On November 12, as Web Summit entered its final day in Lisbon, Nvidia announced that co-founder Jensen Huang had resigned as CEO. Chief financial officer Colette Kress took over on an interim basis. Nvidia shares fell roughly 15% at the open before recovering part of the loss. For a conference already debating whether AI investment had run ahead of demand, the market reaction gave speakers a concrete question to consider. How much of investors’ confidence in the AI boom rested on one company and its founder?
The assumptions behind AI infrastructure spending
At a panel moderated by Euronews, Adam Kobeissi, founder of capital markets publication The Kobeissi Letter, argued that losing Huang would weigh on Nvidia investors over the next few years. He remained optimistic about the company. James Galbraith, an economist at the University of Texas at Austin, took a longer view. A company of Nvidia’s size can outlast its founder, he argued. A potentially more consequential change would be a Chinese competitor able to deliver comparable technology at a lower price.
That possibility matters well beyond Nvidia. Microsoft, Alphabet, Amazon and Meta are committing hundreds of billions of dollars to AI infrastructure. A cheaper generation of competing chips could make AI services more affordable while making it harder for operators to earn the returns they expected on equipment they have already bought. Demand for AI could keep growing even as some of today’s investments lose value.
Galbraith saw an advantage for China in its ability to coordinate research, manufacturing and deployment. The US, by comparison, can mobilize private investment quickly, while its government often takes longer to agree on policy. He tentatively favored China over the long term, an assessment that puts the pace of industrial deployment alongside access to capital as a factor in the AI race.
Why a technology panel turned to US debt
When the discussion turned to what might trigger a financial crisis, both speakers pointed to government debt markets. The US spent about $1.1 trillion on net interest in the fiscal year ending in September, equivalent to roughly 19% of federal revenue. Kobeissi saw persistent borrowing costs as a risk that could build over the next decade or two. He stopped short of predicting an imminent repeat of 2008. Eu
For the AI industry, the connection is the time between paying for infrastructure and earning money from it. A data center requires capital years before its operator knows how much customers will pay to use it. If financing stays expensive while competition pushes the price of computing down, a facility can attract customers and still fall short of the return that justified building it.
What this means for businesses building with AI
For a company integrating AI into daily operations, the same economics apply on a smaller scale. Consider a document-processing system. The model may cost only a few cents per document, but the business still pays for retries and staff time when an extraction needs correction. The figure worth tracking is cost per correctly processed document. Comparing that figure with the existing workflow gives the team a basis for deciding whether to expand.
It also gives engineering teams a practical way to respond when the market changes. A representative set of real documents, with known correct results, lets a team test whether a lower-priced model delivers the same accuracy and how often a person still needs to intervene. That evidence remains useful whichever chipmaker leads the market or whoever runs it.
We have helped 20+ companies in industries like Finance, Transportation, Health, Tourism, Events, Education, Sports.