NextFin News - Allianz chief economist Ludovic Subran is pushing back against one of the market’s most powerful assumptions: that artificial intelligence will quickly deliver a broad, economy-wide productivity boom large enough to justify today’s enthusiasm. His warning lands at a moment when investors are leaning heavily on AI to explain elevated valuations, aggressive capital spending and a surprisingly durable growth narrative.
The issue is not whether AI matters. It clearly does. The question is whether the speed and scale of the payoff are being overstated. Markets are already pricing in a future where firms squeeze more output from the same labor force, margins expand, and the economy’s speed limit rises without an inflationary cost. Subran’s point is that this can become “exuberance” if the evidence is still too thin to support the confidence embedded in prices.
That distinction matters because the AI debate has moved beyond technology. It is now a macro story, a valuation story and a policy story all at once. If productivity improves slowly, the market may be paying today for gains that arrive in stages. If productivity improves quickly, then the optimism may prove justified, but the economy may also stay hotter for longer than many investors expect. Either way, the consequences reach far beyond the semiconductor trade.
AI’s appeal is easy to understand. It can draft text, summarize data, assist coding, automate support tasks and reduce the time workers spend on repetitive work. Those uses are real, and companies are spending to capture them. But the leap from a useful tool to a measurable lift in aggregate productivity is not automatic. Firms still need to change workflows, retrain staff, clean data and integrate the tools into daily operations. Those steps are slower than the market narrative suggests.
That gap between promise and proof is where exuberance starts to show. Investors often price the destination before the route is visible. The result is a familiar pattern: a genuinely important technology gets a valuation premium long before the economic payoff is visible in the data.
The timing makes Subran’s warning especially relevant. In recent months, AI has been used to justify not only earnings growth in mega-cap technology, but also stronger long-run tax receipts, better fiscal arithmetic and a more resilient growth outlook. That is a large claim to hang on still-early evidence. Productivity is notoriously hard to measure in real time, and the first visible gains often show up as faster service, fewer errors or better quality rather than an immediate surge in output per hour.
The market’s enthusiasm is also shaped by concentration. A narrow group of technology companies has become the clearest beneficiary of the AI buildout, while the broader economy is still waiting for a full diffusion of gains. That leaves investors dependent on a small number of assumptions: that adoption will accelerate, that monetization will hold up and that the infrastructure spending required to power AI will translate into durable returns.
When those assumptions are priced as near-certainties, caution becomes harder to find. Subran’s message is less a rejection of AI than a warning about timing. A technology can be transformative in the long run and still be expensive in the near term if the market treats eventual gains as already secured.
Why The Productivity Story Is Doing So Much Work
The reason AI matters so much to markets is that it touches nearly every major macro variable at once. Higher productivity can support higher earnings, stronger GDP growth and better fiscal outcomes. But it can also alter wage dynamics, the rate path and inflation expectations. That makes AI unusually powerful as a narrative: it explains why stocks can rally even when the macro backdrop is mixed.
It also explains why some economists are urging restraint. Productivity improvements take time to diffuse across an entire economy. They are not captured fully by the first wave of enterprise pilots or by anecdotal reports from tech-heavy firms. A company may report meaningful savings from AI tools, but that does not automatically mean the national accounts will show a similar jump.
“AI will impact the economy, how productivity will play with inflation and much more,” Wells Fargo chief economist Tom Porcelli said in a recent television interview.
That framing is useful because it shows why the AI debate is not just about equity returns. If the technology lifts output without a matching rise in labor supply, it could support demand, profits and public revenues. But if the investment boom required to build the infrastructure is large enough, it can also add to costs, power demand and the strain on financing conditions before the productivity benefit is fully visible.
That is why the word “exuberance” matters. It does not require a collapse in the AI story. It only requires a mismatch between what is known now and what is already embedded in prices. In that sense, Subran’s warning is narrower and more practical than a dramatic bubble call. He is not arguing that AI will fail. He is arguing that markets may be assuming too smooth and too quick a transmission from code to cash flow to macro growth.
The broader history of technology adoption supports that caution. General-purpose technologies tend to work through the economy in phases. First comes the excitement about capability. Then comes the capital spending. Only later come the organizational changes that turn tools into productivity. The market, however, often prefers to collapse those phases into one.
That is especially true when a technology seems to offer a solution to several different problems at once. AI is being asked to explain stronger profits, lower unit costs, faster innovation, better public finances and even more room for central banks to manage growth. That is a lot of weight for one story to carry, particularly when the measured gains remain uneven and early.
The Market Is Testing How Much Of The Future Is Already Priced In
For investors, the central question is not whether AI will be important. It is how much of its eventual benefit is already embedded in the market. That matters because the biggest gains in financial markets often come from changes in expectations, not from the underlying technology itself.
If the current wave of AI spending produces clear margin expansion, the valuations will look more defensible. If the spending produces only gradual gains, then the market may need to re-rate the companies that have benefited most from the enthusiasm. The same is true for the broader economy: slower diffusion would mean less of a near-term productivity shock and more of a drawn-out adjustment.
That is also why the AI discussion has become inseparable from rates. A genuine productivity acceleration could, over time, support faster growth without necessarily forcing inflation higher. But in the short run, the buildout itself is capital-intensive, and the labor market effects are uncertain. That leaves central banks with a difficult signal to parse: some of the optimism is disinflationary in theory, while some of the investment cycle can be inflationary in practice.
There is a second layer of risk in concentration. When a small group of companies carries a large share of the market’s AI exposure, the index may look healthier than the underlying breadth of the economy. That can make the whole market appear more resilient than it is. If the AI narrative weakens, those same concentrated positions can turn from support to vulnerability very quickly.
The larger lesson is that AI remains a real economic force even if some of the current enthusiasm is premature. Technologies rarely need to fail in order for the market around them to become overextended. They only need to arrive more slowly than investors assumed. That is the risk Subran is pointing to.
For now, the most important evidence to watch will come from three places: corporate earnings, capital-expenditure plans and labor-market data. Earnings will show whether AI is converting into revenue and margin gains. Capex will show whether the infrastructure spend remains justified. Labor data will show whether productivity is being realized in the real economy or mainly in market rhetoric.
The answer to the AI debate is unlikely to be binary. The technology can be transformative and still be overhyped in the near term. It can improve productivity and still leave investors ahead of themselves. That is the uneasy space where Subran’s warning sits.
The market may be right that AI changes the economy.
It may still be wrong about how quickly that change arrives.
Explore more exclusive insights at nextfin.ai.
