Transcription
Today we're going to talk about a prediction of stunning proportions by Gita Gopinat. But first, the preface.
Reuters has reported that OpenAI is laying the groundwork for an IPO that could value the company at $1 trillion. If this happens, it would be one of the biggest IPOs of all time across the world. Do remember that OpenAI started out as a nonprofit in 2015, and it launched Chad GPT as recently as 2022. Its journey from there to today captures perfectly the stunning rise of generative AI.
And then there's Nvidia. It has just made history by becoming the first ever company to reach a $5 trillion market value. The company and its CEO, Jensen Huang, have become legends in Silicon Valley. Since the launch of Chad GPT in 2022, Nvidia's shares have climbed 12-fold as the AI fever pushes the S&P 500 to record highs. The latest milestone came just 3 months after Nvidia breached the $4 trillion mark.
And again, the debate on whether AI is a bubble has gathered steam, and this is where Gita Gopinat comes in. Now, we have previously done an episode of Everybody's Business on the debate around AI possibly being a bubble, analyzing the arguments on both sides. I'll leave the link in the description box. But today's edition of Eyesight is all about Harvard professor and former IMF executive Gita Gopinat's prediction about the possibility of a bubble and what AI holds in store for the world. Let's dive right in.
We all know that fueled by a wave of optimism around AI, tech stocks have soared high, and that's predominantly in the US. This has drawn inevitable comparisons to another similar euphoria of the past, the dotcom boom of the late 1990s. But according to Gita Gopinat, former first deputy managing director of the International Monetary Fund, this new AI-fueled rally could be setting the world up for something far more dangerous than what we saw in 2000. The cost could be $35 trillion. You heard that right, $35 trillion.
In her analysis published in The Economist, Gopinat warned that while tech innovation was genuinely transforming productivity and industries, the scale of exposure to American equities made the situation precarious. Over the past decade and a half, US households have poured unprecedented amounts of money into the stock market, pushed by the dominance of big tech and consistently strong returns. And it's not just the Americans. Foreign investors, especially from Europe, have also piled in, attracted both by US tech giants and by the strength of the US dollar. This means that if the US market takes a hit, the shock waves will be global.
Now, let's look at what Gopinat's calculations are. She estimates that a market correction similar to the dotcom crash could wipe out more than $20 trillion in US households' wealth. That's around 70% of America's GDP in 2024. Just to be clear, this is several times higher than the losses during the 2000 crash. She warns in her analysis that this kind of shock could slash consumption by 3.5 percentage points, dragging overall GDP growth down by about 2 percentage points. And the global hit could be just as severe. Foreign investors could see $15 trillion in losses, which is about 20% of the rest of the world's GDP. By comparison, during the dotcom crash, the global losses were roughly $4 trillion in today's money. And that, according to Gopinat, shows just how vulnerable global demand has become to US-centric market shocks.
Historically, the world found some cushion in the flight to safety effect. That basically, when any type of crisis hits, the US dollar usually strengthens as investors rush to dollar assets. And this strength helped offset some of the pain of lost wealth abroad. But this time, things might be different. Despite tariffs and expansionary fiscal policy, both factors that should have boosted the dollar, it has weakened against most major currencies. According to Gopinat, that doesn't mean the dollar's dominance is over, but it does reflect growing unease among foreign investors. Many are now hedging against dollar risk, a clear sign of waning confidence.
And this unease is not without reason. Investor trust in American assets depends heavily on the independence and credibility of US institutions, especially the Federal Reserve. But recent legal and political challenges, Gopinat notes in The Economist piece, have raised doubts about whether the Fed can still operate freely. If that confidence erodes further, the dollar's role as a global safe haven could weaken too.
So, what's the way out? Is there one at all? Gopinat argues that the problem isn't just unbalanced trade. It is the unbalanced growth. Over the past 15 years, productivity and high returns have been heavily concentrated in the US, leaving the rest of the world dependent on a narrow base for global growth. If other regions like Europe or other emerging markets can strengthen their own growth engines, it will help stabilize global markets and reduce fragility.
In short, her message is clear. The AI boom may be real, but so are the risks. That's far more wealth at stake today than in the year 2000, and far less policy space to cushion a fall. She warns that if the market crashes this time, it won't be a brief or benign downturn. It could mark a new kind of global economic shock, born out of innovation but powered by overexuberance.
Now, in March 2024, at the AI for Good Global Summit, Gita Gopinat laid out a different lens on the AI risk, one that was rooted in economics and history. She framed a central worry, which was that AI could amplify the next recession by hitting three key channels all at once: jobs, finance, and supply chains. The rest, she warned, could be a deeper, longer crisis than past slowdowns. Here's how.
