Imagine a world where the dazzling promise of artificial intelligence isn't just exciting innovation—it's potentially inflating an economic bubble that's ready to burst. That's the unsettling reality economist Jason Furman dives into, and it's enough to make anyone sit up and take notice. Stick around as we unpack the most alarming highlights from Bloomberg's thought-provoking Q&A on the AI bubble, giving you the full picture while clarifying the complexities for those just getting started in this fascinating yet daunting topic.
Jason Furman stands out as one of the more approachable figures in the economics world. He's a respected professor at Harvard (check out his profile at https://www.hks.harvard.edu/faculty/jason-furman), and he previously headed the White House Council of Economic Advisers under President Barack Obama (you can read more about that nomination here: https://obamawhitehouse.archives.gov/blog/2013/06/10/president-obama-nominates-jason-furman-chairman-council-economic-advisers). Back in October, he appeared on the podcast of conservative New York Times opinion writer Ross Douthat (find the episode details at https://www.nytimes.com/2025/10/23/opinion/ai-bubble-economy-bust.html) to discuss this very issue. Interestingly, that earlier chat only touched on a single concern. So, what could explain the recent surge in worries?
Right off the bat, Furman reveals his primary anxiety: 'I'm more concerned about the financial valuation bubble than a technological bubble.' At first glance, this sounds like he's drawing a subtle distinction—perhaps the underlying tech is impressive, but companies might still be priced way too high, making the latter the true threat. But as he elaborates, it becomes clear we should be equally vigilant about both sides of the coin.
To put it simply, for AI companies' stock prices to make sense, you need two key ingredients: groundbreaking technology that performs exceptionally, and the ability to turn that into real profits. The risks that could deflate these valuations include hitting a point where further improvements yield less and less benefit—think of it like the 'diminishing returns' in farming, where adding more fertilizer stops boosting crop yields dramatically. Additionally, Furman points out that the 'scaling laws'—those patterns we've seen where bigger models lead to better results—might not always translate to economic gains. For instance, just because your computer's microchip doubles in speed doesn't mean you'll suddenly type up Word documents or reply to emails twice as quickly. In many cases, that extra power sits unused, like extra capacity underutilized in our devices. And this, he warns, could very well happen with AI, even if its scaling patterns hold true.
But here's where it gets controversial... This description eerily mirrors one of the year's biggest AI headlines (read the Gizmodo article here: https://gizmodo.com/it-took-just-24-hours-of-complaints-for-openai-to-start-bringing-back-its-old-model-2000640912). When OpenAI unveiled GPT-5 in August, users flocked to it, but many felt the upgrades didn't justify the hype. Sure, it might have been a minor advancement technically, but ChatGPT enthusiasts didn't perceive enough value to outweigh the frustration—it felt like chatting with a friend who suddenly got less engaging, not more. In essence, the model had 'excess capacity' without the warmth or insight people craved. Critics might argue this is just a blip in consumer tech, but Furman uses it to illustrate broader economic pitfalls.
If you're scratching your head trying to differentiate between a 'tech bubble' and a 'valuation bubble,' don't fret—you're in good company. Bloomberg's interviewer, Shirin Ghaffary, admits she was puzzled too. Furman expands on this, noting that beyond stock prices, there's a tangible reality: billions upon billions of dollars annually pouring into data centers, energy systems, and related infrastructure. This isn't vaporware; it's 'actual, real activity,' he emphasizes, drawing parallels to the dot-com boom when massive investments in internet infrastructure laid the groundwork for future growth. Yet, he adds a crucial caveat:
The real concern arises if all this investment fails to boost productivity. Currently, AI is primarily stimulating the 'demand side' of our economy—in other words, it's creating demand for goods and services, but not yet fully enhancing overall production.
He reinforces this by stating:
We don't have a US economy operating at full capacity. It's more like it's running on a single engine right now.
These insights are crucial for understanding how mainstream economists view AI today. Labeling AI as 'demand-side' might seem odd—who among us actively seeks out AI in daily life? If you're like me, it's probably minimal or nonexistent. But let's clarify: Picture the global economy as a vast, somewhat vacant Home Depot store. AI acting on the demand side means it's a colossal shopper, snapping up drills, bags of cement, and ladders to keep the shelves stocked and the lights on for now. It's a voracious customer sustaining the business.
However, AI can't forever be the sole big spender in this metaphorical Home Depot. The structures it builds with those purchases must generate enough economic buzz to attract even more customers—ideally, more than ever before—to shop and create their own things. Without that, the store risks emptying out, leaving everyone high and dry.
And this is the part most people miss... Returning to that ChatGPT fiasco earlier this year, while it's a prime example of consumer AI interactions, Furman doesn't see it as the kind of application that will propel widespread economic growth. He also dismisses the notion that AI will massively disrupt jobs through efficiency gains, calling it a low-risk scenario. 'History shows that predictions of job obliteration from new tech have consistently been wrong,' he notes, pointing to past technological shifts that ultimately created more opportunities than they destroyed.
Instead, Furman's forecast paints a hazy picture of AI's future role in bolstering the economy:
People in the real world are deliberate and intricate in their approaches—they might discover one practical application this year, another the next, and thoroughly test it in multiple ways before implementation. Various businesses, industries, and sectors will adapt at their own paces, not in some overnight revolution. Of course, this is just my best educated guess, with the huge disclaimer that anything is possible.
Your reaction to this might range from reassured to skeptical, but here's my take: Essentially, Furman is suggesting that AI will prove genuinely valuable to unknown individuals at unspecified times in the future. It's not an entirely far-fetched outlook—think of how smartphones evolved from novelty to necessity over years. The truly alarming aspect, though, is that for our economy to thrive, this prediction absolutely must come true.
Now, let's stir the pot a bit: Is Furman downplaying the bubble risks to avoid panic, or is his cautious optimism the reality check we need? Could AI's 'excess capacity' be a sign of overhyped tech, destined for a bust like past bubbles? Or will gradual adoption save the day, proving the skeptics wrong? What are your thoughts—does this vision of AI align with yours, or do you see a different outcome brewing? Jump into the comments and share your perspective; we'd love to hear agreements, disagreements, or fresh angles!
