Insight Partners co-founder Jerry Murdock: AI bubble parallels the dot-com era — and companies on old platforms won't survive the next wave

Summary

Jerry Murdock, co-founder of Insight Partners, built the firm on the proceeds of a fortunate early bet. He and co-founder Jeff Horing backed OpenVision, which merged with Veritas in a 60-40 deal and reached a $30 billion valuation in 1999. That windfall seeded Insight, which shook hands on its founding in August 1994, deployed its first investment in December 1994, and closed its debut fund in July 1996. Early credibility came from an advisory board that included Eric Schmidt (then VP at Sun), Scott Cook (founder of Intuit), and Ray Lane (who succeeded Mike Fields as President of Oracle). The firm now manages approximately $110 billion in AUM.

The Dot-Com Parallel Is Not Reassuring

Insight's experience through the dot-com crash is a direct warning for the current AI cycle. The firm deliberately avoided consumer internet plays, betting instead on infrastructure and applications as the safer path. The strategy failed anyway. Tech dropped 40% in March 2000, and the September 11 attacks compounded the damage in 2001. Insight's 1999 vintage fund returned roughly 1%, which was still enough to rank top quartile — a data point that captures how severe the industry-wide destruction was.

The recovery was grinding. Insight spent 2001 to 2005 working portfolio companies back to breakeven and did not raise Fund V until four years after Fund IV closed. Many firms that launched around the same time, 1995, did not survive.

The central lesson Murdock draws from that period is directly applicable now. Companies built on old technology platforms become unattractive when the next wave arrives. Web-native businesses looked strong before 2008 but lost relevance to mobile after 2010. He argues the same transition dynamic will repeat when the current AI cycle turns.

Where the AI Bubble Could Break

Murdock does not predict a timing, but he identifies three specific pressure points.

First, power. He believes chip performance, driven by Nvidia CEO Jensen Huang's roadmap, is outpacing the electrical infrastructure needed to run the resulting data centers. Existing power capacity is, in his view, already insufficient to fully utilize the chips that have been sold.

Second, enterprise adoption. He argues the AI boom cannot sustain itself without the Fortune 5,000 getting meaningfully on board. A recession or a significant AI security incident could delay that adoption and stall demand.

Third, CapEx overhang. He suggests the hyperscalers may come to regret their infrastructure commitments if token prices fall faster than anticipated, a dynamic that would pressure revenue assumptions underlying those investments.

On private credit, Murdock does not forecast a banking crisis but acknowledges the sector could play a role in the next downturn, particularly if a liquidity crunch materializes against a backdrop of record M2 money supply.

LLMs Are the Old Platform

In March 2025, Murdock hosted a 40-person research conference at the Santa Fe Institute that included senior scientists from major hyperscalers. The takeaway he highlights is that the field is over-indexed on large language models and under-focused on the complexity of human interaction with those systems.

He expects a wave of open-source models, citing DeepSeek's release as an early signal, to erode the commercial value of proprietary LLMs within roughly a year. A multi-model world with abundant free alternatives would significantly degrade the defensibility of companies built on today's foundational model stack. Buying current-generation AI companies as a post-correction strategy is, in his view, a flawed thesis for exactly this reason.

Power Concentration and the Have-Not Problem

Murdock frames the Magnificent Seven CEOs and Sam Altman as the primary architects of the next three to five years of economic and political reality. He adds a handful of chip executives — Hock Tan and CC Wei — and names Fei-Fei Li, Mira Murati, and Lin at Fireworks AI as significant emerging figures.

His longer-range view is that AI wealth concentration will displace immigration and current social debates as the dominant political issue within three years. He cites a proposal he attributes to Jensen Huang and Brad — giving every child $1,000 in stock market exposure — as an early indication that major AI players are aware of the coming distributional tension and are beginning to think about responses.

Portfolio Positioning

Murdock's current personal bets reflect a preference for infrastructure adjacent to the human-AI interface rather than foundational model companies. He names Avan (led by CEO Sadiq Khan, whom he describes as one of the standout executives of the generation, with Vinod Khosla also a significant investor), Fireworks AI, E2B, and Lotus AI.

He is also backing Dynasty, a trust services platform aimed at founders with concentrated equity positions. His argument is straightforward: approximately 80% of venture investments return less than 1.3x, but the 20% that generate meaningful multiples represent founders who should be using trust structures to protect gains and access the QSBS tax exemption, a provision extended by every administration since Clinton. He frames Dynasty as democratizing a tool previously limited to established wealth, applicable to both equity and crypto holdings. Annual startup formation, he notes, has grown from roughly 500,000 in the 1990s to close to 5 million today, expanding the addressable market for that kind of planning.