The software industry is being devoured by its own creation: AI
The Software as a Service (SaaS) model, which for years dominated enterprise software thanks to subscriptions and its high margins, is changing at an accelerating pace. Artificial intelligence, which was initially integrated as an enhancement within these programs, has now become a factor that reduces their value. Many of the functions for which companies paid expensive licenses can now be performed directly with AI systems, without the need for multiple applications or long-term contracts.
AI is destroying the software business
As AI automates tasks and makes it possible to create tailored solutions, traditional software is losing its competitive edge. This is forcing the sector's major companies to rethink their business model, since the basis of SaaS —recurring payment per user— no longer makes sense when a single AI tool can replace several programs and numerous jobs.
Some analysts describe this process as a form of technological self-cannibalization. The artificial intelligence that is now weakening the SaaS model does not come from external players; it has been developed by the software companies themselves, which for years benefited from traditional subscriptions. By integrating increasingly advanced automation and software generation capabilities, these companies are helping to reduce the value of their own products, accelerating a shift in the model in which software ceases to be the center of the business and is replaced by more general and flexible AI systems.
The beginning of the collapse of the SaaS paradigm
The first sign of this situation began to appear in 2025, when the broader market showed notable resilience, the software sector experienced one of its most violent contractions since the 2008 financial crisis. The S&P North American Software index plunged 15% in a single month, wiping out approximately 2 billion dollars in market value, at a time when the overall S&P 500 was showing growth of 9%. This 24-point performance gap points to a systemic loss of confidence in the sustainability of recurring revenue models based on per-user subscriptions.
The market logic is clear: sif a general-purpose AI can perform specialized tasks such as legal research, document drafting or data analysis, the "sticky" nature of high-priced SaaS subscriptions evaporates. Companies that for a decade saw SaaS as an invincible fortress of 80% gross margins now face an environment where software has become a commodity.
The drop in valuations has hit even the most established names. Companies such as Intuit, Atlassian and HubSpot showed sustained weakness throughout 2025. The specialists' concern is that AI does not only assist the human worker, but replaces the "seats" those workers occupy. This fundamental shift in the unit of value of software is forcing a total reassessment of what constitutes a sustainable competitive advantage.
Vibe Coding: The disintermediation of software engineering
One of the deepest attacks on the software ecosystem comes from the democratization of its creation through "vibe coding." The term refers to the use of natural language to build complex applications, removing the technical barrier of traditional programming languages. The ability to generate highly customized solutions has allowed companies to consider building their own internal tools instead of paying for generic, expensive SaaS solutions.
There are documented cases of companies terminating contracts worth more than 350,000 dollars with established vendors such as Salesforce, opting to develop custom tools on AI agent platforms such as Base44. A former Amazon executive reported building a complete CRM system over a weekend, a task that traditionally would have required months of development and massive investment in platforms such as Nice or Salesforce. In fact, Salesforce itself is guilty of this kind of progress, since the company has an open source AI model designed to facilitate vibe coding, known as xLAM.
Of course, this technological shift means that the features, the user interface (UX) and the dashboards that once differentiated a software product can now be replicated almost instantly by AI agents. Thus, the "stickiness" of traditional software, based on implementation complexity and user training, is being eroded by conversational interfaces that eliminate the learning curve.
The impact on the gaming industry
But the most recent attack can be seen in the gaming industry, where the tension between AI and human talent is more palpable. In 2025, 90% of game developers used some type of AI in their workflows, a massive adoption driven by the need to mitigate production costs that now exceed 660 million dollars for AAA titles such as Call of Duty or Grand Theft Auto.
However, this adoption has come with a wave of unprecedented layoffs. One in ten developers lost their job in 2024-2025, with the narrative area the hardest hit at 19% of layoffs. Even though executives see AI as an efficiency tool, 30% of developers believe generative AI has a negative impact, citing ethical concerns, intellectual property theft and a drop in the quality of the final content.
Transforming the sector
But this has not stopped companies such as NetEase and Krafton from linking their record revenues directly to the deep integration of AI in development and advertising optimization. At the same time, venture capital in the gaming sector has become extremely selective, preferring «gametech» startups that develop foundation models for asset and voice generation, rather than traditional development studios.
And that situation could be pushed even further with the arrival of models such as Google Genie 3, capable of taking a prompt and making a fully navigable virtual world. Of course, the limitations of the current model make it impossible to apply to massive video games. The idea that a world only lasts 1 minute, and then that whole world «is forgotten», makes this unplayable, and if presented to the average gamer, that game would most likely end up in the gamers' trash bin.
Genie 3: Creating dynamic worlds that you can navigate in real-time
But the big picture is that AI is also gradually transforming the sector, and although it has very clear limitations now, in about 5 years those limitations may have been overcome.
The capital exodus
Faced with all this, venture capital in the sector is reconfiguring. In 2025, 58% of all capital invested in VC-backed companies in the United States went to AI companies. Investors are prioritizing hardware and «Deep Tech» over pure software. For the first time in recent history, the robotics and deep tech sectors have surpassed software in deal volume for four consecutive quarters.
This trend reflects a shift in risk perception. Traditional software, once considered low-capital and highly scalable, is now seen as a sector vulnerable to rapid commoditization. By contrast, hardware infrastructure and foundation AI models are perceived as the new defensive moats that protect capital.
«Big Tech» investment (Microsoft, Alphabet, Amazon, Meta) in AI infrastructure reached 320 billion dollars in 2025, a massive increase from the 230 billion of 2024. This level of spending on hardware underlines the belief that technological sovereignty in the AI era will depend on control over compute and data, not simply over end applications.
But despite this, we must be clear about one thing: the software industry is not disappearing, rather its model is being devoured by its own invention, and it will have to reinvent itself.
After all, the ability of artificial intelligence to automate software creation, personalize processes and act autonomously has transformed the «tool» into an «agent». And companies that continue selling access to tools based on user interfaces and per-seat licenses face a permanent decline in their margins and global relevance.