Anyone who looked only at the index on 30 September 2026 saw a calm quarter: the S&P 500 2% higher, the Nasdaq 100 almost unchanged, the VIX at 16.3, exactly where it stood at the end of June. The AI Macro Cockpit, which tracks the concentration of the equity market and the financing conditions for the build-out of artificial intelligence (AI) through eight signals, paints a different picture of the same three months. The US ten-year yield climbed from 4.44% to 5.29%, and Bunds and Japanese government bonds followed. The equal-weighted S&P 500 lost 4.3% against the cap-weighted index, hitting its low on the last trading day. Semiconductors fell almost 12% against the Nasdaq, while NVIDIA alone beat its sector by 29%. And in the last week of September the risk premium on high-yield bonds jumped from 2.66% to 3.12% in six trading days.
The common reading of the quarter is that the AI trade survived a wobble in July and carries on. The data support a narrower thesis. The AI trade has survived, but its price has changed. Since the summer the capital market has paid for data-centre investment only where it shows up in reported cloud revenue, and it demands noticeably higher interest on the debt that funds a growing share of it. That is not a diagnosis of a bubble. The finding concerns financing: until now the large groups paid for the AI build-out almost entirely from operating cash flow. Since the summer, debt has carried a growing share, through bonds issued by the large cloud providers (hyperscalers) and loans to smaller infrastructure providers. This shift changes where the first problems become visible.
This report places the quarter in context in seven steps: the macro backdrop, concentration, tech and semiconductors, credit, consulting and IT services, banks and the car industry. It closes with a risk assessment and a risk-based outlook for the fourth quarter, whose first four weeks already feed into the assessment.
Interest rates were the driver. The Federal Reserve (Fed), the European Central Bank (ECB) and the Bank of Japan (BoJ) raised policy rates between June and September, in response to energy inflation in a robust economy. The inflation traces back to the war with Iran and the disruption of shipping through the Strait of Hormuz.
The index narrowed. The ratio of the equal-weighted to the cap-weighted S&P 500 reached its quarterly low on 30 September. In October breadth recovered, while several heavyweights fell after their results.
The market wants evidence for the investment. Microsoft and Amazon were rewarded the day after their results for reported cloud growth, while Alphabet was punished for negative free cash flow and Meta for spending plans without a comparable revenue line. NVIDIA remained the exception in a weak semiconductor chain.
Debt carries a growing share. The hyperscalers and Oracle issued around USD 220 billion in bonds by August, spreads and private credit defaults rose, and several large private credit funds limited redemptions.
In consulting, AI shows up in prices first. The largest share price swings at Accenture and IBM had operational triggers with no direct link to AI. The effect of AI shows up in pricing pressure, slower growth and less hiring.
Banks profited from volatility. Their AI risk sits in credit lines to private credit funds and is not yet visible in the numbers.
The car industry is losing in China. So far AI is mainly a cost item there.
Oil and a robust economy pushed three central banks to tighten
The cause of the rate shock predates the quarter. With the US-Israeli strikes on Iran from 28 February 2026 and the closure of the Strait of Hormuz, Brent rose to as much as USD 126.41 on 30 April. After a ceasefire in April the price fell to USD 72.92 by the end of June, and in the third quarter attacks on tankers and American counterstrikes flared up again. Brent ended September at USD 103.50, a gain of 42% over the quarter. US consumer prices rose by 3.7% in September, according to the Bureau of Labor Statistics, with energy alone up 20.4%. For the euro area, Eurostat reported 3.3% for August, with the core rate at 2.4%.
The central banks reacted even though this is a supply shock that monetary policy cannot fix. The ECB raised rates by 25 basis points in June and again on 10 September, taking the deposit rate to 2.50%. The Bank of Japan raised its rate to 1.25% on 18 September, the highest level in 31 years. The Federal Reserve under Kevin Warsh held in July, with three dissenting votes in favour of a hike, and raised the target range to 3.75% to 4.00% on 16 September. It was the first rate rise since July 2023. The Fed’s projections point to further hikes, and the futures market does not price in cuts until 2028.
