← Back
Here is something most people in the software industry have not yet grasped. The business model that created more wealth in technology than any other, the model that turned Salesforce into a $200 billion company, that convinced private equity to pour hundreds of billions into recurring revenue, that made “per seat, per month” the most bankable phrase in capitalism, is breaking apart. Not slowly. Not at the margins. At the foundations.
In a single trading session in early February 2026, roughly $285 billion in market value vanished from global software stocks. Analysts called it the SaaSpocalypse. Within days the selling spread to private credit markets, insurance brokers, and financial services firms. The contagion was so swift and so broad that Macquarie Group’s chief executive disclosed publicly that a quarter of the bank’s balance sheet was exposed to software as a service businesses, and that the firm was “forensically” stress testing what could unfold over the next seven years. Apollo Global Management halved its lending exposure to the sector. Blackstone’s president compared the moment to what happened to the Yellow Pages when the internet arrived.
Something is happening that goes beyond a market correction or a rotation in investor sentiment. Something structural. And to understand it, you need to see a pattern that almost nobody is talking about.
I started my working life as a software developer. My degree was in computer science, with honours in artificial intelligence, at a time when most people had never heard the term. If a department needed a payroll system, someone like me built it from scratch, line by line, over months. There were no products to buy off the shelf. Custom software was the only software there was.
A few weeks ago, I watched a founder in Sydney open a browser, type a paragraph describing the customer management system she needed, and watch an artificial intelligence model generate a working application in under an hour. No development team. No consulting engagement. No licence fee. The application was imperfect, but it was functional. And it cost almost nothing.
Between those two moments lies the entire history of the modern software industry: a fifty year journey from custom to packaged to rented. And that journey, I believe, is now completing a full circle, returning to custom, but at a cost structure that makes the original era look like ancient history.
This paper argues that we are living through the third structural reset in enterprise software. The first moved us from custom development to packaged products. The second moved us from licence fees to subscriptions. The third is unwinding the subscription model itself. Its consequences, for founders, for investors, and for the competitive dynamics of the entire sector, will be unlike anything this industry has experienced.
For companies that own production software today: the valuation multiples that justified your investment decisions, your hiring plans, and your growth strategies have compressed and are unlikely to return to previous levels. A founder who committed to a product roadmap assuming a 12 times revenue exit may discover the achievable multiple is now 6 to 8 times. The expense commitments have already been made. The valuation to justify them may never arrive. Every assumption about the next 12 to 36 months must be tested against a world that no longer operates by the old rules.
For new founders building applications: the barriers to entry have collapsed. AI tools can now generate functional software from a plain English description in hours rather than months. This is liberating for those entering the market, but it is equally lethal for those already in it, because the same tools that empower new entrants also empower your existing customers to build their own alternatives. The window for launching undifferentiated products into uncrowded categories has effectively closed.
For the competitive landscape: the next two years will produce a consolidation wave. Compressed valuations will turn many software companies into acquisition targets. Private equity and strategic acquirers will distinguish ruthlessly between businesses with genuine structural moats, those with proprietary data, regulatory embedding, and deep integration complexity, and businesses that are little more than graphical interfaces sitting atop commodity functionality. The former will attract capital. The latter will struggle to survive.
For investors: the repricing of software as an asset class is structural, not cyclical. The recurring revenue model that attracted hundreds of billions was built on the assumption that software usage scales with headcount. AI agents are breaking that assumption. The investment thesis that treated software subscriptions as quasi annuities needs to be fundamentally reconsidered.
The rest of this paper explains how we arrived here, why the disruption is structural rather than transient, and what it means in practice for anyone who builds, owns, or invests in software companies.
The first reset happened in the late 1980s and early 1990s. For three decades before that, software was something you built because there was nothing to buy. It was expensive, slow, and fragile. Only large organisations could afford it. Then companies like SAP and Oracle offered a seductive proposition: why build when you can buy? A tested product, maintained by the vendor, for a large upfront licence fee plus modest annual maintenance. The total cost of ownership was a fraction of custom development. The trade was irresistible, and within a decade it remade the industry.
The irony, of course, was implementation. “Packaged” software turned out to require enormous customisation to fit actual workflows. Consulting firms, Accenture, Deloitte, the Big Four, built vast practices on the gap between what the vendor promised and what the organisation needed. In Australia, the wave arrived in force through the 1990s. Government departments, banks, and mining companies adopted SAP and Oracle, and a generation of consultants made careers from the complexity of making those systems work.
The second reset arrived at the turn of the millennium. Marc Benioff, a former Oracle executive, proposed something radical: what if you did not need to own the software at all? Salesforce, founded in 1999, offered a model where you paid a recurring monthly fee, per user, to rent access to an application maintained by someone else. No upfront capital. No servers. No upgrade cycles. The customer traded ownership for convenience.