First, labor markets. She says that history offers a clear warning. Firms often invest in automation when times are good and keep workers on the payroll. But when profits fall, those same firms use automation to cut costs, laying off staff rather than rehiring them. Gopinat cites research showing that nearly 90% of automated-related job losses in the US since the mid-1980s occurred in the first year of recessions. The aftermath of the global financial crisis showed this pattern. Rather than rehiring, many companies automated, and the recovery became a jobless one. And AI changes the game entirely because it threatens a much broader set of tasks. These include higher-skill cognitive roles, not just routine work. Gopinat used clear estimates. About 30% of jobs in advanced economies are at risk of replacement by AI, 20% in emerging markets, and 18% in low-income countries. That means the pool of potentially displaced workers in the next downturn could be far larger than ever before.
A second area of concern was the financial system. Before the AI boom, finance already used automated models for trading and risk management. Now, these models are being rapidly replaced by self-learning complex AI systems whose inner workings can be difficult even for experts to understand. Gopinat pointed to the growing role of automated investment. Assets under robot advice are projected to reach $2.3 trillion by 2028. Now, what is the real problem here? It is that AI models typically perform poorly when faced with novel events, basically conditions different from the data they were trained on. If many models respond at the same time by shifting portfolios towards a safe asset, markets could see a rapid simultaneous move away from risky assets, triggering fire sales, hurting behavior, and broad price collapses. Because the models are often black boxes, diagnosing and managing such dynamics would be particularly challenging for regulators.
And a third area of concern was the supply chains. Companies are increasingly relying on user-friendly AI forecasting tools to manage production and inventories. In normal times, these systems raise efficiency. But when they're trained on the wrong or stale information, they can cause serious forecasting errors. The COVID crisis is a recent reminder of how costly supply chain disruptions can be. Gopinat warns that AI could make those swings faster and more damaging.
But Gita Gopinat, being Gita Gopinat, she doesn't leave the diagnosis unresolved. She proposes three policy directions designed to lower the chances that AI amplifies the next recession.
First, tax systems should not favor automation over people. IMF analysis she cites finds that in several countries, including Germany, the US, the Netherlands, New Zealand, Singapore, and Hong Kong, taxes on software and hardware that substitute for labor are effectively lower than taxes on labor. That creates a bias towards capital-intensive, labor-replacing investments. She's not calling for an AI tax, do keep in mind that. Rather, she urges policymakers to rethink incentives that make automation artificially attractive.
Second, help workers adapt. That means much greater investments in education and training, lifelong learning, and digital infrastructure. Beyond skills, social safety nets must be modernized. More generous and portable unemployment insurance, wage insurance, and better coverage can give workers the time to find suitable new roles and soften wage declines tied to automation.
And third, reduce financial and supply chain amplification risks. Regulators must improve their own skills and oversight, require stronger disclosures about how institutions use AI, and stress-test models for events like no other. Companies should build human oversight and circuit breakers into AI systems to prevent cascading breakdowns. These measures are meant to reduce synchronized reactions that can deepen a downturn.
To see how Gopinat's thinking evolved, it's useful to step back to an essay she wrote in 2023. There she sketched the bigger picture. Generative AI may be a transformative technology on a scale similar to past industrial revolutions. It could spread faster than earlier technologies and reshape economies and societies in profound ways. She cited experimental evidence showing big productivity gains in specific settings. For example, a study she mentioned found that ChatGPT reduced task time by 40% and increased output quality by 18% in an exercise with college-educated professionals. Firm studies suggested AI could increase annual labor productivity growth by two to three percentage points on average, with some results approaching 7 percentage points.
But Gopinat also stressed uncertainty. AI's effects vary by country and occupation. Advanced economies tend to have a large share of jobs in high-exposure occupations simply because they have more professionals and managers. Those workers are both more exposed to displacement and more likely to gain from complementary AI tools. By contrast, emerging markets may be less exposed now but could lose comparative advantages if tasks are reshorced to advanced economies. She also raised the ethical and social risks: the ability of AI to mimic human language, privacy and bias problems, and the danger that a small set of models could magnify her behavior in finance. Her prescription in 2023 was structural: global cooperation, harmonized standards, public investment in AI research, stronger education systems, and social safety nets adapted to changing labor markets.
Gopinat has also spoken about AI in terms of the Indian lens. At the World Economic Forum in Davos, speaking to NDTV, she highlighted India's exposure to AI. Her estimate was that about 26% of India's workforce is exposed to AI, and of that share, roughly 14% stand to benefit from AI adoption, while about 12% face a displacement effect. She emphasized that this is not all doom and gloom. Some jobs would be transformed in ways that make workers more productive. Others will vanish or change beyond recognition. The net effect depends on how governments and societies choose to respond.
Across these pieces—the market warning, the AI for Good Summit speech, the 2023 essay, and her Davos comments—a consistent view emerges from Gita Gopinat. AI brings real benefits and risks just as real. The benefits include higher productivity, better health outcomes, and faster scientific progress. The risks include widespread displacements, fragile financial reactions, and brittle supply chains that could amplify any recession. Gopinat emphasizes that the question is not whether AI will change the world. We all know that it will. It's about whether we'll shape that change in ways that protect workers, stabilize markets, and share the gains broadly. And if the world doesn't, the same technology could amplify the next downturn into something far more damaging.
>> [music]