The bond market moved so forcefully because the hikes met a robust economy. According to the first estimate by the Bureau of Economic Analysis, the US economy grew at an annualised 3.4% in the third quarter. On 23 September, after strong purchasing managers’ indices and weak demand at an auction of five-year Treasuries, the ten-year yield rose by 13 basis points in a single day to its highest level since 2007. The real yield, measured on inflation-linked bonds, rose from 2.20% to 2.93% over the quarter and accounted for most of the move. A higher real yield raises the discount rate in every valuation that depends on distant cash flows, and so hits growth stocks hardest. The US yield curve steepened slightly in the process (the gap between the two-year and ten-year yield, 2s10s for short, widened from 30 to 41 basis points), while the European curve flattened (from 40 to 36 basis points), because the two-year Bund yield rose by 74 basis points, more than the ten-year.
There was no abrupt exit from yen funding. The yen gained only 2.8% after the Japanese rate rise, from 161.9 to 157.4 per dollar. The cockpit’s carry unwind signal, which flags the unwinding of positions funded with cheap yen loans, stayed quiet. The Nikkei lost 4.7% over the quarter, but without the momentum of August 2024.
The index rose on ever fewer shoulders
The first cockpit signal answers the question of how heavily the American equity market depends on its largest stocks. It sets the equal-weighted S&P 500 (ETF RSP) against the cap-weighted index (SPY). When the ratio falls, a few large stocks are carrying the index. In the third quarter it fell by 4.3%, while SPY rose by 2.1%. The average stock in the index lost value, in other words, while the index itself gained.
The chart shows two phases. Breadth held up until mid-August. The ratio even peaked on 29 July, the quarterly low for equities, because the large technology stocks fell by more than the average stock that day. From mid-August concentration took over, driven by Microsoft (up 37.5% over the quarter) and Meta (up 28.7%), which had more than recovered its loss after its July results by the end of September. The ratio ended the quarter at its low. Higher rates hit smaller, more indebted companies harder than groups with net cash. In the third quarter the signal therefore measures both the AI lead of the large groups and their lower sensitivity to interest rates.
In October the picture turned. Since 30 September the ratio has recovered 4.9%: the equal-weighted index rose by 5.8% to 29 October, the cap-weighted index by only 0.9%. The average stock recovered, while several heavyweights fell after their quarterly results, Meta by 7% on 22 October and Alphabet by almost 13% on 29 October. When an index depends on a few stocks, every disappointment at one of them becomes a move in the index, and that works in both directions.
The market wants reported cloud revenue as proof that data centres pay
29 July was the day on which almost every stock in the cockpit marked its low for the quarter. The S&P 500 fell 1.5%, and the Nasdaq 100 stood 11% below its June record. Three triggers coincided: the Fed held rates with three votes for a hike, SK Hynix had missed expectations the day before, and Vertiv, a maker of cooling and power equipment for data centres, lost 17% despite raising its full-year forecast, because revenue came in below expectations. Vertiv and SK Hynix raised the same question: how much longer will the hyperscalers’ investment budgets (capital expenditure, or capex for short) hold up?
The answer came that evening and the following day, and it differed from company to company. Microsoft reported growth of 43% for Azure, and annual cloud revenue passed USD 100 billion for the first time. The shares rose 15.5% on 30 July, adding around USD 450 billion in market value, the largest one-day gain in dollar terms according to Reuters. Amazon followed with AWS growth of 37% and a backlog of USD 496 billion, and its shares rose 15.3%. Meta, by contrast, planned capital spending of up to USD 145 billion without a comparable revenue line to show for it, and lost 8% on 30 July. The shares more than recovered the loss by the end of September, but the verdict on results day was unambiguous. A week earlier, Alphabet had reported its first negative free cash flow since its IPO, and its shares lost 7% on 23 July.
In October the pattern repeated itself, and the punishment was harsher. On 28 October Microsoft reported Azure growth of 48% and free cash flow around USD 10 billion above expectations, and the shares held their level. Alphabet reported negative free cash flow of USD 9.1 billion on the same evening, with capital spending of USD 51 billion in the quarter and weaker search revenue, and lost almost 13% on 29 October. S&P Global and Visible Alpha calculate that infrastructure spending by the large technology groups will rise from USD 404 billion in 2025 to around USD 790 billion in 2026. The debate is therefore less about the total than about who can pay for it out of operating cash flow.