For investors, software as a service created what many considered the most attractive business model in capitalism: predictable, recurring, high margin revenue with the remarkable property of growing faster from existing customers than it lost from departing ones. This was the financial alchemy that attracted venture capital, growth equity, private equity, and ultimately the vast private credit pools that now underpin a $3 trillion market.
But the entire model rested on a single assumption so deeply embedded it became invisible: software usage scales with headcount. More employees meant more seats. More seats meant more revenue. The per seat pricing model was not merely a commercial convenience. It was the structural premise upon which two decades of investment, valuation, and leverage were built.
The concentration of capital behind this assumption was extraordinary. Private credit markets accumulated an estimated 20 to 25 per cent exposure to software. Private equity firms acquired SaaS businesses at an average of 24 times earnings. In Australia, Xero traded at 102 times earnings, WiseTech at 95 times, TechnologyOne at 96 times, Pro Medicus at a staggering 276 times. Even Canva, still private, was valued at $42 billion. These were not speculative startups. They were mature businesses whose valuations reflected an almost religious faith in the perpetuity of recurring revenue.
Each reset solved real problems while creating new vulnerabilities. The first relocated complexity from development into implementation. The second relocated it from capital expenditure into subscription dependence. What neither reset changed was the fundamental bargain: organisations needed someone else to build their software because they could not do it themselves. That bargain is what the third reset is breaking.
Here is what changed. In late January 2026, Anthropic launched a system called Claude Cowork, designed to handle tasks across CRM platforms, analytics dashboards, and customer support without continuous human oversight. Within days, similar capabilities emerged across competing platforms. The market’s response was swift. But the trigger was not merely another product launch. It was the moment investors recognised that three forces had converged to undermine the foundations of the SaaS model simultaneously.
The per seat pricing model was always a proxy for human labour. An organisation paid for Salesforce seats because it had salespeople. It paid for ServiceNow licences because it had IT staff. The seat measured human effort channelled through software.
AI agents break this proxy. When an autonomous system can operate a CRM, generate reports, and manage workflows without human input, the number of seats an organisation needs compresses sharply. A marketing agency that once employed ten people using a complex stack of software tools can now achieve comparable output with two people and a set of autonomous agents. The work still gets done. The subscriptions do not.
IDC, the global technology research firm, has forecast that by 2028, seat based pricing will come under material pressure across the sector, with the majority of software vendors restructuring their pricing around consumption or outcomes. That is not a prediction about a distant future. It describes the next 24 months.
For the first time since the 1980s, it is economically practical for organisations to build their own software. The phenomenon has acquired a name, “vibe coding,” coined by Andrej Karpathy, a cofounder of OpenAI, to describe the use of AI tools to generate functional applications from plain English descriptions rather than handwritten code.
The tools enabling this, including Cursor, Replit, Lovable, and Bolt, have matured rapidly. The economics are measurable. Industry estimates suggest a custom CRM that would have cost $400,000 to build through conventional development can now be produced for approximately $100,000 with AI assistance. Against a SaaS subscription costing $180,000 per year, the custom build breaks even in under seven months, and from that point the organisation owns the asset rather than renting access to someone else’s. The build cost, it should be noted, is only the beginning. Maintenance, security, compliance, scaling, and integration carry their own ongoing burden, and anyone seduced by the economics of the initial build without accounting for the full lifecycle is making a familiar and expensive mistake. But the calculus has shifted decisively nonetheless, and for a growing number of use cases the numbers now favour building over renting.
Mustafa Suleyman, who leads Microsoft’s AI division, has publicly predicted that conventional software applications will be replaced by AI agents and vibe coding in the near term. He is not alone. Multiple major platform leaders and research firms have made similar projections. Whether the exact timelines prove correct is secondary. The direction is not in dispute. The barrier between “describing what you want” and “having a working application” has narrowed to a degree that would have been inconceivable two years ago. The circle back to custom software is closing, but at a fraction of the cost and time that defined the original era.
Perhaps the most significant shift is the migration of value away from the application and toward the data and intelligence beneath it. When a general purpose AI subscription costing $20 to $50 a month can perform cognitive tasks, analysis, strategy, reporting, customer interaction, that previously required thousands of dollars in specialised software, the “application” becomes a thin interface sitting atop a commodity.
Users are discovering they can feed raw data into a powerful AI model and ask questions in natural language, bypassing the structured interfaces that SaaS companies spent decades refining. The application layer, the graphical interface, the predefined workflows, the templated reports, is no longer where the value resides. It has shifted downward, into the proprietary data and the intelligence that interprets it.
