The “AI Killed Software” the narrative has a few very vocal beneficiaries and a lot of quiet evidence against it. The companies that will survive the next five years are the ones that refuse to accept hyperscalers as the new gods.
Whenever I have an endorsement, I like to do my research first, not like a LinkedIn post. I wish more people in this industry would do this because there is a sentiment that we think big numbers are the whole story.
When the Black Death came, people probably thought it was the end. When wars came to our societies, people thought it was the end. Yet, in a strange way, we have the natural power to overcome obstacles and turn change to our advantage.
As AI began to infiltrate our work and then our personal lives, a large group of people declared that “AI will replace humans,” that this technology, while not new, would conquer our minds, hearts, and jobs and take us wherever it wanted.
We are still working; people still write, think, create, build.
💜 of EU technology
The latest rumblings from the EU tech scene, a story from our wise founder Boris and some questionable AI art. Free in your inbox every week. Register now!
Over the past two years, more and more people have been saying “SaaS is dead.” Of course, this statement came from someone’s mouth, from someone who had enough influence to shape the general opinion, and everyone was dressed in black and ready for the funeral.
In August 2024, Klarna CEO Sebastian Siemiatkowski sat in on an earnings call and remarked, almost in passing, that Swedish fintech “Salesforce” closed. It was the next working day.
Klarna would build its AI-driven replacements, a lightweight stack freed from the constraints of traditional enterprise software. The quote moved the markets. Articles followed with headlines about the death of SaaS. Onstage at Salesforce’s Dreamforce, Marc Benioff was asked to respond to a customer who decided the future was AI and the past was its product. He looked embarrassed, by his own admission.
Six months later, Siemiatkowski he calmly clarified what was actually going on. Klarna had not replaced Salesforce with AI. It replaced Salesforce with other SaaS: Deel for HR, third-party tools for CRM, Swedish graph database Neo4j for data consolidation.
Klarna still uses Slack, a Salesforce product. Siemiatkowski himself admitted in X that he was “very embarrassed” by how the story turned out.
“No,” wrote, “We didn’t replace SaaS with an LLM.”
This is the most instructive story in enterprise software in the last two years. The gap between what is said and what is done reveals the mechanics of the whole “SaaS is dead” story. The title traveled. It didn’t fix it.
An industry of analysts, venture capitalists and fund model CEOs has built a year of marketing louder.
Start by asking who is winning from the story of software-as-a-service being replaced by AI, because the answer is surprisingly narrow. Hyperscalers are doing this because AI workloads are driving the $660-690 billion in capital expenditures committed by the five largest US cloud and technology companies by 2026. Future Group analysisnearly twice as much as the previous year.
Base model labs benefit because enterprise software spending directed to their APIs validates otherwise hard-to-defend assessments. OpenAI ended 2025 with approximately $20 billion in annual recurring revenue. Anthropic passed $9 billion in January 2026. These are really big numbers. They also account for about three percent and just over one percent of the hyperscalar capital spent to service them, respectively.
Venture capitalists benefit because their portfolio revaluations depend on the narrative that homegrown AI companies will outperform the incumbents they once funded. And Nvidia, the purveyor and financier of the boom, benefits until it no longer exists.
CEO Jensen Huang in March 2026 confirmed His recent investments in OpenAI and Anthropic will likely be his last. Funding rounds, Nvidia invests in OpenAI, OpenAI buys Nvidia chips, even to the point where the chipmaker is ready to stop calling it a virtuous cycle.
Michael Cusumano of MIT, Bloomberg quotesexpress the arithmetic clearly: “Nvidia invests $100 billion in OpenAI stock, and OpenAI says they will buy $100 billion or more of Nvidia chips.”
You can call it demand. You can also call it accounting.
A 95% figure that should put an end to the hype
The harder question is whether any of this has business consequences. The data here is less generous than the pitch decks.
MIT’s NANDA project in July 2025 “The GenAI Divide: Business in AI State 2025” has been publishedBased on analysis of 150 executive interviews, 350 survey responses, and 300 public AI deployments. Headline finding: 95% of pilots had no measurable impact on the bottom line, despite roughly $30-40 billion in enterprise generative AI spending. Only 5% reached production.
