AI Hype vs Economic Reality: A Wake-Up Call for Businesses and Investors

AI-obsessed bosses are about to get a rude awakening

Fear of losing out has driven big tech to take a trillion-dollar bet it looks likely to lose It's not yet autumn, but the technology industry has just felt the first icy blast of an AI winter. These winters have haunted the field for decades - the periods of wild exuberance such as we see today are quite short. Now, following ominous warnings from Sequoia Capital and Barclays, Goldman Sachs has published a report titled: "Gen AI: too much spend, too little benefit?" If some of their more pessimistic observations come true, we're all in for a bumpy ride. The tech giants are predicted to throw $1 trillion at new data centres in the expectation of fulfilling demand for AI services. But "build and they'll come" supposes a lot of people will soon arrive who want to pay real money. Sequoia's David Cahn highlights the vast disparity between what the industry is splurging on new data centre capacity and the returns that AI can generate for companies like Google and Microsoft. Even by his generous calculations, there's what he calls a "$500bn hole". "To justify those costs, the technology must be able to solve complex problems, which it isn't designed to do", explains Jim Covello, Goldman's head of global equity research. "Replacing low-wage jobs with tremendously costly technology is basically the polar opposite of the prior technology transitions I've witnessed in my 30 years of closely following the tech industry," he adds. MIT economics Professor Daron Acemoglu, a contributor to the Goldman analysis and co-author of the best-seller Why Nations Fail, thinks that the hype obscures the reality of what is really a quite limited technology. "It was always a pipe dream to reach anything resembling complex human cognition on the basis of predicting words," he says. Far fewer jobs are exposed to automation, he thinks - just 4.6pc of tasks can be reliably automated. Over a decade, Acemoglu envisages a mere 0.53pc improvement in total factor productivity. So instead of a "fourth industrial revolution", generative AI's impact more closely resembles that of the Excel macro. Useful but not exactly epoch-defining. "There is pretty much nothing that humans do as a meaningful occupation that generative AI can now do," Acemoglu warns. And he suspects the vast social cost of fraud enabled by AI will bury any advantage in the public's mind. Covello recalls the arrival of e-commerce, which even in its crude infant state, took off like a wildfire. Sellers discovered a global market and that transaction, inventory management and fulfilment could all now be done at a lower cost. The benefits to a business were so emphatic, e-commerce didn't really need any hype. But that kind of impact is startlingly absent from many of the generative AI trials. Businesses report that hallucinations render it useless for many potential use cases: half of those surveyed recently see no upside at all. The financial benefits of implemented projects have been "dismal", one recent survey found. Even when it's reliable, Covello notes, it may not be worth deploying. At Goldman Sachs, generative AI can update historical data in the firm's models more quickly than doing so manually, but at six times the cost. Two more factors are making investors very nervous. Generative AI isn't "scaling" as expected. Nor is it becoming more reliable. Scaling means it gets vastly better the more resources you throw at it, and the industry bet the farm on this: it has no plan B. Bill Gates warns that this assumption is now close to being exhausted and new approaches are needed. Then there are the so-called hallucinations. They're an intrinsic feature of guessing machines that can't say, "I'm sorry, I don't know". So they go and make stuff up instead. If it's a bubble, how did we get here? We have our policy elites and a gullible business class to thank for that. From Sun Valley to Jackson's Hole to Davos, all want to be seen to be at the forefront of a new era. Chief executives leave dazzled by what was always a marketing term - "AI" - vowing to do their bit. Further down the organisational tree, managers want to add it to their resumés, before moving on. Careerists play this game well. "It's a top-down directive," one senior manager told me recently. "We don't actually need it". The influential VC investor Roger McNamee considers it "unimaginable" that big tech will ever see a return on its trillion-dollar bet. "America loves financial manias in a way no other country in the world does," he quipped. But he wonders if, like in previous bubbles, something useful may be left behind. I'm not so sure, and neither is Sequoia's Cahn. The railway mania of the 1840s left behind new infrastructure. The fibre boom of the 1990s connected the world. There was barely a blip before those assets were being used again. But spending $1 trillion on data centres will look very foolish in a few year's time when chips are four generations more powerful. This is capital incineration on a vast scale. Fear of losing out has driven the industry insane. After Nvidia's stock price fell last month, Trade Nation analyst David Morrison warned of "a danger of contagion, with selling spreading to other big tech names". Even for a boom and bust business, this crash may be quite spectacular.

Investors May Be Ignoring Weakening Economy Amid the AI Hype | Investing.com UK

After Sequoia recently suggested that the AI bubble may be reaching a tipping point, Goldman Sachs (NYSE:GS) came out and made a similar point. Tech blogger Ed Zitron took it a step further, concluding "Generative AI is decidedly not the future." Famed tech investor Roger McNamee went so far as to say that, "More evidence supports the view that LLMs are a scam than the Next Big Thing." And venture capitalist, Bill Gurley pointed out that there is much more potential from smaller, less expensive generative AI models than from the massive and obscenely expensive ones Big Tech is hyper-focused on developing. Meanwhile, investors in these "hyperscalers" of generative AI are so enamored they may be overlooking important developments in the real economy.