The fourth cockpit signal asks whether the whole supply chain is holding up or only NVIDIA. In the third quarter the chain did not hold, and NVIDIA was the exception. The SOXX semiconductor ETF lost 11.3%, and its ratio to the Nasdaq 100 fell by 11.7%. NVIDIA, by contrast, rose by 14.1% and on 26 August reported quarterly revenue of USD 96.2 billion, up 106% year on year, with guidance of USD 108 billion. The rest of the chain suffered for reasons that have little to do with AI demand: Micron lost as much as 36% at one point because the IPO of the Chinese memory maker CXMT called pricing power in the memory market into question, ASML lost 8.9% on reports of a Chinese lithography machine, and Vertiv 27.9%. According to analysts at Bernstein, memory capacity itself is sold out for 2026 and 2027. The share price losses reflect concern about future margins.
In software the split ran right through the sector. Palantir rose 60.3% over the quarter after revenue growth of 93%, and ServiceNow 35.0%. On 29 October ServiceNow then lost 27% in a single day after barely raising its forecast for the subscription business. The market read this as confirmation of a thesis that has been circulating for months under the label SaaSpocalypse: that AI agents will displace the per-user licensing models of classic rented software (software as a service, SaaS). The revenue of the large software vendors does not show this yet. The market still prices the thesis in at the first disappointment, and the ECB and the Bank for International Settlements (BIS) already see it as one trigger of the stress at credit funds that finance many software companies.
The two largest private AI companies are carried in the cockpit as cards without a share price. Anthropic filed its IPO prospectus with the US Securities and Exchange Commission (SEC) on 26 October. It shows revenue of USD 16.3 billion for the first half of 2026, up from USD 1.2 billion a year earlier, and compute commitments of around USD 417 billion over several years. According to Reuters and CNBC, OpenAI is negotiating a funding round of around USD 30 billion at a valuation between USD 1.25 and 1.4 trillion, which has not closed. Both sets of figures show the same structure as the hyperscalers, only more compressed: fast-growing revenue and even faster-growing commitments.
Financing the AI build-out is becoming a credit question
The second cockpit signal tracks the price of credit. The risk premium on American high-yield bonds over Treasuries (ICE BofA High Yield Option-Adjusted Spread, or HY spread for short) moved in a narrow band around 2.7% from July to mid-September, with its low at 2.60% on 28 August. From 23 September it rose to 3.12% in six trading days, and in October further to 3.40% on 22 October. That level is far below earlier stress readings, but the rise was the fastest of the year and coincided with the peak in yields. The HY spread is not a pure AI signal, as it also responds to rates and the economy. It does, however, measure the price of credit for weaker borrowers, among them the leveraged infrastructure providers, while the hyperscalers themselves issue in the investment-grade segment.
A second signal in the same chart seems to point the other way. The ratio of the HYG high-yield ETF to the LQD investment-grade ETF rose by 3.1% over the quarter, as if money were flowing into riskier credit. The explanation lies in duration. LQD holds longer-dated bonds and lost 6.3% as yields rose by 85 basis points, while HYG lost only 3.5%. Here the ratio mainly measures the rate effect. For the same reason the ratio of seven-to-ten-year Treasuries to money market paper (IEF ÷ BIL) fell by 5.6%. Anyone holding short-dated paper during the quarter lost nothing.
At the same time, the financing of the AI build-out is shifting. By 10 August the hyperscalers and Oracle had together issued around USD 220 billion in bonds. Microsoft did not join until 29 October, with a USD 25 billion issue, its first bond sale since 2017. At the edge of the market it is getting expensive: the 9.75% bond due 2031 issued by CoreWeave, a provider of rented computing power with around USD 50 billion in debt and commitments, yielded 15.8% on 15 October, and the shares have lost 30% since the end of the quarter. In the Global Financial Stability Report published on 13 October, the International Monetary Fund (IMF) counts fourteen publicly known AI financings since 2023, totalling around USD 650 billion of debt and USD 300 billion of equity, against future revenue commitments of USD 1.4 trillion.
