The Case for the Diversified AI Stack: Avoiding the 'Bet the Farm' Trap

In the current AI gold rush, the prevailing narrative suggests a winner-take-all outcome. While many venture capital firms are betting heavily on a few frontier labs, as TechCrunch first reported, the real strategic win for operators may not be picking a single winner—it's building a diversified stack.

According to reporting from TechCrunch, Insight Partners has adopted a diversified strategy even as other VCs have "piled into the frontier AI labs." Devin Parekh, who has co-run the investment firm for 26 years, noted that Insight Partners holds stakes in both OpenAI and Anthropic. While investing in rival companies was once considered taboo in the venture world, Parekh explained to TechCrunch that the firm viewed the two through different strategic lenses: OpenAI as the dominant consumer play and Anthropic as having a clear enterprise strategy.

From an operational standpoint, this approach mirrors the necessity of avoiding vendor lock-in. When an organization ties its entire AI infrastructure to a single provider, it inherits that provider's risks and pricing power. Parekh noted that as these frontier companies require massive capital—raising roughly $30 billion to $100 billion—they have lost the ability to dictate exclusivity. This shift suggests that the market is moving away from a closed ecosystem toward one where multiple powerful models can coexist.

Furthermore, the risk profile of AI is shifting. Parekh told TechCrunch that venture valuations are currently rising at a pace reminiscent of 2021, a period he noted "didn't end well." He observed that current funding rounds are moving so quickly that there is almost no incremental data available, meaning investors are paying higher prices without a corresponding reduction in risk. For businesses, this volatility is a signal to avoid over-indexing on a single "buzzy" solution.

Instead, the operational win lies in vertical AI. Parekh highlighted that while AI infrastructure talent is concentrated in San Francisco, vertical AI talent is more geographically diverse—citing Ramp as an example of talent concentrated in New York for financial services. This means the most impactful AI wins may not come from the general-purpose frontier models, but from specialized applications tailored to specific industry nuances.

**Opinion:** In my view, the most resilient tech stacks will be those that treat frontier models as interchangeable utilities rather than core identities. By diversifying the AI layer, operators protect their margins from the pricing whims of a single vendor and ensure they can pivot as the "enterprise strategy" Parekh mentioned continues to shift in real time.