The SaaS Apocalypse: Who Lives and Who Dies

A 30-year veteran of growth equity investing sits down to explain why everything his industry taught a generation of VCs is being rewritten. Jerry Murdock, co-founder of Insight Partners (managing $90+ billion), doesn’t use the word “disruption” as abstraction. He has portfolio companies that have already declared Cursor obsolete, agents that are writing code autonomously, and a clear thesis about what comes next: the orchestration layer, ASIC chips, and a labor market crisis that could decide the 2028 presidential election.

The Tsunami Analogy

Murdock’s framing is precise. A tsunami is harmless at sea; it’s only dangerous when it hits the beach. The AI wave isn’t a single product. It’s autonomous agents. And we’re in the anticipatory period, the water pulling back before the main wave hits.

The pre-peak waves? Claude’s legal plugin tanking legal tech stocks. The security tool dropping CrowdStrike and Cloudflare. The COBOL modernizer cratering IBM. These are tremors. The real impact comes when autonomous agents reach enterprise adoption, which Murdock estimates at roughly a year out.

“What we have with the tsunami happening is a wakeup call to move to higher ground. Don’t get caught on the beach when the damn thing hits the beach.”

His advice to portfolio companies: bolting on AI isn’t enough. You need to be AI-native. The companies that survive will be the ones building for a world where agents, not humans, are the primary buyers and users of software.

Cursor Is Already Obsolete

The most provocative claim in the episode comes from Murdock’s own portfolio. Companies like E2B, Eventual, Lotus AI, Get Dynasty, and Aven are all using OpenClaw, NanoClaw, or homegrown autonomous agents to write code. Their view, as directly reported to Murdock: Cursor is obsolete.

Murdock qualifies this: Cursor’s team is smart, well-funded, and has time to pivot to autonomous agents. But the gap between an AI coding assistant and an autonomous agent is fundamental. An assistant helps a human write code. An agent writes code on its own, spins up 100,000 sandboxes in seconds, and makes decisions probabilistically.

One telling detail: E2B builds sandboxes that respond in 80 milliseconds versus the industry standard of 400 milliseconds. Murdock asked why. The answer: “The agents notice it, and that’s what matters.” When your customer is a machine, human-perceptible latency thresholds are irrelevant. Machine-perceptible ones are everything.

The Coming Claw Stack

Murdock draws a historical parallel to the LAMP stack (Linux, Apache, MySQL, PHP) that enabled the web explosion in 2004-2005 after the post-9/11 slump. He predicts an equivalent “Claw stack” for autonomous agents, likely emerging from the open-source community.

The architecture he envisions: an orchestration layer that triages workflows across multiple LLMs. Expensive Claude tokens for complex reasoning tasks. Cheaper open-source models (DeepSeek, Llama) for simpler workloads. The agents themselves will make these routing decisions probabilistically, running workloads across 10 different sandboxes with different libraries to see which performs best.

This leads to a critical prediction: the rise of ASIC chips. As open-source models get routed by orchestration layers, you’ll want to put the model directly on the chip. ASICs will be cheaper, more tunable for specific workloads than Nvidia’s general-purpose GPUs. Murdock believes this is exactly why Nvidia acquired Groq: not just for different workload types, but to ensure CUDA remains viable as the ASIC explosion arrives.

Meta’s decision to say “no” to Jensen and bet on ASICs is, in Murdock’s view, a tell. The companies that see the trajectory are already positioning for it.

Agents as Employees: The Pricing Revolution

Murdock describes a paradigm shift in how software gets bought and sold. Today, all software is purchased by human beings. Soon, software will be purchased and consumed by agents. An autonomous agent becomes an employee: you give it credentials, identity, and authorization. You review its decisions like you’d review an employee’s work.

This changes pricing models fundamentally. Seat-based SaaS pricing makes no sense when the “user” is an agent that might spin up 100,000 instances in seconds. The future is consumption-based: you open it up, start using it, and pay based on what you consume. Docker, one of Murdock’s portfolio companies, is already moving dramatically toward this model.

For SaaS companies, the survival question becomes: is your software built for autonomous agents? If agents have a reason to use it and it’s valuable in an agent-driven workflow, you’re fine. If you’re still building for human buyers, you’re on a 6-18 month clock.

“Autonomous agents are going to write software faster, cheaper than any human being on the planet. And they already are.”

Systems of Record: Mount Everest or Melting Ice?

On whether systems of record survive, Murdock offers a conditional answer using Carta as an example. If stock tokenization comes and uses Carta’s cap table infrastructure, Carta becomes infinitely more valuable. If tokenization bypasses Carta with a new system of record, Carta’s value collapses. It’s pure execution dependency.

