NextFin News - Global central bankers left their annual Sintra gathering with a message that was much clearer than their usual rate-path debate: artificial intelligence is no longer just a productivity story, but a financial-stability question. Across panels and side conversations at the European Central Bank’s forum in Portugal, AI was described as a force that could lift growth, rewire labor markets and improve lending and supervision, while also creating new vulnerabilities in asset prices, credit, cybersecurity and power demand.
The timing mattered. The Bank for International Settlements used its Annual Economic Report 2026, released on 28 June, to warn that the current AI investment boom carries the hallmarks of an overheated cycle. The BIS said the five largest hyperscalers are set to spend over $1 trillion on AI-related capital expenditure from 2025 through 2026, that those commitments are outpacing earnings and free cash flow, and that some firms are turning to debt to bridge the gap. The central bank of central banks also said a major equity-market correction could have larger economic consequences today than in the past, a warning that gives the Sintra discussion a more urgent edge.
The combination of those two developments is what made this week notable. The policymakers in Sintra were not merely talking about whether AI will make economies richer over the long run. They were also talking about whether the current phase of AI spending, market enthusiasm and infrastructure buildout is already creating a balance-sheet risk that could travel from stocks to credit markets and then into the broader economy. That is a very different question from the one investors usually ask when they focus on model quality, chip demand or software adoption rates.
Torsten Slok, chief economist at Apollo Global Management, captured that duality with a line that became one of the meeting’s recurring refrains:
“If AI overdelivers, it will impact financial stability. If AI underdelivers, it will impact financial stability.”
That framing is not a rhetorical flourish. It is a policy problem. If AI works better than expected, valuations, capex and competitive pressure can all intensify at once. If it disappoints, the unwind can hit the same channels in reverse. In both cases, central bankers are left with the same concern: the system may become more fragile even if the technology itself is economically useful.
That is why AI kept surfacing in discussions that, on the surface, should have been about more traditional central-bank subjects. It came up in bank lending because lenders need to assess rapidly changing business models and infrastructure-heavy borrowers. It came up in supervision because advanced systems can make decisions that are difficult to explain or audit. It came up in labor markets because productivity gains could change hiring, wage growth and the distribution of income. And it came up in power demand because the AI buildout is also an electricity and data-center story.
Why The Financial-Stability Lens Took Over
The central bankers’ unease rests on a simple point: AI can be economically positive and still financially destabilizing. That is especially true when the story is being financed by a narrow group of giant firms whose spending plans are large enough to move capital markets. The BIS said the current AI surge resembles earlier innovation waves in one important respect — expectations are running ahead of visible returns. When that happens, the risk is not just disappointment. It is over-commitment, leverage and a repricing that can move through the funding system.
The BIS also said policymakers need to pay attention to vulnerabilities in the financial system, strained public finances and supply shocks that are still playing out. The AI buildout fits into that broader warning because it links a speculative narrative to real-world capital expenditure, debt issuance and power demand. If the economics of the buildout weaken, the effect is not limited to tech stocks. It can show up in bond spreads, collateral values and credit appetite.
That is why central bankers are likely to keep treating AI as a cross-cutting macro issue rather than a narrow technology trend. A lot of the market conversation has focused on whether AI can lift productivity enough to justify current spending. The policy conversation is different. It asks whether the path to those gains is so concentrated and capital intensive that the transition itself becomes destabilizing.
The distinction matters because the market can survive optimism, and it can even survive disappointment, but it does not always survive both at once. If AI spending remains strong, valuations can stretch further and infrastructure spending can keep climbing. If the returns disappoint, the correction can be sharp because expectations are already embedded in the price of the ecosystem. Either path can create financial-stability pressure.
“This is the biggest time of consequence to each of our economies, I think, in our lifetime,”
Kevin Warsh said in Sintra, describing the AI moment as early in its development. The comment was notable not because it predicted a specific outcome, but because it reflected how central bankers are now thinking about AI: as a force with macro implications large enough to sit beside inflation, labor supply and financial regulation.