The private credit market is also showing fatigue. In mid-September the rating agency Fitch reported a default rate of 6.3% for the twelve months to August, a record. Blackstone’s BCRED fund, Apollo, Cliffwater and HPS limited redemptions to 5% of fund assets in the third quarter, in some cases for the third time in a row. The ECB and the BIS name AI disruption in the software sector as one trigger, since software companies make up a large part of private credit portfolios. In Bank of America’s fund manager survey in October, 39% of respondents saw hyperscaler capital spending as the most likely source of a systemic credit event, down from 42% in September. The concern is widespread, but it is no longer growing. On 29 October ECB President Christine Lagarde attributed part of the global rise in yields to bond issuance by the hyperscalers, although so far this has not really crowded out other borrowers.
In consulting, AI shows up in prices first
No industry was written off as an AI loser as often in 2026 as consulting. In the third quarter the share prices of the large firms seem to supply evidence and counter-evidence at once. Accenture lost around 18% in a single day on 18 June and then rose 47.4% in the third quarter. IBM lost a quarter of its market value on 14 July. Cognizant gained 49.2% over the quarter, EPAM 36.6% and Capgemini 18.7%, while Wipro lost 25.8%. The two largest single moves, Accenture in June and IBM in July, had operational triggers with no direct link to AI.
The slump in Accenture’s shares on 18 June, still in the second quarter, came after tight corporate budgets, lower bookings (USD 19.3 billion, down from USD 19.7 billion) and a trimmed full-year forecast, according to Barron’s, with weaker US federal government business also weighing. The subsequent recovery was mainly a re-rating: the price-earnings ratio rose from 10.0 to 16.6 between the end of June and mid-October, while adjusted earnings per share grew by 8%. On 1 October the shares then gained 15.8%, after Accenture reported 7% growth in local currency for its fiscal fourth quarter and a record 141 large deals worth more than USD 100 million each. The IBM slump on 14 July was a hardware event: chief executive Arvind Krishna cited a revenue shortfall in the mainframe business and delayed large deals. IBM’s consulting segment stagnated at around USD 5.3 billion in both the second and the third quarter.
The effect of AI shows up elsewhere, in prices and in growth. On 23 October Infosys cut its growth forecast for the fiscal year from a range of 1.5% to 3% to a range of 1% to 2%, citing AI-driven pricing pressure, tougher competition and weak demand. Chief executive Salil Parekh spoke of deflation in the existing business on the analyst call. Accenture chief executive Julie Sweet acknowledged lower prices in many areas on her analyst call, and said that AI efficiencies were more than offset by new work. At the Indian IT providers, AI revenue is growing at double-digit rates (78% at HCLTech), and at TCS it accounts for more than 10% of revenue, yet overall growth remains between 0% and 3%. AI revenue is therefore replacing part of the existing volume rather than adding to it. Wipro stagnated over the quarter and cut its workforce by around 6,700 jobs, while Cognizant grew by 4.8% and narrowed its full-year forecast to 4.0% to 4.4%. The strong share price gains at Cognizant, EPAM and Capgemini mainly reflect a recovery in valuations after the June sell-off.
For the German market, the 2026 Lünendonk list already shows the trend for 2025. The 20 largest German management consultancies grew by only 4.1% in 2025, after 7.5% the year before, six of them lost revenue, and revenue per consultant is falling despite the use of AI. The pyramid is losing its base: starting salaries in strategy consulting have stagnated for the third year running, and graduate intakes at PwC and KPMG in the UK have been cut. Anyone who spent twenty years on a trading floor knows the pattern from the electronification of trading. First the margin per trade fell. Only years later did the number of traders follow. Consulting is in the first phase of that development.
The AI risk at banks sits in credit lines to credit funds
In mid-October the large US banks delivered quarterly results that reflect the environment well. Income came from equities trading (up 49% at JPMorgan, 72% at Goldman Sachs and 41% at Morgan Stanley) and from net interest income, which rose by 13% at JPMorgan. Fixed income trading stagnated or shrank, IPOs were weaker, and loan loss provisions remained unremarkable. In Europe, Deutsche Bank reported a record profit after tax of EUR 1.9 billion, and BNP Paribas a 13.5% rise in net profit. Higher rates and heavy equity market volumes make for a good environment for banks with a large trading business.