For Salesforce specifically, Murdock uses a proxy metric: watch the health of the dozens of companies built on top of Salesforce’s system of record (like Veeva). If those companies start getting knocked off one by one, Salesforce’s underlying value erodes. But Salesforce is “Mount Everest,” an 8,000-meter peak that won’t melt overnight. The question isn’t survival; it’s valuation.

His broader insight on software valuation: new technologies typically expand markets before contracting them. Don’t expect software to fall off a cliff. But fear changes prices, and then the underlying quality of the business changes based on how well management adapts.

The Labor Crisis and the 2028 Election

Murdock makes a direct political prediction: AI-driven job displacement will be a major issue in the 2028 presidential election. It could decide the outcome.

The displacement sequence he sees: first, companies stop hiring the next person (junior developers, executive assistants, marketing coordinators). The existing employees aren’t fired first; the pipeline dries up. Then, depending on how fast agents evolve, actual displacement begins. Small-medium businesses move first because a single secretary or customer support rep makes a huge proportional difference. Enterprise is last, as it has been throughout the AI revolution.

His prediction: some form of minimum viable income (essentially UBI) will become at least a ballot question within 2.5 years. No administration will want to preside over 10-15% unemployment, so they’ll restructure unemployment programs to include retraining paths.

The optimistic spin: the jobs going away are mostly suffering jobs, people struggling to pay healthcare while doing data entry. If those people can move out of expensive cities, perhaps to rural areas with technology-enabled small-scale agriculture (he cites a Wyoming program leasing ranches to veterans who can now run operations with 1-2 people instead of eight), the quality of life could actually improve.

Intuition vs. Wishful Thinking

The second half of the conversation shifts to investing philosophy, and Murdock’s most counterintuitive admission: 80% of his investments returned less than 1.3x. The 20% that made real returns made all the money and all the impact.

His framework for decisions: every decision has two components, logic and intuition. Autonomous agents will handle logic brilliantly. They won’t have intuition anytime soon. The human edge remains in that gap.

But he learned the hard way to distinguish intuition from wishful thinking. His biggest mistakes came from backing entrepreneurs he liked as people, ones who were smart but too comfortable, not obsessed enough, without a chip on their shoulder. The most successful founders are “challenged socially.” Peter Thiel’s observation resonates: the best founders make you feel uncomfortable.

“How do you know it’s really intuition and not wishful thinking? Intuition was almost never wrong. But what I was wrong about was me thinking it was intuition. It was nothing but wishful thinking.”

Timing Is the Single Most Important Factor

Across 30 years of venture data, Murdock identifies one variable that correlates most strongly with fund performance: vintage timing. A 2005-2006 fund caught the mobile revolution when Steve Jobs hit 10 million AT&T subscribers in 2008. A 2009 fund missed being early in Twitter, Facebook, and Uber.

His conclusion: right now is absolutely the best time to start a new fund. “Humans are no longer going to be the decision makers about software. It’s going to be autonomous agents.” A sea change this fundamental gives new entrants an advantage over incumbents who are already rich and moving slowly.

Murdock himself hasn’t left the game despite retiring from Insight’s day-to-day. He’s made over a hundred personal investments, each brought to him by someone he trusts, each born from trying to help someone think through a problem.

The Twitter Bet

The most revealing anecdote: in 2009, Murdock invested in Twitter when it had 30-something employees and zero revenue. His own fund questioned the decision. What he saw: “The idea of the status update for the world was so brilliant, so over the top. It was far better than anybody on the team’s ability to execute against it.”

He put his entire reputation on the line, convinced partners Jeff Horing and Devin Parekh to join, and closed the deal in less than 30 days. It became a watershed investment for Insight, opening doors to consumer investments and demonstrating that the quality of an idea, not the current execution, can be the deciding factor.

Closing Notes

This conversation has the quality of a veteran telling you what the weather looks like before a storm. Murdock isn’t an AI researcher or a founder building agents. He’s a capital allocator who has survived three decades of technology cycles, and he’s speaking with unusual directness about what’s coming.

• The most important signal isn’t Murdock’s words; it’s his portfolio companies’ actions. When multiple AI-native startups independently declare Cursor obsolete and switch to autonomous agents for code writing, that’s market intelligence from the front lines, not speculation from the sidelines.
• The ASIC prediction deserves more attention. If the orchestration layer successfully routes workloads to purpose-built chips, Nvidia’s moat isn’t CUDA itself but whether CUDA can run on ASICs. The Groq acquisition may be the most strategically important move in that chess game.
• Murdock’s “agents as employees” framework changes every SaaS metric that matters. If the buyer isn’t a human, then seat-based pricing, user experience design, sales cycles, all of it needs to be rethought. The companies that figure out agent-native product design have a window that may not last long.
• His distinction between intuition and wishful thinking is the kind of insight that costs billions to learn. It’s a warning to every investor who backs founders they like rather than founders who make them uncomfortable.