It is important to be precise about what is driving the repricing, because the temptation to attribute everything to AI obscures a more complex picture. Three forces are operating at once. First, interest rates remain elevated, compressing the present value of long duration growth assets, which is exactly what high multiple software stocks are. Second, SaaS growth rates have been decelerating every quarter since 2021, a trend that predates the current AI anxiety and that the market had been reluctant to reprice. Third, and most consequentially, the emergence of AI agents and vibe coding tools is challenging the structural assumptions beneath the per seat revenue model. Any one of these forces would pressure valuations. The three arriving together have produced a reckoning.
The WisdomTree Cloud Computing Fund fell roughly 20 per cent in the opening weeks of 2026. The S&P North American Software Index posted its worst monthly decline since October 2008. By mid February, software price to sales ratios had compressed from 9 times to 6 times. Globally, HubSpot fell 39 per cent year to date on top of a 42 per cent decline in 2025. Figma dropped 40 per cent, Atlassian 35, Salesforce 26. ServiceNow fell 28 per cent despite beating earnings expectations for nine consecutive quarters. That last fact is the most telling. The repricing is structural, not performance driven.
In Australia, the pain has been acute. Xero suffered its steepest single day fall since 2013, plunging 16 per cent in one session to a three year low. WiseTech has lost more than 60 per cent from its 52 week high. The ASX technology sector dropped 9.4 per cent in a single session, its worst day since the index was created in 2000. Roughly $50 billion was wiped from the combined valuations of just three companies: Xero, WiseTech, and Pro Medicus. Even Canva’s anticipated IPO, once expected to be a landmark event for Australian technology, now faces a dramatically changed valuation environment.
The contagion then spread into private markets with startling speed. Apollo Global Management’s John Zito told a gathering of investors in Toronto that “the real risk is, is software dead?” Apollo subsequently halved its lending funds’ software exposure. Arcmont Asset Management and Hayfin Capital Management hired external consultants to assess portfolio vulnerability. Blue Owl Capital fell 13 per cent in a single session, Ares dropped 12, KKR nearly 10. UBS modelled a scenario in which US private credit default rates climb to 13 per cent if AI disruption accelerates, more than triple the stress projection for high yield bonds. Morgan Stanley issued a warning that nearly half of outstanding software debt is now rated B minus or lower.
And here in Australia, when Macquarie’s chief executive Shemara Wikramanayake disclosed that Macquarie was “forensically looking” at its SaaS exposures and stress testing what could unfold over the next seven years, it confirmed something the market had been reluctant to say aloud: the institutions that financed the software boom are now questioning whether the boom’s foundations can hold.
Intellectual honesty demands a qualification. Not all software is equally exposed, and the blanket selloff has punished businesses indiscriminately in ways that will eventually create opportunities for those who can tell the difference between structural damage and temporary panic.
There is a credible argument that public SaaS growth rates have declined every single quarter since 2021 and that AI merely gave the market permission to reprice a deceleration that had been building for three years. There is an equally credible argument that deeply embedded platforms, systems like CargoWise in logistics or Xero in small business accounting, possess data moats, regulatory complexity, and integration depth that no vibe coded alternative can replicate. As one experienced developer put it, “shipping a version one is maybe two per cent of the work.” The remaining 98 per cent, maintenance, security, compliance, scaling, integration with dozens of adjacent systems, is where the real costs accumulate.
These counterarguments are real. But they are arguments about degree, not direction. They tell us which companies will be disrupted last, not that disruption will fail to arrive.
The vulnerability runs along a clear spectrum. At one end: horizontal tools with generic functionality, low data barriers, and workflows that can be easily replicated. Project management, basic CRM, simple analytics. These are the most exposed. In the middle: vertical software with domain expertise but limited proprietary data advantages. At the other end: deeply embedded systems of record in regulated industries, platforms with genuine network effects and prohibitive switching costs. These will endure the longest. But “longest” is not “forever.”
The critical question for any software company is not whether AI will reshape this industry. The market has answered that with more than $730 billion in repriced value. The question is where on this spectrum your business sits, and whether the assumptions that underpin your strategy still hold.
Australian technology companies face this reset from a position of particular exposure. The ASX technology sector is disproportionately weighted toward SaaS businesses, many of which traded at extreme valuation multiples that assumed indefinite growth in a stable competitive landscape. That landscape has changed. And when the share prices of large global software companies compress, the valuation multiples applied to small and medium technology companies in Australia follow, always with a lag, but always with certainty.
Many founders base their investment decisions, in new products, in additional headcount, in market expansion, on an expected valuation at the point of eventual exit. That expected valuation is typically anchored to revenue multiples or EBITDA multiples from comparable transactions over the preceding two to three years. Those benchmarks have been destroyed.
