Daniel Ostry, Roger Vicquéry and Emilio Zaratiegui
There is growing concern among policymakers, international organisations, and even big-tech Chief Executive Officers (exhibits I, II and III) that the current artificial intelligence (AI) boom features valuations increasingly detached from fundamentals. The Bank’s February 2026 Monetary Policy Report noted that an asset price correction is a key risk to the global economy, while the Bank’s July 2026 Financial Stability Report presented a scenario for how an AI correction could unfold. In this post, we study how negative big-tech earnings news transmits to global markets, which informed discussions around this scenario. We find that the effects ripple far beyond tech: equity indices decline, credit spreads widen and the US dollar depreciates. This last result, together with the limited response of Treasury yields, suggests muted flight-to-safety dynamics, unlike other financial stress episodes.
Why would an AI crash transmit differently from a typical financial shock? One reason may be that the current AI boom has been linked to expectations of greater economy-wide productivity, since the tech-sector is increasingly a leading engine of growth in the US. This view has been a key driver of private sector capital inflows into the US even as reserve managers have shied away from Treasuries since 2015. Should expected AI-driven productivity gains disappoint, investors may retreat from both US debt and equity markets. That would be a sharp departure from typical stress scenarios such as the 2008 global financial crisis in which heightened global risk aversion precipitated a flight-to-safety alongside a significant correction in (US) equity markets.
Empirical setup
To study how big-tech news transmits across markets, we construct a daily shock series around the earnings announcements of the Magnificent 7 US tech companies. These tech giants have grown to represent over one third of the S&P 500 index, reflecting, in part, their dominance in developing and scaling innovative technologies – most recently related to AI – which may render news disclosed at their earnings relevant for aggregate productivity as well as the broader index’s profitability. By focusing on narrow windows around their earnings announcements, this approach follows the standard logic of event studies. Our contribution is to construct a new series of Mag-7 earnings shocks capturing news intrinsic to these firms. This procedure helps to partial out the effects of aggregate shocks (eg, monetary policy) that would jointly influence big-tech firms’ earnings and asset markets, although news regarding how Mag-7 firms’ earnings load on these aggregate shocks could still be present.
Formally, our ‘Mag-7’ equity price shocks, defined in Equation (1) below, are constructed as the percentage change in the stock price of a given Mag-7 firm around its earnings announcement. The weight on these changes is given by the firm’s share in the S&P 500 to control for the fact that their collective market capitalisations have risen from 3% to 35% of the S&P 500 over our sample. Since these firms report earnings after markets have closed, we measure the change between the closing price on earnings day and the opening price on the following day. If more than one firm announces earnings on the same day, we add them up.
Chart 1: Mag-7 equity price shocks
Source: Staff calculations.
Chart 1 plots our Mag-7 equity price shocks, which we construct from 2000 to the end of 2025. The shocks mostly lie between -0.5% and 0.5% and are larger after 2020, reflecting the Mag-7’s larger share in S&P 500 in recent years. By way of example, the largest negative shock in our sample occurred on 28 April 2022, when Amazon announced 2022 Q1 earnings that were significantly below market expectations. They also materially revised down their expected 2022 Q2 operating income. Together, these announcements led Amazon’s stock price to tumble by 10% in after-hours trading. This event triggered a significant market reaction. Despite Amazon constituting only about 3% of the S&P 500 at the end of 2021, the index fell by close to 4% within a day. The market reaction was not contained only to US stocks: over the coming days, UK stocks declined as well, credit spreads in both jurisdictions widened and the US dollar depreciated. Below, we establish that many of these dynamics represent systematic patterns following Mag-7 equity price shocks, which we argue may provide a plausible base case for how an AI crash scenario could transmit to financial markets.
Specification and results
Armed with our shock series, we now study how big-tech equity price surprises affect global asset prices using a parsimonious local projection framework.
where corresponds to an asset price of interest in country , including equity indices, effective exchange rates, credit spreads, nominal government bond yields and break-even inflation rates. Controls in include lags of the dependent variable as well as other asset price changes. Our coefficients of interest, , measure the marginal effects of a negative 1% Mag-7 equity price shock on business days after the earnings announcement.
Focusing first on the equity market, Chart 2 traces the reaction () of the S&P 500 and the FTSE 100 to a -1% Mag-7 equity price shock. Importantly, since the underlying Mag-7 stock price changes are weighted by each firm’s size in the S&P 500, a -1% shock mechanically implies a 1% fall in the S&P 500 index on impact. In Panel (a) of Chart 2, we see that the S&P 500 declines by nearly 2% to the shock, implying significant spillovers to other firms in the index as well. These spillovers are of comparable magnitude to the mechanical effect. Relatedly, within two days, the FTSE 100 declines by 1% as well despite no mechanical effect, showcasing that equity indices outside the US, in this case the UK, are significantly affected by news intrinsic to major US tech firms. Both effects are relatively persistent as well.
