Katie Wood: Open AI is not a threat, but a catalyst for superprofits for OpenAI and Anthropic
Contrary to the dominant narrative, ARK Invest CEO Cathie Wood argues that the proliferation of open models with public weights is not an existential threat to industry leaders, but rather a powerful driver of their future dominance. The logic I see in this analysis flips conventional investor assumptions about competitive dynamics in the AI market on their head.
Wood's key thesis is built not on the traditional "displacement" model, but on a security paradox. As open models from players like Meta, Mistral, and DeepSeek become more powerful, they expand the attack surface for cyber threats. It is this growing vulnerability, according to Wood, that will become the main catalyst for demand for the most advanced and expensive proprietary security solutions. Companies will be forced to invest in premium security, which will directly enrich OpenAI and Anthropic.
Open Source as an Accelerator of Threats
Research conducted by the UK AI Safety Institute confirms this alarming trend: the offensive cyber capabilities of open models are already comparable to those of leading closed developments from four to seven months ago. This means the gap is narrowing, and risks to the corporate sector are growing exponentially. In its latest analysis, ARK AI Primetime explicitly calls the rise in AI infrastructure costs the trend of the decade, and a significant portion of these funds will be directed specifically toward security.
Wood names OpenAI, Anthropic, and the promising SpaceXAI as companies that will capture the lion's share of revenue from AI models. ARK Invest already holds positions in all three organizations. "Paradoxically, open models are becoming the main reason why OpenAI, Anthropic, and, in the future, SpaceXAI will take the bulk of revenue in the AI space," she states. This bet looks especially bold amid Anthropic's preparation for an IPO with a valuation of nearly $1 trillion and OpenAI's potential stock market debut as early as September 2026.
My analysis: Wood's counterargument is not just a defense of her own investments, but a deep bet that security will become the primary commodity in the era of ubiquitous AI. However, this logic is vulnerable: if open models achieve parity with closed ones in quality, the premium for "closedness" could disappear faster than sustainable demand for protection forms. Nevertheless, in the short term, the thesis about rising corporate spending on cybersecurity looks flawless.