The industry’s response was not to recalibrate. It was to argue that the wrong metric was used. UC Berkeley has issued a rebuttal suggesting that ROI is an “industrial age” measure that is inconsistent with the “cognitive age transformation.”
It says every hype cycle in its late stages, profit is a distraction, what is being built is too big for conventional standards. The same argument was made about WeWork, metaverse and blockchain.
Each time, the underlying assumption was that people with capital and megaphones understood the future better than people who were actually trying to build a business.
MIT found that 5% of successful AI projects share specific characteristics. They are built by specialized vendors, not an in-house effort. They focused more on automating the back office than the sales theater. They are deeply integrated with existing workflows. More than half of enterprise AI budgets were going to sales and marketing tools, which had the lowest ROI.
This is not an enterprise-sweeping revolution. Many companies buy demo-friendly products that don’t generate revenue, while a minority do extraordinary integration work that quietly adds value.
Nondispersive deposition
Still, I have to admit that there are real signs of stress in the SaaS market. About $285 billion in market value evaporated from software stocks in one trading session in February 2026, Wall Street “SaaaPocalypse.”
ServiceNow is down 7%. Intuit fell 11%. LegalZoom lost about 20%. Salesforce is down nearly 30% year-to-date. The business rationale that the cost per seat starts to collapse when one worker with AI tools can do the work of five people is not wrong.
But Bain & Company, looking at the broader record, has suggested a helpful correction: technological transitions are rarely destroyed.
They create heterogeneity. The desktop got rid of the mobile phone. The cloud didn’t kill it in place so much as push it into specific niches. The history of software is the history of layering, not replacement.
SaaS vendors are becoming an agent-orchestration platform. Salesforce has Agentforce. HubSpot has AI tools. Snowflake is partnered with Anthropic. Incumbents are forced to adapt, but adaptation is not death.
IDC’s European experience precisely framed in February: “SaaS is not dead, but it is metamorphosing.”
Evaluation varies towards results. Interfaces are mostly agent driven. But the real business logic remains where auditing, versioning, compliance, and data weighting is concerned. The transformation is real. The extinction event is marketing.
New gods are not new
Every great technology wave creates a brief period in which the companies at its center are regarded as the reinventors of reality. For the cloud, it was AWS. For mobile, Apple. Before that, Microsoft.
The rhetoric around big technologies like Nvidia, OpenAI, Anthropic, Meta and xAI is at the same pace: they are building a new infrastructure of civilization, rewriting how humans work, inevitable. There is a grain of truth in this. Artificial intelligence, and agent AI in particular, is a real technological leap.
The companies most likely to thrive are the ones disciplined enough to recognize the pattern. Every enterprise that survived the dot-com crash, the mobile transition, and the cloud migration did so by embracing what was useful and ignoring the hype, weighing results against costs, refusing to pander to platform vendors.
Losing companies have bought the whole story: customers will wait while they rebuild, the new paradigm will reward early and full commitment.
We reported in February In a pattern seen among dozens of SaaS companies ranging from $20 million to $80 million in ARR: shipping AI features, net income retention is quietly collapsing.
Eighteen months after going AI first, one company saw its NRR drop from 108% to 94% and lost $2.8 million in upgrades, not because the product got any worse, but because everyone was building the future and no one was looking at the present. The AI ​​features were legitimately good. Existing customers were confused anyway.
None of this is an argument against AI. Pprevious AI eras ended with research freezes, shuttered startups, and survivors quietly doing useful work while everyone else claimed to own the moon. This cycle will likely end the same way.
Some hype will be real. There won’t be many revenue projections. A handful of existing “AI-native” startups will become viable businesses. Many will be swallowed or exposed as bandages.
Incoming companies avoid both extremes. They don’t miss the trend, because abandoning AI in 2026 is as serious a strategic mistake as abandoning the mobile phone in 2010. They don’t free up their engineering teams for AI-first rebrands while their existing revenue base is walking out the door. They don’t treat big tech companies as gods, but as what they are: very large for-profit entities with very vested interests in what you believe is the future.
Note that Klarna still charges for SaaS. It still pays OpenAI. This is probably the honest form of the future: not the death of anything, but a quieter restructuring in which operators who keep their feet on the ground are the winners while everyone watches the sky.
The funeral of SaaS was held with a huge crowd. The body is still breathing on closer inspection.
