OpenAI's Brad Lightcap argues that incumbent software companies are adopting AI faster than Wall Street assumes, making the recent software stock sell-off look like an overreaction.
Wall Street panicked in February, and the casualties were some of the most established names in enterprise technology. Salesforce, Snowflake, and Microsoft saw their share prices plunge between 24% and 30% in a rout that analysts dubbed the "software apocalypse." The trigger was a wave of anxiety that AI tools could let companies simply build their own software, making traditional vendors obsolete. Brad Lightcap, OpenAI's chief operating officer, thinks that narrative is wrong.
Speaking on the "Uncapped" podcast, Lightcap pushed back against the idea that legacy software companies are falling behind. Based on his experience working directly with these firms at OpenAI, he said they are moving as quickly as startups to integrate AI into their products. More importantly, they have something most young companies spend years trying to build: deep, entrenched customer relationships.
"If you're kind of long AI, and long startups, then it might even make sense, maybe, as a contrarian opinion, to be long legacy software, too," Lightcap said, as reported by Business Insider. He added that these companies are not just bolting AI onto existing features. They are rethinking the entire customer journey and looking at how AI can unlock entirely new markets.
The February sell-off was catalyzed by growing excitement around AI agents, particularly after Anthropic demonstrated a tool capable of handling clerical tasks for legal professionals. The implication seemed clear: if AI can do the work that software used to do, why pay for the software? But that logic misses a crucial detail about how enterprise technology actually operates.
Asana's CEO Dan Rogers made this point clear when he addressed the sell-off. His company, which builds work-management software, was hit particularly hard. But Rogers argued that AI agents do not eliminate the need for coordination platforms. Instead, they make those platforms more essential. When you have thousands of AI agents working alongside humans, the coordination problem does not disappear. It scales dramatically. Someone, or some system, has to manage that complexity.
Then there is the economic argument. Anish Acharya, a general partner at Andreessen Horowitz, pointed out in a February podcast that using AI to rebuild core enterprise tools like payroll, ERP, or CRM systems only saves around 10% of costs. The innovation happening with AI models, he suggested, is better directed at genuinely new problems rather than reconstructing software infrastructure that already works well.
What the Market Missed
Nvidia CEO Jensen Huang also dismissed the sell-off logic at a February event, calling the idea that AI will replace the software tool industry "the most illogical thing in the world." His reasoning is straightforward: AI models do not operate in a vacuum. They rely on the APIs, databases, and workflows that legacy software companies have spent decades building. AI will use these tools, not reinvent replacements for them.
The broader context matters here. The global enterprise software market is projected to exceed $300 billion in 2025, driven by cloud adoption and digital transformation efforts that accelerated during the pandemic. Companies like Salesforce, Oracle, and ServiceNow have spent years embedding themselves into the daily operations of millions of businesses worldwide. Ripping out those systems is not a simple procurement decision. It involves retraining staff, migrating years of data, and accepting operational risk during the transition. Most CIOs are not eager to take that leap, especially when the vendors they already trust are actively bringing AI capabilities to the table.
For startups, this creates a nuanced competitive landscape. The window to challenge incumbents with a purely AI-native product is narrower than venture capital hype might suggest. Incumbents like Microsoft have already proven they can leverage their distribution advantage, as seen with the rapid enterprise adoption of Copilot across its Office 365 suite. Salesforce has rolled out its Einstein AI assistant, and ServiceNow is embedding generative AI directly into its workflow automation platform. These are not companies caught flat-footed.
The real opportunity for startups lies in areas where legacy software has no foothold: entirely new categories of AI-native workflows, specialized industry applications, and tools that simply did not exist before large language models made them possible. Competing head-on with Salesforce or Oracle on CRM, using AI as the differentiator, is a tough bet when those companies can match your AI capabilities and offer a mature platform with established integrations.
Lightcap's perspective carries weight because OpenAI sits at the center of this ecosystem. The company has a clear financial incentive to champion startups, yet its COO is publicly making the case for legacy software. That says something about what he is seeing in OpenAI's partnerships and enterprise deployments.
The takeaway for investors and founders alike is to watch how effectively incumbents convert AI features into actual revenue growth over the next two to three quarters. If adoption and retention metrics hold steady, the February sell-off will look less like a warning sign and more like a buying opportunity. If, on the other hand, customers start churning in favor of leaner AI-native alternatives, then the "software apocalypse" narrative may have been early rather than wrong. The data is still coming in, and the next earnings season will be revealing.
