Fidji Simo says OpenAI can't afford not to strike massive compute deals: 'What's much riskier would be to not lean into compute'

Fidji Simo, the CEO of Applications at OpenAI, emphasized that the true risk lies in avoiding substantial computing investments rather than making them. In a recent interview with Wired, Simo stated, "While these deals may appear risky on the surface, the greater danger would be to shy away from investing in compute." This sentiment echoes the views of other tech leaders, including Meta's CEO Mark Zuckerberg, who has previously collaborated with Simo at Facebook. Zuckerberg highlighted that the greater peril is not taking bold steps forward in AI development. He remarked, "If we end up misallocating a few hundred billion dollars, that would be unfortunate. However, I believe the risk of playing it safe is far greater." The capital expenditure by major tech firms in AI is so significant that economists argue it is bolstering the overall economy. Yet, some investors on Wall Street express concerns reminiscent of the Dot-Com bubble, wary of the massive investments being made. OpenAI’s CEO Sam Altman recently addressed concerns regarding potential bailouts if data center investments fail. He clarified on social media, "If we make mistakes and can't rectify them, we should face the consequences. Other companies will continue to thrive and serve their customers. That’s the essence of capitalism; the ecosystem will remain intact. Our goal is to succeed, but if we falter, that's on us." With Simo's appointment, Altman indicated a stronger focus on research and computing advancements. Simo, who transitioned from Instacart to OpenAI in May, acknowledged external apprehensions regarding the company's substantial $1.4 trillion commitment to data center investments over the next eight years, particularly given that OpenAI is currently operating at a loss. However, she asserted that the demand for computational power within OpenAI is so pressing that there is no alternative. "When we assess our internal constraints and the potential of what we could achieve with more GPUs, it becomes evident that investing in compute is the right choice," Simo explained. Simo also expressed her ambition to make ChatGPT Pulse, a feature that personalizes updates based on user interactions and connected applications, accessible to all users. Currently, the limitation of available compute resources makes this ambition challenging. "This is just one example, but there are many similar instances," she concluded.