Why Palantir Declared 'SaaS Is Dead' After 85% Growth

Why a company with 85% revenue growth said “SaaS is dead”

Forbes ran an interesting piece recently. Danny Lutkus, a market strategist at Palantir, made this remark while discussing supply-chain software:

“SaaS is dead.”

It sounds like a bold marketing line. Palantir’s Q1 2026 revenue came in at $1.63 billion (~₩2.2 trillion), up 85% year-over-year, with its U.S. commercial segment growing 133%. Still, one company’s strong quarter doesn’t prove the end of SaaS as a category. What’s worth examining is the specific shift in product and pricing this statement is really pointing to.

When AI agents start handling parts of the work themselves, buyers can evaluate pricing not just by the number of user seats, but by throughput or outcomes delivered. That shift ripples into how products are designed, how pricing models are built, and what kind of workforce a company actually needs.

“The End of the ‘Time Charge’ in the AI Era… What Will Lawyers Sell Now?”The forecast here is that a lawyer’s value will increasingly be defined not by “how long they worked” but by “what judgment and accountability they provided.”lawtimes.co.kr

I actually made the same argument recently, delivering a keynote at an event hosted by Lawtimes, Korea’s legal affairs newspaper. My point was that software once sold per seat or per license is now shifting toward pricing based on cases resolved and the actual value customers realize. I read Palantir’s comment as pointing to that same shift. That said, it doesn’t mean every one of the company’s contracts has already converted to outcome-based pricing.

🔥 What Palantir Actually Said

Seat-based SaaS pricing is built around humans using software. Not all of SaaS runs on seat-based pricing, of course. Users log into a dashboard, organize data, and generate reports. Companies pay a subscription fee for this “access right” scaled to the number of users (seats). More people using the product means more revenue.

The future Palantir describes is different. A growing share of work gets handled directly by AI agents. Agents make judgment calls, invoke tools, move data, and carry a workflow through to completion. Buyers can then compare cost not just against access rights, but against completed outcomes.

The customer breakdown mentioned in the article is worth noting too. Manufacturing is the industry where Palantir has the most commercial customers, and nearly all of its manufacturing clients use Palantir solutions in their supply chains. Domains like supply chain management, compliance, and defense are exactly where agentic systems can demonstrate clear value first. Shortening procurement cycles, auto-classifying exceptions, eliminating human-to-human handoffs — for customers in these fields, what matters isn’t how polished the interface looks, but whether the job actually got done.

This expectation could also reshape how investors evaluate the revenue structures of existing SaaS companies.

📉 What the February 2026 Software Stock Selloff Revealed

Three months before Palantir’s comments, this happened in the market.

In the first week of February 2026, tallies showed that SaaS stocks had lost more than $800 billion (~₩1,100 trillion) in market capitalization. The S&P 500 Software Index plunged 13% in five trading days, and Salesforce’s stock dropped 29% in two weeks. Some called it the “SaaSpocalypse.”

Coverage at the time pointed to a string of newly announced AI agent tools as one catalyst for the decline. Agents were demonstrated autonomously reviewing legal contracts, performing financial analysis, and managing workflows. The market’s fear was simple: if AI reduces the headcount needed to do the same work, revenue tied to per-seat pricing could shrink too. How much headcount actually falls, and how software spending actually shifts, has to be checked product by product, task by task.

There were, of course, counterarguments. SaaStr’s Jason Lemkin said, “This isn’t AI killing SaaS — it’s the market finally pricing in a growth slowdown that’s been building since 2021. BofA analyst Vivek Arya’s team called the selloff an irrational rout that priced in two logically incompatible scenarios at once.” Deloitte offered its own take: “SaaS isn’t dying — it’s evolving into a hybrid model.”

These counterarguments interpret the cause and pace of the decline differently. There’s no consensus that everyone is moving to the same pricing model. Bain argues that while the 2016 downturn was about purchase timing, in 2026 what matters is how customers allocate budget between existing software and AI.

The Financial Times reported that during due diligence on software acquisitions, Bain uses AI coding tools to build similar functionality from scratch. The point isn’t to classify whether a target is a competitor — it’s to gauge competitive advantage by testing how easily the acquisition’s product could be replicated.

💰 The Pricing Model That’s Already Changing

There are also forecasts anticipating a shift in pricing models.

A Gartner forecast cited by Deloitte projects that by 2030, at least 40% of enterprise SaaS spending will shift to usage-, agent-, and outcome-based pricing models. A Bloomberg forecast referenced by RSM likewise suggests that the share of subscription-based pricing could decline over the long term. These projections still need to be validated against real-world contracts, and hybrid models combining usage-based billing with subscriptions are also possible.

Examples of charging by throughput or outcome already exist. The prices below were publicly disclosed at the time, and additional costs may apply depending on product configuration and contract terms.

• Intercom’s AI agent Fin charges $0.99 per customer inquiry it resolves automatically.
• Zendesk: $1.50–$2.00 per ticket resolved automatically by AI.
• Salesforce Agentforce: an example of conversation-based pricing, at $2 per conversation.

