Deutsche Bank Warns AI Will Split Every Industry Into Adopters and the Dead

Deutsche Bank Warns AI Will Split Every Industry Into Adopters and the Dead

TREE NEWS reports: Deutsche Bank Research has published a 20-page report arguing that the defining market risk of the artificial intelligence era is not a broad unwind of the AI trade, but a “bimodal” split in which every industry fractures into two camps: those who adopt superintelligence and those who are forced out. There is, in the bank’s framing, almost no middle ground.

The report anchors its thesis in the collapse of AI costs. Citing Epoch data, it notes that comparable AI capability has been roughly halving in price every quarter for three years — an annual decline of about 13x. By contrast, computing power took six decades to travel the same distance; AI has done it in five years. When the price of intelligence approaches the cost of computation, firms using superintelligence gain an overwhelming edge over those that do not, making adoption effectively mandatory — much as every industry eventually had to wire itself for electricity.

“In this scenario, waiting is not an option,” the report states. “There is no stable ‘non-adoption’ alternative. A sufficiently high level of capability will force a tipping point of rapid diffusion and competitive repricing. Firms that stick with yesterday’s production model must change, retreat to a different market, or die.”

Why This Matters for Markets

The fastest path to disruption is rarely incumbents transforming slowly; it is “AI-native” entrants replacing them outright, since these newcomers “never have to bear integration costs, because there is nothing to integrate.” Deutsche Bank adds that capital migrating toward early adopters can raise the cost of capital for laggards — in other words, the bond market may decide who is out before customers do.

The report identifies what retains value when intelligence is cheap: proprietary data, operating licenses (especially in heavily regulated fields such as healthcare and finance), sticky customer relationships tied to mission-critical workflows and databases, excludable physical assets like machinery and land, and new demand that emerges as human needs expand. Conspicuously absent from that list: code, analytical work, and most hourly-billed professional services in the white-collar sector. The report expects superintelligence to hit “contestable” industries hardest — large-scale customer service, routine software engineering, and freelance creative work.

Markets have already begun pricing this. In the four years since ChatGPT launched, semiconductors and semiconductor equipment names in the S&P 500 have risen nearly sevenfold, while software stocks are up only about 80%.

Three Investing Rules Are Breaking Down

• Diversification may backfire. Normally, idiosyncratic risk cancels out across a portfolio. In a bimodal world, “one variable sits above everything else, and spreading holdings only spreads exposure rather than diluting risk.” Deutsche Bank suggests classifying holdings as adopters or the displaced, rather than relying on mean positioning, and sizing positions by confidence in that classification.
• Calm prices do not mean low risk. Historical volatility measures realized dispersion, but bimodal divergence is a “jump” — “the two regimes look identical until, at some moment, they no longer are.” The report cites Chegg, which “behaved like an ordinary small-cap until one day in May 2023, when management blamed ChatGPT for slowing growth and the stock halved in a single session.” Ahead of this week’s AI selloff, implied volatility on affected names sat in the 0th to 6th percentile — quiet, until it cracked.
• Buying the dip may no longer be prudent. “A falling price may be signaling that a business model is obsolete, not that an opportunity exists.” For investors who piled into software stocks during this week’s AI correction, that warning is worth heeding.

The Solow Reprieve

Deutsche Bank does leave an exit for laggards. It concedes a reprieve exists, citing Robert Solow’s 1987 quip that the computer age was visible everywhere except in the productivity statistics — which, the bank argues, may still hold in the short run even in a superintelligence era, thanks to the productivity J-curve identified by Stanford’s Erik Brynjolfsson: intangible investment drags on measured output first, then accelerates it.

Regulatory barriers offer a buffer too. Banking, utilities, pharmaceuticals, and defense are “hard nuts to crack,” with existing licenses, high switching costs, and depreciated infrastructure protecting incumbents for “a healthy transition period.” The report offers Pakistan as a cautionary tale: 3.6 GW of imported coal plants locked into decades-long take-or-pay contracts, even as rooftop solar pushed their utilization below 20%. Cheaper technology does not always win quickly.

But the conclusion is that mandatory adoption “is about relative value, not the end of the world.”

Five Early-Warning Signals

The report lists five indicators to watch for whether the bimodal world is arriving: how fast AI accelerates AI research (per METR evaluations); whether AI can complete full work tasks (the Remote Labor Index shows the best AI agents still automate only about 20% of real freelance projects); whether compute remains scarce (one-year GPU rental prices are up roughly a third over the past year); revenue divergence between top and bottom quartile firms on revenue per employee, even as the overall mean stays flat; and whether junior-role hiring comes under pressure and spreads to senior positions.

Deutsche Bank does not predict the arrival of superintelligence; its base case is that AI remains economically important without a regime-breaking breakthrough. But the investment implication holds either way: in a bimodal world, “buy the dip” is a classification judgment, not a free option. The first question an investor must answer is whether a company is an adopter — or one of the displaced.