Chart 2: Equity-market response to a -1% Mag-7 equity price shock
Notes: The chart plots the response of the S&P 500 (left) and FTSE 100 (right) to a -1% Mag-7 equity price shock, as described in Equation (1) and estimated according to Equation (2). 68% (blue) and 90% (light blue) confidence intervals constructed with Newey-West standard errors.
Source: Staff estimates.
Turning to exchange rates, Chart 3 plots the responses of the dollar and the pound nominal effective exchange rates (NEER), which move opposite to one another. We estimate that a -1% Mag-7 equity price shock depreciates the dollar on impact, with a peak effective depreciation of 0.3% a week after the earnings announcement. Sterling, on the other hand, appreciates in effective terms by about 0.5% over a similar timeframe.
Overall, in combination with the earlier equity price responses, this USD depreciation suggests that negative US tech-sector earnings news leads investors to pivot away from US equities, depreciating the dollar by outweighing any other flight-to-safety dynamics into US government bonds. While our identification strategy imposes no restrictions on the underlying mix of structural shocks driving our Mag-7 equity surprises, these results are consistent with investors interpreting negative US big-tech earnings news as downward revisions to future US productivity (Chahrour et al (2024)).
Chart 3: Exchange rate response to a -1% Mag-7 equity price shock
Notes: The chart plots the response of the USD NEER (left) and GBP NEER (right) to a -1% Mag-7 equity price shock, as described in Equation (1) and estimated according to Equation (2). 68% (blue) and 90% (light blue) confidence intervals constructed with Newey-West standard errors.
Source: Staff estimates.
In search of these flight-to-safety dynamics, we study the responses of US and UK 10-year government bond yields. Panels (a) and (b) of Chart 4 show that yields in both jurisdictions fall by a couple of basis points on impact, in line with a muted flight-to-safety or a decline in aggregate demand, before rising, although this rebound is not statistically significant. A rise in yields following a negative Mag-7 earnings news is consistent with new evidence by Andrews and Farboodi (2026) on the release of new AI models, which is understood to be good news about future productivity. Lustig et al (2026) argue that this is the result of US fiscal sustainability increasing in productivity, meaning that US government bond holders are effectively long AI. These offsetting channels may explain the overall muted response on bond markets.
Interestingly, Panel (c) of Chart 4 shows that 10-year breakeven inflation rates meaningfully decline in the US following negative Mag-7 earnings news, potentially indicating important demand-side effects to the shock. As a result, US 10-year real yields rise. The response in the UK, however, is more attenuated on impact, but grows over time.
Chart 4: Government-bond market response to a -1% Mag-7 equity price shock
Notes: The chart plots the response of US 10-year sovereign nominal yields (top left), UK 10-year sovereign yields (top right), US 10-year breakeven inflation (bottom left) and UK 10-year breakeven inflation (bottom right) to a -1% Mag-7 equity price shock, as described in Equation (1) and estimated according to Equation (2). 68% (blue) and 90% (light blue) confidence intervals constructed with Newey-West standard errors.
Source: Staff estimates.
Finally, Chart 5 plots the responses of US and UK credit spread indices to the negative Mag-7 equity price shock. US credit spreads rise on impact, with the effect growing to between 5 and 10 basis points after three weeks. The results are similar for the UK spreads, although the on-impact reaction is more muted. Overall, these findings showcase that big-tech equity price shocks spillover to corporate credit borrowing rates as well.
Chart 5: Corporate credit market response to a -1% Mag-7 equity price shock
Notes: The chart plots the response of US (left) and UK credit spreads to a -1% Mag-7 equity price shock, as described in Equation (1) and estimated according to Equation (2). 68% (blue) and 90% (light blue) confidence intervals constructed with Newey-West standard errors.
Source: Staff estimates.
Conclusion
The current AI equity boom has been characterised by expectations of greater US productivity growth. In this post, we have explored how an AI correction could unfold, leveraging a novel high-frequency equity price shock around the earnings announcements of the magnificent-7 US tech companies.
These shocks provide evidence that what originates in Big Tech does not stay confined to Big Tech. On negative Mag-7 earnings news, equity markets decline, credit spreads widen, the US dollar depreciates, and bond yields are muted, suggesting that a downward revision to US productivity outweighs flight-to-safety dynamics. These findings have two policy implications. First, significant cross-border spillovers to equity and credit markets imply that policymakers should not rely on corrections to elevated US tech-sector valuations remaining confined to the United States. Second, a USD depreciation could exacerbate the macroeconomic effects of these spillovers, increasing the downside risks to an AI crash globally. In particular, foreign economies have historically benefitted from a USD appreciation in times of stress in terms of export competitiveness and a net wealth transfer that helps offset the valuation losses on their USD asset holdings. Accounting for a potential USD depreciation is therefore paramount for policymakers when considering AI crash scenarios.
Daniel Ostry, Roger Vicquéry and Emilio Zaratiegui work in the Bank’s Global Analysis Division.
If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.