Will Generative AI Follow the Internet Bubble? Insights from MongoDB’s CEO

MongoDB CEO Dev Ittycheria says that, if we view the timeline of the internet era, commercial applications of artificial intelligence can be traced back to 1996.

Speaking at a Goldman Sachs event, Ittycheria told investors that the technology sector is still in the early stages of an AI rise and that AI will transform business and enterprise technology.

He emphasized that AI is not a question of "if" but "when" it will appear, comparing today’s world to the period just after Netscape launched, when the internet was still static and nascent.

Gartner’s 2024 Emerging Technology Hype Cycle warns that generative AI (GenAI) is approaching a "trough of disillusionment," with expectations high but long‑term impact expected.

ServiceNow CFO Gina Mastantuono recently acknowledged a possible consumption bubble around GenAI, while Equinix’s survey found many enterprises still struggling to find business cases for GenAI projects a year after the hype began.

Three Use Cases of Generative AI

Ittycheria identified three current GenAI use cases: chatbots, research summarization, and automation.

Chatbots

Chatbots can handle simple social interactions, but when placed on corporate sites they often fabricate answers or return nothing, making them less useful than well‑crafted FAQ sections.

Research Summarization

Large language models (LLMs) tend to hallucinate, so research summaries can be inaccurate; a recent Australian study found LLMs still lag behind humans in document summarization.

Automation

Machine learning, a subset of AI, can automate low‑value tasks, but the limits of what LLMs can reliably do remain unclear, and no current application promises revenue comparable to the massive AI investment levels.

Ittycheria draws a parallel between today’s AI surge and the late‑1990s internet bubble, noting that while Netscape was only a few years old in 1996, the subsequent dot‑com crash saw the Nasdaq lose two‑thirds of its value within a year.

He cautions that people often overestimate the short‑term impact of new platforms while underestimating their long‑term influence, suggesting the AI journey may be rocky and that significant downturns could take up to five years to manifest.

In conclusion, as a veteran internet user from 1996, the author argues that the current AI excitement resembles the late‑1990s era, and we may be closer to a 1999‑style environment than to a stable, mature market.