The 50% Warning That's Shaking Software Stocks

More than 50%. That's how much of current enterprise software could be replaced by AI, according to Arthur Mensch, CEO of Mistral AI. It's a prediction that's sending shockwaves through an industry already grappling with an AI-driven transformation.

The Market Already Got the Message

Software stocks are bleeding. The iShares Expanded Tech-Software Sector ETF, packed with names like Microsoft and Salesforce, has tumbled over 20% this year. Major Indian software giants TCS and Infosys have joined the selloff.

Investors are spooked by a simple question: What happens to the reliable subscription revenue of Software-as-a-Service (SaaS) when AI can do the job faster and cheaper?

"We Build Apps in Days, Not Years"

Mensch's vision isn't theoretical—it's happening now. "With the right infrastructure, enterprises can connect their data to AI systems and create custom applications in a couple of days," he told CNBC from New Delhi.

Think about that timeline. Tasks that required months of development and specialized SaaS vendors five years ago—procurement workflows, supply chain management—are now being built in days by Mistral's customers.

The company has over 100 enterprise customers exploring what Mensch calls "replatforming"—ditching expensive legacy systems bought 20 years ago in favor of AI-powered alternatives that are both more efficient and cost-effective.

Not Everything Gets Replaced

But it's not a complete wipeout. Mensch draws a crucial distinction: "Software focused on systems of records are not going to change." These are the foundational data management systems that actually enable AI to work—they're becoming more valuable, not less.

Rubrik CEO Bipul Sinha echoes this view. "Workflow software could be significantly disrupted by AI," he said, "but system of record or data infrastructure software that enables AI will be positive."

The Speed of Disruption

What's striking isn't just the scale of potential disruption—it's the velocity. Traditional software development cycles measured in quarters are being compressed into days. That's not just faster; it's a fundamentally different business model.

For established software companies, the math is brutal. Why pay monthly subscriptions for generic solutions when you can build exactly what you need, when you need it?

The India Expansion Play

Mistral isn't just making predictions—it's betting on them. The company plans to open its first Indian office this year, partnering with local infrastructure providers rather than building its own data centers.

India's push for locally-run AI models with domestic data storage creates opportunities for companies that can adapt to local languages and regulations. It's a market where global tech giants are all scrambling for position.