Share prices followed only in part. The European banking sector gained 6.0% in the third quarter and has lost 10.1% since (measured by the iShares Stoxx Europe 600 Banks, not in the cockpit). On 22 October the euro area banking index fell by 4.2%, its steepest one-day loss since March, after Bankinter warned of rising deposit costs and the spread of French government bonds over Bunds rose to a 15-year high. Rising rates help banks only as long as they do not make deposits more expensive at the same pace.
The AI theme affects banks in two ways. The first is operational: on 22 October Barclays announced around 5,000 job cuts linked to AI, and Citigroup has booked more than USD 1 billion in severance since the start of the year and points to automation. The second is credit risk, and it matters more. Banks rarely finance the AI build-out directly. They finance the funds that finance it. According to an analysis cited in the financial stability reports, 63% of bank credit commitments to business development companies (BDCs), the listed funds that provide direct lending (private credit), sit with the five largest US banks. The IMF considers the banking sector as a whole robust in its stress test, but warns of a tail of weakly capitalised banks. None of this shows in the quarterly figures. It would become visible if funds under redemption and default pressure drew on their committed credit lines to pay out investors or to support their portfolios. The risk would then move from the fund onto the bank’s balance sheet.
The slump in the car industry has its roots in China
The European car industry is having a difficult year, but its third-quarter story is not primarily an AI story. On 29 October Volkswagen reported an operating loss of EUR 5.86 billion, its largest since 2015, driven by around EUR 8 billion of one-off charges, including a goodwill impairment on Porsche of EUR 6.4 billion. Deliveries in China fell by 38.5%. On 27 October Mercedes-Benz cut its margin forecast for the car business for the second time, to 2% to 3%, while BMW had already warned in June and reports its results on 4 November. The Chinese domestic market shrank in August for the eleventh month in a row, while Chinese manufacturers exported record volumes.
The European auto sector barely fell further in the third quarter (down 1.3%). Most of its loss of around 21% over the year came in the first half, and the Stoxx Europe 600 Automobiles & Parts did not reach its 52-week low until 8 October. Individual stocks diverge widely: Stellantis lost 21.7% over the quarter, Tesla 15.6% and Mercedes-Benz 7.6%, while Volkswagen’s preference shares ended the quarter almost unchanged (up 0.5%), even though they lost around 13% between the profit warning on 18 September and the end of the quarter. The manufacturers’ own reports offer only indirect evidence that rates and oil prices are to blame, for example higher costs in the Mercedes-Benz outlook.
So far AI appears in the industry on the cost side. On 16 October Mercedes-Benz extended its partnership with NVIDIA to autonomous driving functions up to Level 4 and to AI-supported production. Tesla counts 1.74 million active subscriptions to its driver assistance system and runs a driverless fleet in Texas, but does not expect robotaxis to contribute to earnings until 2027. Until then, spending on AI and on the humanoid robot Optimus is weighing on free cash flow, which fell by 62% in the quarter. For the European manufacturers, AI is an investment whose return arrives later than the bill, and in a year when cash is tight anyway.
The telecoms wave around 2000 shows where investment booms break first
Any comparison of the AI build-out with the dotcom bubble falls short if it starts with share prices. The more revealing comparison is the telecoms investment wave around the year 2000. Network operators and equipment makers then invested enormous sums in fibre and mobile licences. The German UMTS auction in August 2000 alone raised around EUR 50 billion. Much of this was financed with bonds and bank loans. Demand for bandwidth did grow, only more slowly than capacity. The most severe breaks happened in the credit market: risk premiums on telecoms bonds widened, smaller and more indebted providers lost market access and went bankrupt, while the large groups survived, albeit with heavy write-downs.
The parallels in the third quarter of 2026 lie in the structure, while the scale is different. Demand is demonstrably growing, with Azure and AWS up 43% and 37% over the summer. Debt is taking on a growing share of the financing. And the smaller players pay first: CoreWeave at a yield of more than 15%, while Microsoft places its 40-year bond at a spread of 105 basis points. The difference lies in the borrowers’ balance sheets. The large hyperscalers still finance their investment mainly from operating cash flow, and the share of debt is growing mainly at the edges. The telecoms groups of 2000 had no such reserve. Veterans of those years still recognise the pattern: the default rate at the edge of the market rises while the centre remains carefree. That is why the cockpit tracks credit and not just share prices.
