.7% and industrial-robot growth of 34.6%. Singapore electronics exports rose 131.8%. The Bank of Japan explicitly said AI demand is raising both output and semiconductor prices. The RBA said AI is supporting world growth while increasing prices for scarce technology, and that domestic business investment is being driven by data centres and renewable energy.

That is a remarkable change from even a year ago. AI is appearing simultaneously in exports, industrial production, capital expenditure, electricity planning, inflation forecasts and central-bank decisions.

There is an important investment distinction inside that boom. The fact that AI demand is real does not mean every AI investment earns an adequate return. The capital-efficient versus capital-hungry distinction remains crucial. Revenue growth has to be compared with capex, free cash flow and incremental return on invested capital. With sovereign yields around 5%, the tolerance for businesses that consume unlimited capital in exchange for distant profits should fall.

The fifth conclusion is that the short and medium end of sovereign curves still offers the cleanest fixed-income proposition. Short-duration government securities provide substantial income without requiring investors to assume that inflation, fiscal supply and term premia will quickly normalise.

The medium part of the curve is becoming more interesting as yields rise. A five- to ten-year security purchased near current yields can provide meaningful carry while also offering capital appreciation if inflation eventually subsides. That is a different proposition from deliberately maximising duration in thirty-year bonds while fiscal supply, oil and global capital demand remain uncertain.

This distinction is increasingly important. Owning duration is not the same thing as owning the longest bond available. There is a point at which additional maturity adds far more volatility than expected return. In the present environment, the short and medium end can provide much of the income benefit without requiring a heroic macroeconomic forecast.

The Bank of England’s decision this week supports that argument in an interesting way. By ending market sales of long-dated gilts and pausing gilt auctions for six months, it removed a technical pressure from the long end. That makes long gilts somewhat more attractive. It does not eliminate inflation risk. The same applies to US Treasuries: a 5% ten-year yield is increasingly attractive, but a thirty-year bond remains far more exposed to fiscal policy, term premia and structural capital demand.

The next conclusion is that oil remains the variable most capable of changing the story quickly. Brent settled around US$103.9 after approaching US$110, which is welcome. If the decline continues, headline inflation will ease, household purchasing power will improve and some of the urgency behind central-bank tightening will disappear. If oil returns above US$110 and stays there, the opposite happens.

For the moment, the data do not suggest consumers have broken under the pressure. That is why the Fed, BOJ and others still have room to tighten.

ESG is also becoming a much more concrete investment theme. The most investable part is increasingly physical infrastructure rather than labelling. Australia’s data-centre investment requires renewable energy. China’s AI and automation industries require batteries, grids and power. Singapore needs electricity imports and efficient cooling. Japan is maintaining climate-finance facilities even as monetary policy normalises. These projects exist because economies physically need them, not because someone has attached an ESG score.

For equities, the preference remains quality growth backed by actual cash generation. AI infrastructure, semiconductors, memory, networking, automation, electrical equipment, power generation and grids all retain genuine demand. Banks can benefit from positive nominal rates provided credit quality remains sound. Energy remains useful both as an earnings exposure and a partial hedge against the inflation shock.

The vulnerable areas remain equally familiar: highly leveraged property, companies dependent on repeated refinancing, speculative long-duration growth and businesses whose capital requirements are rising faster than their eventual economics can justify.

The broad portfolio message therefore remains quality, liquidity and selectivity.

But there is an important shift in emphasis.

For much of the past few years the question was whether economic weakness would force central banks to cut. The present question is almost the reverse:

How high can the cost of capital rise before resilient economies and the AI investment boom finally begin to slow?

We are not there yet.

Economic growth remains alive. Labour markets are generally stable. AI demand remains exceptionally strong. A synchronised global recession is still not the base case.

The warning is coming instead from bonds. Capital is no longer cheap, and the world wants an extraordinary amount of it.

The winners will increasingly be the economies and companies that can turn that expensive capital into real productivity, real revenue and real free cash flow.