Because “resolution” and “conversation” are different billing units, the figures can’t be compared on price alone. Some vendors are pricing a single AI agent at $800–$2,000 a month. The comparison isn’t to a $20-a-month software license — it’s to a $60,000-a-year employee. Some sellers are even framing the pitch around labor-cost savings. That said, this doesn’t mean one agent replaces one employee.

🧭 The Questions Startups Need to Ask Right Now

Here’s Palantir’s argument, translated into terms a startup founder can actually use.

First, you need to define what job your product actually finishes for the customer. A dashboard that helps manage a task has value, but you need to ask whether a competitor could get that same job done with less effort.

Second, you need to settle on a pricing unit the customer will actually accept. If you want to charge based on volume processed or outcomes delivered, you need agreement upfront on what counts as a success and how errors or rework get billed. Basing pricing on outcomes outside the customer’s control is a recipe for disputes.

Third, hiring changes. The center of gravity shifts away from teams optimized for shipping features and polishing interfaces, toward systems thinking, workflow design, and building evaluation frameworks. Agentic products fail differently than traditional software does. You need reliable tool-calling, guardrails¹, memory, and feedback loops. Teams that treat this as an engineering problem rather than a demo problem move faster.

Fourth, the gaps in legacy SaaS are the opportunity. The incumbent SaaS giants have broad product lines, but they’re also locked into rigid pricing structures and seat-based revenue. An agent startup that fully handles one labor-intensive task in a specific industry can make a far more compelling pitch than a general-purpose SaaS product with a polished interface.

Oswarld’s Lens

To be honest, I don’t agree with Palantir’s claim that “SaaS is dead” — but I strongly resonate with the direction it’s pointing toward.

In my years doing technology strategy consulting, I’ve seen this pattern repeat itself. Whenever a new technology disrupts an existing business model, the first thing that changes is the pricing benchmark. When cloud replaced on-premise, the shift went from one-time license sales to subscriptions first. What we’re seeing now — the move from seat-based to outcome-based pricing — follows the same sequence.

The core question, as I see it, is what basis customers pay on. When you move from charging for access to software to charging for processed outcomes, you have to rethink everything from product design to how you define customer success. It’s not just a technical shift — it forces a review of the business model itself.

Palantir’s high growth rate is evidence that demand exists for its business. But that alone doesn’t mean every product needs to move to outcome-based pricing. I think startups need to start by concretely explaining exactly what work their product removes for the customer, and what outcome it delivers.

Closing

• The phrase “SaaS is dead” is an overstatement. But as more work gets handled directly by AI, more products are re-examining usage and outcomes as the basis for billing.
• Charging for outcomes means deciding what counts as an outcome and how to measure it. That reshapes not just the pricing model, but product design, hiring, and even the definition of customer success.
• For startups, this transition is also a chance to stake out territory. While incumbent SaaS companies stay anchored to seat-based revenue, there’s room to move in with outcome-based pricing tailored to specific industries.

The next time you review your product roadmap, ask this question: “What work does our product take off the customer’s plate, and how can we measure that impact?”

References & Further Reading

Primary sources

• Palantir Technologies, “Q1 2026 Earnings Press Release,” SEC Filing, May 4, 2026. : This is where today’s core numbers come from — Palantir’s $1.633B revenue, 85% YoY growth, and more.
• Forbes / Yahoo Finance, “Palantir Says SaaS Is Dead,” May 16, 2026. : This is the original interview where Danny Locktos said “SaaS is dead.”
• Deloitte, “SaaS meets AI agents: Transforming budgets, customer experience, and workforce dynamics,” TMT Predictions 2026, Feb 2026. : This report covers Gartner’s forecast (40% shift to outcome-based models by 2030) and the size of the agent market. It’s the best single source for getting the full picture of the SaaS transition.

Background

• Bain & Company, “Why SaaS Stocks Have Dropped — and What It Signals for Software’s Next Chapter,” Feb 2026. : This report analyzes the structural causes of the February 2026 crash. Its “2016 vs. 2026” comparison framework is striking.
• Financial Times, “Bain tests software takeover targets by vibecoding AI replicas,” Jun 21, 2026. : An article on how Bain is testing whether software solutions can be replicated through vibe coding.
• Jason Lemkin / SaaStr, “The 2026 SaaS Crash: It’s Not What You Think,” Jan 30, 2026. : A counterargument that “AI isn’t killing SaaS — this is the reflection of a growth slowdown that’s been building since 2021.” Worth reading for a balanced perspective.
• RSM US, “SaaS vendors must adjust pricing models as agentic AI transforms the industry,” Mar 2026. : An analysis of the concrete mechanics behind the shift to outcome-based pricing, along with Bloomberg’s forecast that subscription revenue share will fall from 60% to 30%.
• EY, “Agentic AI: How SaaS companies can embrace the future,” Feb 2026. : A report that digs deep into the operational complexity of outcome-based pricing — namely, the question of what counts as an “outcome” in the first place.

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.

Footnotes

1. Guardrails: safety mechanisms set up to prevent an AI agent from taking unintended actions. They can define which tools the agent may access, which actions are forbidden, and which conditions require human confirmation. ↩