Despite nearing 1 billion weekly users, ChatGPT's growth is 7 months behind schedule. With only ~5% of users paying, OpenAI is prioritizing enterprise products like Codex and aims for businesses to constitute 50% of its revenue by year-end, signaling a major strategic shift to where the money is.
Stripe is considering acquiring OpenRouter for over 70 times its revenue, a multiple three times higher than the recent pricey Cursor deal. This inflated valuation suggests that fear of missing out (FOMO) and competitive bidding from other players are driving the price far beyond traditional fundamentals.
Reflection AI shifted its entire strategy overnight to building foundational open-source models after a meeting between its founder and Nvidia's CEO. This move serves Nvidia's strategic interest in diversifying the AI ecosystem and chip demand away from giants like OpenAI who develop proprietary silicon.
Unlike closed-source models where release timing is constrained by inference costs, open-source models benefit from being released "as soon as possible." This strategy helps capture developer loyalty and community engagement, as users run the models on their own infrastructure, freeing the lab to focus on training the next generation.
Cisco's rationale for joining the Open Secure AI Alliance is a philosophical one: in cybersecurity, the primary threat is external adversaries, not rival companies. This mindset prioritizes industry-wide collaboration and information sharing to build a united defense against attackers who operate at machine scale.
Despite the economic and performance benefits of Chinese open-source models, Cisco refrains from using them. As a provider for critical infrastructure like hospitals and power plants, the company prioritizes national security and safety over cost savings, deeming the potential unknown risks too high.
In a future where open-source models commoditize the model layer itself, closed-source labs will likely adapt their business models. Monetization will move up the stack to the application layer (where the "last mile" value is) and down to the infrastructure layer (optimizing costs with custom chips).
Contrary to fears of mass job displacement, step-function improvements in AI will likely increase jobs in the short to medium term. Each technological leap creates new bottlenecks that require human intervention and expertise to manage, from implementation to oversight, thus creating new roles and responsibilities.