The quarter’s risk assessment points to financing
The table below summarises which risks visibly increased in the third quarter and what supports that picture. The assessment is the author’s own, based on the cockpit signals and the reported figures.
| Risk | Evidence in the quarter | Trend |
|---|---|---|
| Rates and inflation | US 10-year yield +85 basis points to 5.29%, real yield +73 basis points, Fed hike on 16 Sep | Increased |
| Energy | Brent +42% to USD 103.50, US CPI energy +20.4% | Increased |
| High-yield credit, AI infrastructure | HY OAS 2.66% → 3.12% in six days, CoreWeave bond > 15% | Increased |
| Private credit | Fitch default rate 6.3%, redemption limits at four large funds | Elevated |
| Index concentration | RSP ÷ SPY −4.3%, low on 30 Sep | Elevated |
| Software disruption | ServiceNow −27% on 29 Oct, pricing pressure at IT service providers | Elevated |
| Yen carry | USD/JPY −2.8% despite BoJ hike, no abrupt exit | Calm |
On the other side of the ledger there are solid supports. The US economy grew strongly at 3.4%. With cloud growth rates of 37% to 48% and an AWS backlog of USD 496 billion, Microsoft and Amazon show that demand is real. NVIDIA beat expectations again and raised its guidance. Accenture reported a record in large deals, the big banks strong equities trading income, and Deutsche Bank a record profit. The quarter confirmed the AI thesis and at the same time tightened its financing conditions.
Oil and the credit market will decide the fourth quarter
The first four weeks of the fourth quarter have eased some tensions and sharpened others. The central banks stood still: the Fed on 28 October at 3.75% to 4.00%, with two dissenting votes for a further hike, the ECB on 29 October at 2.50% and the BoJ on 30 October at 1.25%. The US ten-year yield reached its high for the year of 5.31% on 5 October and eased afterwards (5.05% on 28 October), Bunds traded at 3.38% on 30 October, index breadth recovered and the yen remained stable. At the same time Brent rose by a further 9.4% to 30 October, the HY spread hit a high for the year of 3.40% on 22 October, the VIX jumped sharply on 16 October and again on 22 October, and Alphabet, ServiceNow, Meta and CoreWeave lost double digits after quarterly results or financing news. The outlook therefore describes a base case, two risk paths and one opportunity, each with the cockpit signals that show which path is materialising. Probabilities are assessed qualitatively.
Yields stay high but below the 5 October high for the year, and the Fed raises rates once more in December, as the futures market prices in. The credit market continues to draw a sharp line between hyperscalers with positive cash flow and highly indebted infrastructure providers. The indices fluctuate without breaking.
Confirmed by: HY spread between 3.0% and 3.4%, US ten-year yield within its October range of 4.96% to 5.31%, semiconductors against the Nasdaq 100 (SOXX ÷ QQQ) without a new low.
A highly leveraged provider loses access to the capital market or has to renegotiate contracts. The stress reaches the banks through private credit funds and credit lines, and the hyperscalers cut their spending plans for 2027. Semiconductors and infrastructure suppliers fall first, and the index follows through its concentration.
Early warning: HY spread above the October high of 3.40%, high-yield against investment-grade ETF (HYG ÷ LQD) falling despite lower yields, NVIDIA losing ground against the semiconductor sector.
Further attacks on shipping in the Gulf drive Brent above its October high. Inflation picks up again, and the Fed and the ECB raise rates more than expected. Short-dated yields then rise faster than long-dated ones, because the market prices in further rate steps, so the yield curve flattens, unlike in the third quarter, when the long end did most of the rising. Highly valued growth stocks lose the most.
Early warning: Brent above its October high of USD 116.72, US ten-year yield above its high for the year of 5.31%, ten-year Bund above its September high of 3.66%, US 2s10s curve below 30 basis points.
A durable agreement reopens the sea lanes. Brent falls sharply, inflation expectations decline, and the Fed refrains from a hike in December. Yields fall, equity market breadth increases, and cyclical sectors such as autos and banks outside the interest business benefit.
Confirmed by: Brent below its October low of USD 100.20, US ten-year yield below 4.96%, equal-weighted S&P 500 continuing to gain against the cap-weighted index.
