Steve Hou argues AI is already a real bubble, but that label should not be used to dismiss its macro and market impact. He says the buildout cycle is large enough to matter for GDP, supply chains, and asset prices, and that agentic AI could sharply raise compute demand as adoption broadens and pricing normalizes.
Watch on YouTube ›Get the market thesis, key claims, assets, contradictions, and follow-up questions from any financial video — then unlock a version personalized to your portfolio, watchlist, and favorite speakers.
This Forward Guidance episode is a focused conversation with Steve Hou, a senior quant researcher at Bloomberg, about how to think about AI through a macro and investment lens. Hou says his early bullishness on AI came from seeing it as at least as significant as the internet in terms of buildout cycle, even before he fully believed in its productivity potential. In his view, the key macro point was always that uncertainty about AI’s ultimate capability was enough to trigger a massive capital expenditure cycle, financial repricing, and eventual economic effects. He argues that the AI boom has already had a visible GDP impact through data-center and related infrastructure spending, even if the direct productivity gains from AI adoption are still hard to isolate. He says the U.S. …
AI capex remains the immediate tradable setup: watch for continued strength in data-center, semis, memory, and construction exposure, while headline misses on AI monetization should be treated cautiously if compute demand is still tight.
Over the next few months, the base case is a still-expanding AI investment cycle with pricing and capacity constraints becoming more explicit. The key test is whether agentic usage and enterprise adoption keep compounding fast enough to justify the infrastructure spend.
AI is framed as a genuine bubble with real-world economic consequences, not a false trade. If the thesis holds, the lasting regime shift is that compute and knowledge become priced strategic inputs, changing productivity, labor, and capital allocation for years.
AI is unquestionably a bubble, but being a bubble does not prevent it from having a very large sustained impact.
Hou says bubbles can still matter economically and financially for a long time.
AI differs from the dot-com bubble because it is being adopted and used almost immediately.
He contrasts current AI usage with unused capacity in the internet era.
Agentic AI and AI calling AI tools can multiply compute demand by roughly a hundredfold or more.
Hou argues recursive tool use radically changes demand for compute.
What is your background, education, and what do you do as a quant researcher at Bloomberg?
Steve Hoe is a quant researcher at Bloomberg Indices, which creates systematic investable indices and benchmarks. He has a PhD in macro and financial econometrics from the University of Michigan (2018), previously worked at quant hedge fund AQR Capital for a couple years, and joined Bloomberg in 2020.
Why were you so confident about AI back in 2024 when it was just a chatbot that seemed to hallucinate non-stop? Unpack your AI thesis and how you think about it as an academically trained economist.
As a trained economist, he was naturally skeptical of technology shocks at first. But by 2023 he already saw AI as at minimum as significant as the internet in terms of a buildout cycle. Even if he didn't believe in its full productivity potential, the uncertainty alone meant the investment cycle would have to follow through, creating a massive cycle in real and financial investments. As ChatGPT evolved, hallucinations decreased, consistency improved, and its usefulness in his daily life grew, making him more convinced the trajectory was promising.
Is it useful to look at historical precedents like the internet boom or industrial revolution when thinking about this technological revolution, or is that a red herring?
There are two angles: the economic/financial impact vs whether the technology is the right paradigm. A technology can have massive economic impact without fully working out in the end — the uncertainty alone was sufficient to drive investment. On the second angle, he found the idea of intelligence emerging from brute force signal processing intuitive — similar to how kids learn. He collected charts showing the AI investment cycle is the biggest and fastest capex cycle ever, and believes we may have overlearned lessons from the dotcom bubble, making this cycle more elongated.
Unlock the full claims, asset map, scores, related transcripts, follow-up questions, and AI chat — shaped around your portfolio, watchlist, favorite speakers, and risks.