Why This Cycle Feels More Fragile Than Earlier Tech Booms
The current AI cycle differs from earlier technology booms in one key respect: the financing is easier to see, and potentially easier to stress. The five largest hyperscalers are not simply spending from operating cash flow. The BIS said their capital expenditure from 2025 through 2026 will exceed $1 trillion, and that those commitments are outpacing earnings and free cash flow. Once debt enters the picture, a market revaluation stops being a pure equity story and becomes a credit story too.
That creates a transmission channel central banks care about. Equity losses can reduce household wealth and business confidence. Credit losses can tighten lending standards and raise borrowing costs. When the same sector sits at the intersection of high valuations, heavy capital needs and debt financing, the potential spillovers widen. The BIS warning that a major equity-market correction could have larger economic consequences than in the past is therefore not generic caution. It is a specific signal that the AI boom may have more macro reach than many investors assume.
The labor-market dimension adds another layer. Central bankers do not only worry about whether AI makes firms more efficient. They also worry about how the gains are distributed. A technology that improves output but concentrates benefits in a small number of firms or skilled workers can widen economic disparities, which in turn can affect consumption, wage bargaining and political pressure. The Sintra discussions reflected that broader concern: AI may support productivity, but it can also alter the composition of jobs and the pace at which workers move between them.
That is also why supervisors care about explainability. If AI tools start affecting credit decisions, underwriting, fraud detection and risk management at scale, regulators need to understand how those decisions are made. Black-box systems may be fast and efficient, but they make oversight harder. In financial stability terms, opacity is not a side issue. It is a risk factor, because it makes problems harder to identify before they become systemic.
The electricity and infrastructure side of the story matters for the same reason. AI is not only a software deployment; it is a physical buildout. Data centers, chips, cooling systems and power grids all need financing. That means the AI cycle spills into utilities, industrial supply chains and local infrastructure, broadening the number of entities exposed to the same investment wave. The more sectors that depend on the same growth narrative, the more damaging a reversal can become.
Central bankers are not saying the boom must end. They are saying the boom is now large enough that a change in expectations would matter beyond the technology sector. That is the crucial shift in tone.
What Central Banks Are Really Preparing For
The practical response from central banks is likely to be surveillance, not intervention. They will probably spend more time examining AI-related lending, collateral values, capital-market exposures and infrastructure financing. They will also keep asking how quickly AI adoption is changing productivity, wages and inflation dynamics, because those outcomes feed directly into monetary policy assumptions.
The BIS’s message is useful here because it links the AI boom to a broader set of pressure points. The same report that highlighted AI also warned about high public debt, fragile liquidity in core bond markets and the risk that inflation could become ingrained if expectations de-anchor. That means AI is not being discussed in isolation. It is one of several forces making the global policy environment harder to manage.
For markets, the implication is straightforward. AI is still a growth engine, but it is no longer only a growth engine. It is also a source of balance-sheet risk, funding risk and potentially macro risk if the current spending cycle overshoots its eventual payoff. That does not mean investors should treat the theme as broken. It means the valuation debate has become inseparable from the stability debate.
That is the deeper takeaway from Sintra. Central bankers are not trying to decide whether AI is good or bad. They are trying to decide whether the path to the good outcome is safe enough for the system to absorb. In their view, the answer is not yet obvious.
The next catalysts will be the next round of earnings, capex guidance, and signs of whether AI spending is still being funded comfortably by cash flow or increasingly by debt. Central bankers will also watch whether the technology’s productivity gains become visible in the data or remain mostly a story told by corporate planners and investors. If the gains stay abstract while the leverage grows concrete, the policy concern will intensify.
The central lesson from this week is that AI has crossed from an innovation debate into a financial-stability debate. That makes it one of the few stories that can simultaneously raise hopes for growth and fears about the system that finances it.
Explore more exclusive insights at nextfin.ai.
