Meta pursues muddled AI strategy
Meta CTO Andrew “Boz” Bosworth apology Recruiting staff on Monday over the company’s mass layoffs and AI missteps may be the first step toward relieving morale, but it points to a larger problem: its AI strategy still doesn’t make sense.
The company is borrowing like crazy to build massive data centers to train AI models that are unlikely to keep up with leading labs. Meta does not have the enterprise businesses that Anthropic or OpenAI have built, which require large numbers of tokens. AWS also doesn’t seem to want to be a cloud services provider like Google or Microsoft.
Meta may have gained traction with its hardware strategy (a strategy that even Apple has struggled to emulate), but it can’t do much with the top-end models. Even though there are billions of Instagram, WhatsApp and Facebook users, when you make AI available to so many people, you need to use highly efficient models to control cost. Chasing more than that can be a costly distraction.
