The video highlights Base44’s strategic move to develop its own AI language model to gain control over costs, performance, and geopolitical risks, contrasting this with the instability and high valuations of relying on external AI providers like OpenAI. It argues that specialized, proprietary models tailored to specific business needs offer greater sustainability and defensibility than generalized “frontier” AI models, especially amid emerging concerns over AI sovereignty and unpredictable pricing.
The video discusses the recent move by Base44, a coding platform acquired by Wix, to develop and deploy its own AI language model. This decision reflects a broader trend among AI startups seeking defensibility by owning their models rather than relying on external providers like OpenAI or Anthropic. The speaker questions the high valuations of companies like OpenAI, arguing that the idea of centralized “frontier” AI models as a utility provider is flawed, given the diversity of smaller, open-source models and the architectural realities of AI deployment.
A key issue highlighted is the lack of control businesses face when relying on third-party AI models. Examples include Anthropic restricting usage of its models with third-party tools, and OpenAI frequently updating models in ways that disrupt established workflows. Such unpredictability is problematic for enterprises that require stable, scalable AI solutions. Base44’s approach to owning and training its own model aims to provide greater control over latency, cost, and efficiency, which are critical factors for business adoption and long-term sustainability.
The video also touches on geopolitical concerns, particularly for Israeli companies like Base44 and Wix. The concept of AI sovereignty is emerging alongside data sovereignty, as countries and companies seek to avoid dependence on foreign AI providers that could abruptly cut off access due to political or regulatory reasons. This is especially relevant given recent actions by the U.S. government limiting access to certain AI models for non-Americans. For Israel, with its complex geopolitical situation, developing indigenous AI capabilities is seen as a strategic necessity to ensure uninterrupted AI services.
From a technical and business perspective, Base44’s model is trained on tens of millions of real user interactions specific to their platform, rather than broad internet data. This specialization allows for more efficient, cost-effective, and relevant AI performance tailored to their users’ needs. The speaker contrasts this with the “frontier” models that aim to be general-purpose but may not offer the best return on investment for all use cases. Enterprises increasingly demand AI solutions optimized for their specific tasks to manage costs and maintain performance.
Finally, the video raises concerns about the financial sustainability of relying on external AI providers whose token pricing may change unpredictably, complicating budgeting and operational planning. By owning their models, companies like Base44 gain transparency and control over compute costs, enabling better financial forecasting and potentially stronger profit margins. The speaker invites viewers to consider whether this trend toward proprietary AI models is the future, especially in light of geopolitical risks and the questionable valuations of major AI startups.
