can you describe the j-proudctivity curve for AI - need to invest now with time, foundations, etc which will lower productivity before it bounces back?

The sources describe a phenomenon akin to a J-shaped productivity curve for AI, where substantial upfront investments in time, infrastructure, and foundations lead to initial slowdowns or elusive gains before eventual rebounds. High AI spending, including on models and data infrastructure, has outpaced returns so far, creating a "productivity paradox" with billions invested but no significant firm-wide improvements in output per worker or total factor productivity [5][9]. This lag reflects the need for foundational work, such as integrating AI tools, which temporarily diverts resources without immediate macroeconomic boosts, as seen in scarce evidence of real gains despite vast investments [2][8]. However, emerging data suggests an inflection point, with revised U.S. statistics indicating productivity growth accelerating to about 2.7% in 2025—double the prior decade's average—and potential for 1.8% annual increases over the next decade from current AI models [1][6][7]. Over longer horizons, AI is expected to revive sluggish growth through innovation and efficiency, similar to historical mechanization shifts that eventually lowered prices and raised wages [3][11].