AI’s capabilities are expanding as investors question the cost
AI’s investment case has a tension at its center: the technology is finding new uses, while investors are asking whether the huge build-out can generate enough returns. That debate matters more than any single company’s growth slowdown. New research projects show why demand could reach far beyond chatbots, though they don’t prove that the businesses funding AI will earn attractive profits. At the University of Chicago, researchers mapped 20 million biomedical papers and the 9 million researchers who wrote them. Their AI system tracks where scientists focus, then searches for overlooked links between established fields. The team calls the resulting possibilities “alien hypotheses”: ideas outside the paths researchers would usually pursue. They could point to new ways of treating disease, but each hypothesis still needs laboratory testing. A separate project targets the time required to model climate. Global simulations can keep supercomputers occupied for weeks, limiting the number of scenarios researchers can study. Spherical DYffusion, developed by researchers at UC San Diego and the Allen Institute for AI, simulates 100 years of global climate patterns in 25 hours. The researchers say it runs about 25 times faster than the methods used for comparison, without requiring massive supercomputers. Those results will need validation, but the potential to run more simulations is clear. Here’s the point for investors: AI’s possible uses are widening even as the market reassesses the cost of the infrastructure behind them. A promising research tool doesn’t guarantee paying customers, and a climate model doesn’t establish commercial returns. The capability story and the profit story are related, but they aren’t the same. That tension frames the latest episode of Being Exponential, which examines whether the sell-off signals weaker demand, where opportunity may remain, and what evidence would change a bullish view. Watch the full conversation by clicking the image below:Separate company setbacks from industry demand
Anthropic’s growth is slowing while OpenAI is gaining momentum. That divergence doesn’t, by itself, show that AI demand is fading. The interpretation offered in the episode is that competition is shifting customers among providers as they gain more models and coding tools to choose from. One company can lose momentum while the broader market adds customers. Financing is a more direct risk. Data centers require large upfront investments, while customer payments arrive over time. Borrowing costs, construction delays and access to capital can all affect returns. S&P Global Ratings projects combined hyperscaler capital expenditures will exceed $1.3 trillion in 2027, with external financing taking a larger role. Those pressures won’t affect every company in the same way. The episode argues that concerns about CoreWeave (CRWV) shouldn’t automatically be applied to Nebius Group (NBIS), IREN (IREN) or Applied Digital (APLD). For each business, investors need to examine debt, funding arrangements, customer commitments and the ability to bring capacity online. A full customer pipeline doesn’t remove financing risk; expensive borrowing doesn’t prove that customers have stopped buying.What to watch across the supply chain
Networking companies remain in focus, including Astera Labs (ALAB) and Marvell Technology (MRVL). As AI systems grow, processors need to exchange more data. Extra computing capacity helps less if chips spend too much time waiting for information. Memory is another area to monitor, with Micron Technology (MU) and SanDisk (SNDK) among the names discussed. A low valuation can offer an opportunity, but it can also reflect investor expectations that earnings are near a cyclical peak. Upcoming results need to show whether demand and profit growth can hold up. The warning signs are straightforward: a meaningful slowdown among major AI developers, suppliers missing sales expectations and cutting forecasts, or the largest cloud companies reducing planned investment. Spending outlooks from Meta Platforms (META), Amazon (AMZN), Alphabet (GOOGL) and Microsoft (MSFT) will help test the infrastructure thesis. Stable or higher budgets support it; significant cuts would call for a reassessment. The bullish case rests on expanding demand and continued investment. Earnings season will offer another test. The full episode walks through the setups I find attractive… and the warning signs I refuse to ignore. P.S. My colleague Eric Fry sees a different risk: a widening gap between companies spending heavily to build AI and businesses using it to improve their operations. He argues that the midterm elections could accelerate this “Separation,” with Nov. 4 as a potential turning point. In his new briefing, Eric lays out his case and reveals one AI stock he recommends — free. If you own AI stocks, his argument is worth understanding.About the analyst
Luke Lango is editor of Hypergrowth Investing. His work focuses on early-stage investment opportunities tied to fast-growing industries and major long-term trends. You can learn more about Luke.

This footer identifies the analyst, but it doesn’t add evidence for the article’s investment case. Treat the bio as background on the author, not as support for any market forecast or stock recommendation.