Katie Wood: Open AI is not a threat, but a growth catalyst for OpenAI and Anthropic
Contrary to the prevailing industry opinion that publishing open models with weights undermines the business of leading labs, ARK Invest CEO Cathie Wood offers a radically different view. She argues that the spread of open-source AI not only does not harm OpenAI and Anthropic, but paradoxically strengthens their market positions.
The essence of Wood's argument is built on a reinterpretation of threats. Open models, such as those developed by Meta, Mistral, and DeepSeek, are indeed becoming increasingly powerful. However, the main danger, according to Wood, lies not in competition for users, but in the exponential growth of cyber threats. The more accessible and advanced these tools become, the wider the attack surface for malicious actors.
Security as the main driver of demand
This logic flips the traditional valuation model. Instead of taking market share from paid leaders, open models create a new, more compelling argument in their favor for sales. Businesses, faced with growing vulnerability, are forced to invest in the most advanced protection, and therefore turn to cutting-edge commercial solutions. Thus, every new open release indirectly increases demand for OpenAI and Anthropic products.
Data from the UK AI Safety Institute confirms this trend: open models in terms of cyberattack capabilities have already nearly caught up with leading proprietary developments from four to seven months ago. In the ARK AI primetime report, the growth in AI infrastructure spending is outright called the trend of the decade.
Betting on the leaders
According to Wood's forecast, the main financial flows from the AI revolution will be concentrated precisely with OpenAI, Anthropic, and, in the future, SpaceXAI. ARK Invest is already investing in all three companies. "Paradoxically, open models are becoming the main reason why these giants will capture the bulk of AI revenue," she states. It is worth recalling that Anthropic has already filed an S-1 application for a listing with a valuation of about $1 trillion, and OpenAI's IPO is expected in September 2026.
My analysis: Wood's argument is strong, but not without nuances. It ignores the possibility of "killer" open-source models emerging that would close the quality gap. Nevertheless, her thesis about security as the main commercial advantage in the era of ubiquitous AI looks increasingly convincing. The market will likely value leaders not by the number of parameters, but by their ability to provide protection against AI threats.