AI Race Heats Up ๐Ÿ”ฅ: Wix's Bold Bet ๐Ÿš€

June 30, 2026 |

Tech

๐ŸŽง Audio Summaries
English flag
French flag
German flag
Japanese flag
Korean flag
Mandarin flag
Spanish flag
๐Ÿ›’ Shop on Amazon

๐Ÿง Quick Intel


  • Wix acquired Base44 for $80 million one year ago, with Base44 initially having an eight-person team.
  • Base44โ€™s custom LLM development aimed to optimize latency, cost, and efficiency, contrasting with models like Opus.
  • Lovable reached $500 million in ARR earlier this month, achieving unicorn status in its Series A round last summer.
  • Inference costs have become a significant factor, driving enterprise customer demand for optimized LLM performance.
  • Enterprise revenue now represents a growing share of platform revenue for Base44.
  • Base44 announced $100 million in annual recurring revenue a few months prior to the briefing.
  • A 20% workforce reduction was recently announced by Base44โ€™s parent company.
  • ๐Ÿ“Summary


    One year after acquiring the vibe coding platform Base44 for $80 million, the company, then six months old with a team of eight, was focused on developing a custom LLM. Founder Maor Shlomo aimed to optimize latency, cost, and efficiency, driven by cost reduction goals. However, inference costs increased, prompting a shift toward enterprise customer demands. Base44โ€™s parent company recently announced a workforce reduction, but the company had grown to $100 million in annual recurring revenue. Simultaneously, Lovable reported $500 million in ARR. These developments highlight the evolving landscape of AI startups, where data, distribution, and technology stack are crucial for defensibility, and where the performance of existing models continues to shape strategic decisions.

    ๐Ÿ’กInsights

    โ–ผ


    BASE44โ€™S STRATEGIC SHIFT: BUILDING A CUSTOM AI MODEL
    Base44, a vibe coding platform acquired by Wix for $80 million, is embarking on a significant strategic shift by developing its own AI model, signaling a move towards greater control and potentially enhanced defensibility within the rapidly evolving AI landscape. This initiative, spearheaded by founder Maor Shlomo, directly addresses concerns about relying solely on external, often expensive, large language models (LLMs).

    THE RISE OF CUSTOM LLMS AND DEFENDABILITY
    The development of custom LLMs by companies like Base44 reflects a broader trend within the AI industry. Defensibility โ€“ the ability of a startup to maintain a competitive advantage โ€“ is increasingly tied to factors beyond just the underlying model. Jonathan Userovici, a general partner at Headline VC, highlights data, distribution, and tech stack as key ingredients. Companies are recognizing that simply adopting a leading-edge model isn't enough; they need to build a robust ecosystem around it. This approach aligns with Base44โ€™s strategy of optimizing for latency, cost, and efficiency, directly addressing concerns raised by competitors like Lovable.

    DATA AS A CRITICAL DIFFERENTIATOR
    Data is identified as a core component of AI startup defensibility. Base44โ€™s initial LLM, โ€œBase1,โ€ was trained on โ€œtens of millions of real user interactionsโ€ generated directly from its platform. This focus on proprietary data, combined with other strategic elements, represents a deliberate attempt to build a moat around their technology. The companyโ€™s goal is to create a model more aligned with user needs and optimized for performance, offering a potentially cheaper and faster alternative to utilizing general-purpose frontier models like Opus.

    COST OPTIMIZATION AND ENTERPRISE DEMAND
    Inference costs โ€“ the expense of running AI models โ€“ are becoming a significant factor driving innovation. Userovici notes that enterprise customers are increasingly demanding infrastructure solutions to orchestrate and optimize model selection, preventing runaway costs while maintaining performance. This shift reflects a broader market trend where businesses are carefully evaluating the return on investment (ROI) of utilizing the latest, most powerful AI models across all use cases. The pressure to control costs is a primary driver of Base44โ€™s investment in a custom LLM.

    BASE44โ€™S FINANCIAL PERFORMANCE AND WORKFORCE REDUCTION
    Despite the strategic investment in a custom AI model, Base44โ€™s parent company is undertaking a significant workforce reduction โ€“ a 20% layoff โ€“ indicating a focus on streamlining operations and improving profitability. This decision is partly influenced by the cost of developing and maintaining the new LLM. Conversely, Base44 itself has been experiencing growth, recently surpassing $100 million in annual recurring revenue, demonstrating a strong market position and user adoption.

    COMPETITIVE LANDSCAPE: LOVABLE AND THE โ€œUNISONโ€ EFFECT
    The competitive landscape is populated by companies like Lovable, which achieved unicorn status through Series A funding and relies on external LLMs. However, Shlomo anticipates a trend of increased model training among larger, more established players โ€“ those with sufficient scale and data velocity. This โ€œunisonโ€ effect, where multiple companies train their own models, could lead to a more fragmented and competitive AI market.

    THE ROLE OF SPACEX AND XAI: CURSOR AND GROKโ€™S AI INTEGRATION
    The acquisition of Cursor and Grok by SpaceXโ€™s xAI further illustrates the growing integration of AI across diverse platforms. Claude Code, another vibe coding player, is also leveraging AI, demonstrating the broader adoption of AI technologies within the coding space. This trend suggests a convergence of AI development and application across various industries.

    CONCLUSION: A LONG-TERM INVESTMENT
    Base44โ€™s decision to develop its own LLM is a calculated move designed to address concerns about cost, performance, and long-term defensibility. While a delayed payoff is acknowledged, the potential for improved margins and optimized customer experiences represents a significant strategic advantage for the company and its parent organization. The ongoing evolution of AI models and the increasing emphasis on infrastructure optimization will continue to shape the competitive dynamics of the vibe coding platform market.