Google VP Highlights Challenges for AI Startups: LLM Wrappers and Aggregators at Risk
The rapid expansion of generative AI has led to the emergence of numerous startups. However, as the industry matures, certain business models are facing significant challenges. Darren Mowry, who oversees Google’s global startup initiatives across Cloud, DeepMind, and Alphabet, has identified two types of AI startups that may struggle to sustain themselves: those focusing on Large Language Model (LLM) wrappers and AI aggregators.
Understanding LLM Wrappers
LLM wrappers are startups that build upon existing large language models—such as Claude, GPT, or Google’s Gemini—by adding a user interface or product layer to address specific problems. For instance, a company might develop an AI-driven tool to assist students with their studies by leveraging these foundational models.
Mowry emphasizes that merely relying on the underlying model without substantial differentiation is no longer sufficient. He states, If you’re really just counting on the back-end model to do all the work and you’re almost white-labeling that model, the industry doesn’t have a lot of patience for that anymore.
To thrive, startups must establish significant competitive advantages, either through horizontal differentiation or by targeting specific vertical markets. Examples of successful LLM wrappers include Cursor, a GPT-powered coding assistant, and Harvey AI, a legal AI assistant.
The Role of AI Aggregators
AI aggregators represent a subset of LLM wrappers. These startups integrate multiple LLMs into a single interface or API layer, enabling users to route queries across various models. They often provide orchestration layers that include monitoring, governance, or evaluation tools. Examples include AI search startup Perplexity and developer platform OpenRouter, which offers access to multiple AI models via a single API.
Despite initial traction, Mowry advises caution for new entrants in this space, stating, Stay out of the aggregator business. He notes that aggregators are experiencing limited growth because users seek built-in intellectual property to ensure they are directed to the appropriate model based on their specific needs, rather than being influenced by underlying computational or access constraints.
Historical Parallels and Lessons Learned
Mowry draws parallels between the current AI landscape and the early days of cloud computing in the late 2000s and early 2010s. During that period, numerous startups emerged to resell AWS infrastructure, offering tools, billing consolidation, and support. However, as Amazon developed its own enterprise tools and customers became more adept at managing cloud services directly, many of these intermediaries were phased out. The survivors were those that added genuine value through services like security, migration, or DevOps consulting.
Similarly, AI aggregators today face margin pressures as model providers expand their enterprise features, potentially rendering middlemen obsolete.
Emerging Opportunities in AI
Despite the challenges faced by LLM wrappers and aggregators, Mowry remains optimistic about other areas within the AI sector. He highlights the potential of developer platforms and direct-to-consumer technologies that empower users with advanced AI tools. For example, Google’s AI video generator, Veo, offers film and TV students the opportunity to bring their stories to life, demonstrating the transformative potential of AI in creative fields.
Beyond AI, Mowry also points to significant growth in biotech and climate tech sectors. These industries are attracting substantial venture investment and leveraging vast amounts of data to create real value in ways previously unimaginable.
Conclusion
As the AI industry continues to evolve, startups must critically assess their business models and strive for meaningful differentiation. Relying solely on existing large language models without adding unique value is becoming increasingly untenable. By focusing on innovation and addressing specific market needs, startups can navigate the challenges ahead and seize emerging opportunities in the AI landscape.
