AI at Halftime: What Comes Next?

Sam Altman has described the current moment in AI as "halftime." Phase one — the rapid rise of large language models like GPT-4 and Claude — built the foundation. Companies experimented with chatbots, content generation, and workflow automation. Now, as adoption accelerates, the question is what the second half of the game looks like.

History offers a guide. In 1995, fewer than 1% of people worldwide used the internet. By 2005, that figure had jumped to 16%. Today, over two-thirds of the global population is online.¹ Early internet adoption was driven by browsers and email, but the real value emerged in phase two: infrastructure (broadband, cloud), applications (e-commerce, social media), and integration into everyday business. AI is following a similar curve.

Right now, AI powers only 5–10% of workflows in most companies, according to McKinsey.² But those who have adopted are already seeing productivity gains of 15–40% in targeted areas.³ Phase one was about showing what's possible. Phase two will be about embedding AI into core systems: customer support platforms that resolve queries end-to-end, supply chains that self-correct in real time, and finance tools that forecast and execute without constant human intervention.

Infrastructure is the other half of the story. Just as the internet required fiber optics, content delivery networks, and smartphones to scale, AI will demand massive GPU supply, expanded power grids, and new data pipelines. Nvidia has forecast datacenter power demand from AI could double by 2030.⁴ Meanwhile, startups and enterprises are racing to build "agentic" systems that don't just predict, but act — scheduling meetings, reconciling accounts, even executing marketing campaigns.⁵

The risks of halftime are real. During the dot-com bubble, capital flooded into speculative plays before infrastructure and consumer demand caught up. Many firms collapsed, but those that built with discipline — Amazon, Google — defined the next two decades. The same will likely be true for AI. Overhyped pilots may fade, but systems that deliver measurable ROI will endure.

For executives, the message is clear. Halftime isn't a pause; it's the pivot point. The winners of the second half will be those who stop treating AI as an experiment and start weaving it into operations, infrastructure, and strategy. As with the internet, the tools that seem optional now will soon be non-negotiable.