Marc Andreessen made a dire software prediction 15 years ago. Now it’s happening in a way nobody imagined

On August 20, 2011, legendary venture capitalist Marc Andreessen published a blog post—and an accompanying essay in The Wall Street Journal—that would become the sacred texts of the Silicon Valley bull run. Titled “Why Software Is Eating The World,” he argued that the global economy was undergoing a “dramatic and broad technological and economic shift” and that software companies were poised to take over large swathes of the industry.

Fifteen years later, in February 2026, Andreessen’s prophecy has been fulfilled in a manner that even the biggest bulls failed to predict. Software did indeed eat retail (Amazon), video (Netflix), music (Spotify), and telecommunications (Skype) just as Andreessen predicted, but the market got a $1 trillion shock in February because something was eating software itself. That something, of course, was artificial intelligence.

Morgan Stanley’s software analysts, led by Keith Weiss, offered a “gut check” this week in a major research note, arguing that “AI IS software” but also that software is growing so all-consuming that it is indeed starting to eat work itself. Andreessen’s a16z has a core strategy of investing across enterprise software, including cloud, security and software-as-a-service (SaaS), but the $1 trillion-plus selloff dubbed the “SaaSpocalypse” cuts to the very heart of that model. Andreessen looks like he was more right than he knew about software eating the world.

The original prophecy

To understand the severity of the current shift, one must look back at the skepticism Andreessen was fighting in 2011. Following the trauma of the dot-com crash, he declared that the stock market “hated technology.” While Apple was trading at a price-to-earnings ratio of just 15.2x amid immense profitability, investors constantly screamed “Bubble!”

Andreessen claimed that companies like Amazon and Netflix were not merely speculative bets, but “real, high-growth, high-margin, highly defensible businesses” building a fully digitally wired global economy. He correctly identified that Borders was handing its keys to Amazon, that Netflix was eviscerating Blockbuster, and that “software is also eating much of the value chain of industries… in the physical world,” such as automobiles and agriculture.

For a decade and a half, he was right. The “creative destruction” he invoked—citing economist Joseph Schumpeter—decimated legacy incumbents and minted trillions in value for software insurgents. However, the AI revolution of 2022 onward and the SaaSpocalypse of 2026 suggest that the cycle of creative destruction has arrived at the doorstep of the software industry itself. Morgan Stanley’s Weiss wrote of a “Trinity of Software Fears” currently driving down stock multiples by 33% that cut to a fundamental questioning of the software business model.

While Andreessen saw software disrupting industries, Morgan Stanley sees AI disrupting labor itself. The analysts note that generative AI expands the capabilities of software to “contextually understand unstructured data,” such as emails, PowerPoints, and verbal communications. This unstructured data represents over 80% of information in organizations today.

Previously, software required a human operator to input and manipulate this data. Now, Wall Street fears that software can do it all by itself. “Generative AI represents a continuing expansion of what types of work and business processes software can now effectively automate,” Weiss wrote, revisiting his team’s initial estimate that enterprise software’s total addressable market could grow by $400 billion by 2028. Three risks put that in question, principal among them that “as GenAI automates a broader swathe of work, the increasing productivity gains will result in a reduction in the number of employees necessary to execute those tasks.”

If software allows a company to cut its staff by half, it also cuts the number of software subscriptions it needs by half. After software ate the world, then, it appeared to start eating the revenue of its creators by eating the jobs of its users.

The ‘do-it-yourself’ threat

Andreessen predicted in 2011 that “software programming tools… make it easy to launch new global software-powered start-ups,” viewing this as a boon for entrepreneurs. Today, however, investors are beginning to view this democratized ease of creation as a threat to established software giants.

One of the primary fears cited by Morgan Stanley is the rise of “do-it-yourself” (DIY) software. This is colloquially known as “vibe coding,” where a user will ask the AI to code things in line with a certain vibe that they are going for. As AI code generation tools drastically lower the cost and skill required to write code, there is a growing fear that “companies will choose to develop more software themselves” rather than paying for expensive third-party vendors.

Furthermore, there is the looming threat of “model providers”—the creators of frontier AI models—rendering traditional applications obsolete. The fear is that an AI agent could act as an “intelligent user interface,” pulling together data and tools to automate workflows on the fly. In this scenario, the distinct “app” disappears, replaced by a single, omniscient model that serves as the operating system for the entire enterprise.

Will incumbents strike back?

Like other analysts (and several rattled SaaS executives), Morgan Stanley argued that the market’s reaction is overblown, echoing Andreessen’s 2011 sentiment that investors were ignoring “intrinsic value” right in front of them. The bank suggested that the “bear case arguments around GenAI appear to give too little credence to the ability of incumbent software vendors to participate in this innovation cycle.”

Andreessen once warned that “incumbent software companies like Oracle and Microsoft are increasingly threatened with irrelevance.” In 2026, however, Morgan Stanley identified Microsoft, alongside Salesforce and ServiceNow, as the “Best Athletes” positioned to win. True, Salesforce is in “the eye of the storm” in terms of workflows expected to be disrupted by GenAI. But Weiss said that incumbents such as Salesforce are successfully pivoting to become “fast followers,” integrating AI to solidify their moats rather than losing them. For example, Salesforce has seen its AI-related annual recurring revenue surge 114% year over year.

Zooming out, Morgan Stanley said it sees a “path of innovation that actually looks relatively familiar”: a combination of improving productivity, better use of tools to automate functions and software value “predicated on labor displacement.” The difference now is the rapid pace of innovation compared to prior cycles and better tools on the market. It looking closely at Amazon Web Services and the shift in the early 2010s toward cloud computing. Even with the 33% pullback for software’s equity value/sales multiple since October, the group is trading about 15% above the beginning of the cloud era.

In a sequel of sorts to Andreessen’s famous essay, his own firm has released new thought leadership (as it does quite regularly). Steven Sinofsky of a16z dismissed the idea of the “death of software” earlier this month, arguing that “AI changes what we build and who builds it, but not how much needs to be built. We need vastly more software, not less.” He offered five predictions, including more software being made with new tools in a vastly more sophisticated way, but also an admission that “it is absolutely true that some companies will not make it,” and endless invention and reinvention is the way of capitalism. A look back at the Fortune 500 archives show that is undoubtedly the case.

In his 2011 essay, Andreessen concluded with optimism, calling the software revolution a “profoundly positive story for the American economy.” He acknowledged challenges, specifically that “many workers in existing industries will be stranded on the wrong side of software-based disruption.”

That is where things may be scarily different this time. Even if software finds a way to recover its multiple and continue its upward trajectory, analysts are increasingly seeing a future of growing GDP and productivity without nearly as much human labor involved. Michael Pearce of Oxford Economics recently joined a group including Bank of America Research and Goldman Sachs warning that the U.S. economy is nearing a point where it won’t need to keep creating new jobs to keep increasing output.

Google DeepMind’s Nobel-winning co-founder, Demis Hassabis, recently told Fortune Editor-in-Chief Alyson Shontell that he was excited about the world of “radical abundance,” even a “renaissance” ahead, but there will be a 10- to 15-year shakeout until we get there. That could come as the economy figures out what to do with all the labor that software has eaten.

This story was originally featured on Fortune.com