For banks and financial service providers this leads to four checkpoints for the fourth quarter.
First, their own credit commitments to private credit funds and BDCs, and how much of them has already been drawn.
Second, data centre financing and counterparty risk towards leveraged infrastructure providers whose bonds now yield double digits.
Third, the sensitivity of deposit costs if central banks raise rates further.
Fourth, concentration in their own securities holdings and in client portfolios that depend on a few technology stocks through the index.
Several dates in the coming weeks could decide between these paths: the US midterm elections on 3 November, BMW’s quarterly results on 4 November and EPAM’s on 5 November, the Fed meeting on 8 and 9 December, for which the futures market prices in a hike with high probability, and Accenture’s results around 16 December. NVIDIA’s next quarterly results and Anthropic’s IPO, which media reports expect in mid-November, are also in play. The company has not yet announced a date. Christine Lagarde’s announcement that she will leave the ECB on 7 February 2027 adds a personnel uncertainty to the European rate question.
The third quarter of 2026 tightened the conditions of the AI boom. Demand for computing power is real and growing, but the capital market now pays for it only where it shows up in revenue and cash flow. Financing is moving into the bond and private credit markets, and prices there rose over the quarter. The first problems will therefore not show up in the share prices of the large groups, but in the spreads of smaller borrowers and in redemptions at credit funds. For the fourth quarter, two signals deserve attention first: the oil price as the most likely trigger of further rate rises, and the HY spread as the early warning for the financing of the AI build-out.
Terms and abbreviations
RSP, SPY, QQQ, SOXX, IGV: Exchange-traded index funds (ETFs): the S&P 500 equal-weighted (RSP) and cap-weighted (SPY), the Nasdaq 100 (QQQ), US semiconductors (SOXX) and US software (IGV). The ratio of two ETFs shows which part of the market is performing better.
HYG, LQD, IEF, BIL: ETFs on US high-yield bonds (HYG), corporate bonds with an investment-grade rating, i.e. good credit quality (LQD), seven-to-ten-year Treasuries (IEF) and money market paper (BIL). Together they form the credit ladder from risky to safe.
HY spread: The risk premium of high-yield bonds over government bonds of the same maturity, in percentage points (ICE BofA High Yield Option-Adjusted Spread).
Basis point, US 10y, 2s10s: A basis point is one hundredth of a percentage point. US 10y denotes the yield on ten-year US Treasuries, 2s10s the gap between the two-year and the ten-year yield.
Hyperscalers, capex: The large cloud providers Microsoft, Amazon, Alphabet and Meta, and their capital expenditure on data centres.
Private credit, BDC: Direct lending by funds to companies outside the banking system. Business development companies (BDCs) are listed funds of this kind.
VIX, carry trade: The VIX measures the expected volatility of the S&P 500, derived from option prices. In a carry trade, cheap yen loans are invested in higher-yielding assets.
Methodology: Unless stated otherwise, all price, yield and spread figures come from the proprietary AI Macro Cockpit by the agentic banker and are based on price data from sources including Yahoo Finance, FRED, Deutsche Bundesbank, the ECB and the Japanese Ministry of Finance, with daily closing values up to 29 or 30 October 2026. Quarterly changes refer to the last closing value up to 30 June and up to 30 September 2026. Figures marked as not in the cockpit were also retrieved from Yahoo Finance for this report. Company and central bank information comes from the announcements of the companies and institutions and from Reuters, CNBC and the Financial Times, and was checked against the primary sources in a second review. The scenarios in the outlook for the fourth quarter are the author’s own assessment and not investment advice.
Christian Schablitzki – the agentic banker
Strategy & Management Consultant · Agentic AI expert for financial institutions
More than 20 years in investment banking and derivatives trading, followed by over 10 years advising financial institutions. Currently a Partner at Infosys Consulting in Germany. Certified as Claude Partner – Claude Code (Anthropic), as well as in Google AI, Generative AI Leader (Google Cloud) and IBM RAG and Agentic AI.
LinkedIn profile → · agenticbanker.ai →Keep reading – every 14 days in your inbox.
Capital-markets insights, regulatory updates and AI trends. Concise, well-founded, free.
GDPR-compliant. Unsubscribe at any time.