A central banker saying what a lot of executives think privately
Central bank governors tend to choose their words on asset bubbles carefully, since the wrong phrase can move markets on its own and get quoted back at them for years. Michele Bullock's answer when asked directly about AI valuations broke from that usual caution, conceding plainly that the sector could plausibly turn out to be a bubble even as the outcome remains genuinely uncertain today. That is a notably direct hedge from the head of a G20 central bank, and it lands with more weight than the usual mix of enthusiastic vendor forecasts and vague analyst caveats that dominate most AI investment commentary right now.
Bullock was careful to frame this as a shared concern rather than an Australia-specific judgment, noting that central bank colleagues globally are watching the same dynamic play out in their own economies. The RBA will address the risk formally in its Financial Stability Review the following week, which means this counts as a preview of an institutional position the bank is actively putting into writing, not an offhand remark made in passing during an interview and then walked back later.
The sequencing problem nobody's pricing in
Bullock's most useful framing for enterprise leaders was not the bubble question but the sequencing issue underneath it. She said there are very few signs yet that AI is actually influencing the supply side of the economy, meaning the productivity gains that are supposed to justify the investment. What is visible right now is entirely demand-side: AI providers building data centers and supporting infrastructure at a pace fast enough to add measurably to inflation.
That creates an awkward timing problem specific to economies already running hot. Australia, per Bullock's assessment, faces excess demand, and AI infrastructure investment is amplifying that pressure well before any offsetting productivity benefit shows up in the data. The gap between capital outlay and realized productivity matches the classic J-curve pattern investors recognize from other infrastructure cycles, railroads and telecom fiber both followed a similar arc, and hearing a sitting central bank governor apply that label specifically to AI is a useful reality check against vendor timelines that imply payoff arrives much sooner than a decade out.
Why this matters more than another vendor ROI claim
Enterprise technology buyers hear a constant stream of ROI claims from AI vendors, most of them impossible to independently verify and none of them accountable to an inflation mandate or a currency's credibility. A central bank governor's assessment carries different weight because it comes from an institution whose entire job is separating real economic signal from investment enthusiasm, and whose credibility depends on getting that assessment roughly right rather than simply telling markets what they want to hear.
Bullock's optimism about AI's long-run productivity potential is genuine; she called it the great white hope for lifting an economy stuck with sluggish output growth. She paired that optimism with an explicit acknowledgment that the payoff window runs closer to a decade than a few quarters. CFOs and CIOs building AI capital expenditure cases for 2027 boards should borrow that framing directly, separating the case for AI's eventual productivity value from the separate and much shakier case for how quickly that value actually materializes.
The labor market angle policymakers care about more
Bullock's broader remarks, which also touched on the RBA's likely direction on interest rates, focused far more heavily on labor market heat than on AI-driven job displacement, a useful signal about where central bank attention currently sits relative to popular narratives about AI replacing workers en masse. The rate-setting conversation is being driven by conventional employment tightness, wage growth, and demand pressure, with AI automation effects still largely absent from the unemployment and participation data the RBA actually tracks.
That absence is itself informative for anyone trying to gauge how far AI's labor market effects have actually progressed. If AI-driven productivity or labor substitution were showing up meaningfully in macro data, it would almost certainly be part of the rate-setting conversation, given how directly either effect would move inflation and employment forecasts that feed straight into a rate decision. Its absence from that conversation supports Bullock's core point: the supply-side effects of AI, whatever they eventually turn out to be, simply have not arrived yet in a form central banks can measure with any confidence.
What enterprise finance leaders should take from this
Bullock explicitly stops short of arguing that AI investment itself is misguided. Her actual claim is narrower and more useful: the timeline embedded in most internal business cases for AI capital spending deserves the same scrutiny a central bank is about to apply publicly in its own review. If a G20 monetary authority is building a formal bubble-risk assessment into its next Financial Stability Review, enterprise finance functions evaluating multi-year AI infrastructure commitments should be running a comparable exercise internally rather than defaulting to vendor-supplied payback periods that rarely account for the sequencing problem she described.
Watch for the RBA's Financial Stability Review the following week for the formal version of this assessment, since it will likely include specifics on data center financing structures and valuation methodology that Bullock's interview comments only gestured at in passing. Other central banks, per her own comment, are watching the same dynamics play out in their own jurisdictions, so this is unlikely to remain an Australia-only conversation for long, and enterprise finance teams operating across multiple markets should expect similar language from other monetary authorities within the next couple of quarters.
