Deutsche Bank declares ‘the honeymoon is over for AI’ — here’s why
Central banks and investment firms have warned of a sharp economic correction if AI disappoints investors. Now, Deutsche Bank says 2026 will be the hardest year yet for the technology. “AI will survive,” Adrian Cox and Stefan Abrudan, analysts at the investment bank, wrote in a Jan. 20 note titled “The honeymoon is over.” But they said the frenzied sector will be hit by three themes in particular: disillusionment, dislocation, and distrust. “This matters as AI investment and optimism are buoying the global economy, accounting for most of economic and earnings growth in the US last year,” they said. Here’s what Deutsche Bank expects in the year ahead: AI’s impact has been limited Cox and Abrudan wrote that AI will be transformative — but not yet. The technology’s benefits are “more visible to Silicon Valley and to savvy early adopters,” who tend to work on personal projects, they said, not an “average chief executive” looking for revenue and operational improvements. Most companies also don’t have the necessary data to make it work at scale, Cox and Abrudan added. AI tools for coding have improved “rapidly,” however. AI agents, which can complete tasks independently, have been a big talking point recently, but the complexity of integrating them into workflows is often “glossed over,” they said. “For most people this feels less like changing from a horse to a tractor and more like upgrading to a more comfortable saddle,” the analysts said. Bottlenecks galore There are plenty of bottlenecks slowing the AI race, be that with compute capacity, energy, or talent . “AI depends on the most complex supply chain in history and any one of hundreds of thousands of components can derail the process,” Cox and Abrudan wrote. Memory is a particular crunch point as workloads shift from training models to using AI tools, with the analysts referring to it as the “concern du jour” that has “temporarily distracted from the even more fundamental issue of energy supply to data centres. ” At the same time, demand shows few signs of tapering off. Hyperscalers like Amazon, Microsoft and Google pour billions into infrastructure to fuel their ambitions, while smaller companies are rising to challenge Big Tech . “A scramble to assure sovereign AI” is also fueling demand outside the U.S, the analysts wrote. AWS just launched a sovereign cloud service in Europe while Saudi Arabia is looking to set up data embassies. Anxiety is rife “Anxiety about AI will go from a low hum to a loud roar this year. This will be reflected in lawsuits over everything from copyright to privacy, data centre location and protection of young people from chatbots encouraging self-harm or worse,” Cox and Abrudan wrote. Job displacement and misuse of chatbots are also key concerns, they said. The analysts, however, are skeptical that AI will be responsible for all the job cuts companies attribute to it. “AI redundancy washing will be a significant feature of 2026,” they wrote. Market watchers may also be anxious over a shaky geopolitical backdrop to the AI race between the U.S. and China. DeepSeek “showed how necessity could be the mother of invention , squeezing out more value from second-rate chips at low cost than had been thought possible,” the analysts wrote, adding: “There will be an escalating attempt to own the global standard.”