The forward price to earnings ratio for enterprise software has halved. Private equity acquisition multiples have fallen from 24 times to 18 and are still compressing. Software price to sales ratios have contracted from 9 times to 6 times. In Australia, the declines in Xero, WiseTech, and their peers are not temporary dislocations. They reflect a permanent recalibration of how the market values recurring software revenue in a world where that revenue is structurally less durable than previously assumed.
Consider the practical consequence. A founder who committed to a product roadmap or a hiring plan predicated on a 12 times revenue exit multiple may discover that the achievable multiple is now 6 to 8 times. That is not a minor adjustment. It is a potential existential miscalculation. The expenses have already been committed. The valuation to justify them has not arrived and may never arrive at previously anticipated levels.
And yet the most common posture among founders today is continuity. Revenue is still growing. Customers are still renewing. The pipeline still looks healthy. Why change anything?
Because the market is no longer valuing your company based on what it earns today. It is valuing your company based on what it will earn in a world where AI agents reduce the seats your customers need, where your customers can build alternatives to your product in weeks rather than years, and where a general purpose AI model costing a fraction of your subscription price can perform many of the tasks your software was designed to facilitate.
I have spoken with founders in Sydney and Melbourne over recent weeks who are only now beginning to ask these questions. One described the experience as “like discovering the map you’ve been navigating by doesn’t match the territory anymore.” Another, running a mid market vertical SaaS business, told me his board had not yet discussed what happens to their renewal assumptions if even 15 per cent of customers begin exploring agent based alternatives. These are not hypothetical conversations. They are happening now, in real companies, and the founders having them earliest will be the ones best positioned to act.
The acid test for every founder is confronting what they do not know they do not know. The competitive dynamics of the next 12 to 36 months will be shaped by capabilities that did not exist twelve months ago. Business models and operating assumptions must be tested against scenarios that fall well outside historical experience. This is not a task that can be performed from within the organisation alone.
For those considering an acquisition, the repricing will create opportunities not seen since the global financial crisis. Compressed valuations in the mid market will produce potential targets at prices that would have been unthinkable twelve months ago. But the difference between a bargain and a value trap has never been more consequential. Acquirers must separate businesses that are temporarily discounted by market sentiment from those that are genuinely structurally impaired.
For those considering a sale, the window for premium exits at the multiples that characterised the period from 2021 through 2024 has closed. It may not reopen. Companies that can demonstrate defensible data positions, structural resilience, and the capacity to integrate AI into their value propositions will command fundamentally different valuations from those that cannot. The preparation required to achieve that differentiation takes time. It must begin now, not in the weeks before a transaction process starts.
The broader pattern is already visible. Industry analysts expect a consolidation wave through 2026 and 2027, with software companies becoming acquisition targets for larger platforms and private equity firms seeking to consolidate assets for their data value. For Australian technology companies in the $5 million to $50 million revenue range, the strategic question of whether to acquire, sell, or restructure to compete independently in the new landscape is no longer theoretical. It is immediate.
Every 15 to 20 years, the enterprise software industry undergoes a structural reset in how software is built, deployed, priced, and valued. Each reset destroys enormous value in the incumbent model while creating enormous value in the successor. And each time, the organisations that navigated the transition were not necessarily those with the best technology. They were those with the clearest understanding of what was actually happening to their industry, and the willingness to act on that understanding before the market forced their hand.
I have lived through all three of these resets. I wrote code before packaged software existed. I watched the SaaS revolution transform an industry I had grown up in. And I am watching the AI inflection reshape it again, at a speed and scale that makes the previous transitions look leisurely.
The programmer writing COBOL in the late 1970s could not have imagined software as a service. The SaaS founder in 2010 could not have imagined vibe coding. The lesson is not that prediction is impossible. It is that the capacity to adapt to structural change matters far more than the ability to forecast it. That capacity does not come from technology strategy. It comes from something deeper: the organisational clarity, the structural discipline, and the external perspective to question assumptions that have underwritten every decision to date.
The question for every Australian software company today is not whether this transition is real. The market has answered that question with $730 billion in repriced value. The question is whether your organisation is built to navigate it, and whether you are testing your assumptions with the rigour that this moment demands.
In a world consumed by the next disruption, the next pivot, the next exponential curve, the companies that endure are not the ones that chase volatility. They are the ones that build the foundations to withstand it. And building those foundations begins with an honest conversation about where you stand.
This paper is provided for informational purposes only and does not constitute financial, investment, or professional advice. The views expressed are those of the author and do not necessarily reflect the views of Cube Capital or its affiliates. Readers should seek independent professional advice before making any investment or strategic decisions.
Sell-side technology M&A advisory. Sydney, Australia.
© 2026 Cube Capital Pty Ltd. All rights reserved.
Cube Capital provides M&A advisory services to wholesale clients only. This website does not constitute financial product advice. Advisory services are provided in accordance with applicable Australian financial services law.