Augusto Marietti, Kong’s co-founder and CEO, said that hyperscaling will be worth it, as the AI bubble may pop at any time. He pointed out that AI firms will ultimately need the massive AI infrastructure they are now investing billions in, just as happened with the U.S. railroad buildout.
Marietti said during an interview with Business Insider that the current builders’ era is a singular moment in which the industry is likely to deploy more capex or capital to enable the AI era. He pointed out that energy-related issues will probably be the main bottleneck stunting the growth of AI, adding that the industry lacks sufficient energy to power all GPUs in the coming year.
As Cryptopolitan reported mid-month, the International Monetary Fund (IMF) also warned that the AI bubble is about to burst, similar to the dot-com era. However, the fund clarified that the collapse of the AI bubble is unlikely to spark a financial crisis.
Pierre-Olivier Gourinchas, the IMF’s chief economist, also believes that AI’s transformative promise may be similar to the internet boom and fail to meet near-term market expectations. He, however, noted that only some equity holders are likely to incur losses in the event of a market correction. Gourinchas explained that the sector’s investments are built by cash-rich tech companies, not debt.
Marietti digs deeper into the 19th-century U.S. railroad buildout analogy
Like OpenAI’s Sam Altman and others, the Kong executive compared the current AI spending to building the U.S. railroad during the 19th century. The AI optimist argues that AI will ultimately transform the economy exponentially; therefore, all enormous expenditures are necessary to lay the groundwork for what is to come.
Altman agreed in August that AI could be in a bubble phase, reiterating what others have warned: that current AI spending is unsustainable. He pointed out that capital expenditure is so high that it almost supports the entire U.S. economy.
However, Marietti pushed the U.S. railroad buildout analogy even further, saying that all railroads were ultimately used, although some were deployed ahead of time. He emphasized that the current AI infrastructure is also deployed ahead of time and will be needed eventually.
“Some railroads were deployed ahead of time, but then all the railroads got used…I think in AI, we’re just deploying ahead of time, and eventually something will blow up for a little bit, but we would eventually need the infrastructure that we’re deploying anyways.”
–Augusto Marietti, CEO of Kong
The Kong co-founder said even a down moment will not stop what is coming in the industry. OpenAI President Greg Brockman previously suggested that soon, everyone will want their personal GPU, and this level of demand would demand massive expansion by his firm and others. Marietti believes that when this happens, the AI infrastructure built today will be put to work in the future, just like railroads deployed nearly 150 years ago are still in use.
Nashawaty says Kong is positioning for an agent economy
During an analysis of Kong’s announcement-packed keynote at the Kong API Summit, principal CUBE research analyst Paul Nashawaty said Kong is positioning itself as the cornerstone of the agent economy’s infrastructure. He disclosed that Kong is currently open-sourcing its Volcano development kit and adding new native support for Model Context Protocol (MCP) agents in Konnect.
Marietti echoed this sentiment, saying that Kong stands firm on the notion that opting out of AI means opting for irrelevance, as agents have become the new surface area for execution. According to the Kong boss, it is more about whether to be part of it or become the Yellow Pages. He added that every business will eventually become agentic; it is not optional.
Marco Palladino, Kong’s co-founder and chief technology officer, also opined that enterprises need a playbook to prevent them from going the way of the phonebook. To that end, Palladino says the industry is establishing standards to make real-time, tool-rich interactions predictable, with the interface shifting from clicks to prompts.
He added that in practice, this means protocols like MCP taking over the role REST or RST (Representational State Transfer) played for APIs. The Kong CTO said he is very bullish about agents and the ability to rapidly generate agents by abstracting away the infrastructure requirements.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It’s free.
