Mark Cuban thinks Lovable and Replit can outlast the AI labs

Mark Cuban argued at the RAISE Summit in Paris that AI coding tools like Lovable and Replit can outlast big AI labs by bundling services such as incorporation and payments, creating a defensible workflow. Cuban, an investor in Lovable, countered fears that labs like Anthropic or OpenAI could easily replace these startups by adding similar features, emphasizing the value of localized data and integrated business operations.

At RAISE Summit in Paris on Wednesday, “Shark Tank” investor Mark Cuban made the case for AI coding tools like Lovable and Replit, Business Insider reports https://www.businessinsider.com/mark-cuban-ai-labs-replace-lovable-replit-vibe-coding-2026-7 . They can hold their own against the big AI labs, he argued. The reason: they now bundle services a raw language model does not. Cuban is a Lovable investor, so he has a stake in the answer. But his point lands on the central worry hanging over “vibe-coding” startups. If Anthropic or OpenAI can simply add a feature, what is left to defend? “When we question a lot of the front ends like Lovable, Replit, etc, we wonder if the foundational models, the big guys are going to just replace them,” Cuban said, in conversation with Lovable CEO Anton Osika https://thenextweb.com/news/lovable-ceo-europe-ai-confidence-problem-talent-silicon-valley . “But what you’re saying is you have a base of data that is localized.” From code tool to “AI cofounder” Osika’s pitch is that Lovable has outgrown the label of AI software engineer. Founders can now incorporate a company and wire up payments inside the platform, not just generate an app. “People saw us as an AI software engineer, the product,” Osika said. “Now what we’re increasingly seeing is that people see Lovable as their partner, as literally their AI cofounder.” That framing matters commercially. A code generator competes on model quality alone. A place where you run the whole business resists replacement. Lovable has leaned on that breadth as it scaled to a reported $500M in annual revenue https://thenextweb.com/news/lovable-build-economy-500m-arr-vibe-coding with a small team. The lab-shaped shadow Recent months sharpened the fear. After Anthropic shipped its Opus 4.6 model earlier this year, the pull was obvious. Some founders and developers said on X that they had dropped their Cursor https://thenextweb.com/news/spacexai-cursor-joint-ai-model-launch and Lovable subscriptions for Claude Code. That is the pattern investors fear. They keep measuring vibe-coding apps against the raw models. A single Claude Code https://thenextweb.com/news/alibaba-bans-claude-code-anthropic-tracking-chinese-users update, they worry, could wipe out a startup’s edge overnight. Some tools have answered by building their own models https://thenextweb.com/news/base44-base1-proprietary-ai-model-vibe-coding to lean less on the labs. Why it matters Lovable itself names the threat plainly. In a March podcast, its head of growth Elena Verna said rival startups were not her main worry. “I always worry about the big boys and girls in the world,” she said. “So, OpenAIs, Anthropics, Googles, Apples.” Her reasoning came down to distribution. The giants reach billions of users, and can bolt a coding assistant onto products people already open every day. Cuban’s counter is simple. Own the workflow around the code, and the incorporation, the payments and the data all build a wall the labs cannot easily cross. Whether that wall holds as the models improve is the open question. A lot of European startup value now rides on the answer. Get the TNW newsletter Get the most important tech news in your inbox each week.