TechJun 3, 2024
AI Was Once Thought To Be The Disruptive, Transformative Technology Of Our Time, Now Investors Are Skeptical
Model Stagnation
Noah Weidner
Discover
CEO Jensen Huang may be the god of delivering shareholder value for Nvidia — but whether he can do that for his customers is still up for debate. Billions are being funneled into pursuing artificial general intelligence (AGI), promising a new era for humanity. Yet, this grand vision clashes with an industry bogged down by costly infrastructure — all while customers remain wary of AGI’s potential.
Betting on the future: In 2023, Nvidia sold over $50B worth of AI chips. However, Sequoia estimates only $3B of this revenue came from outside of the chip industry. Unsurprisingly, Nvidia aims to sell even more chips. On Sunday, Huang announced plans to upgrade AI accelerators annually, faster than the previous two-year cycle. This means the tech industry will have to pony up — and while that could eventually pay off, analysts are now questioning if AI will be as transformative as once thought.
Running AI infrastructure is expensive — and for now, it remains more costly and less capable than employees, per Wharton’s Professor of Management Peter Cappelli (W. Compounding those problems, the progress on building AI models is slowing.
GDP hallucinations: A recent MIT study led by famous economist Daron Acemoglu found that AI could boost GDP by a modest 0.93% to 1.16% within the next ten years, far below Goldman Sachs’ forecast of 7%. While Acemoglu AI’s potential to significantly enhance productivity, he also warns of its darker side, including “deepfakes, misleading digital advertisements, addictive social media, or AI-powered malicious computer attacks.”
