Despite hundreds of billions of dollars in AI investments and breathless hype about an economic revolution, Goldman Sachs has delivered an uncomfortable reality check: AI's measurable impact on the US economy in 2025 was "basically zero."The analysis, published in Goldman's latest economic research report, attempts to quantify AI's actual contribution to productivity and GDP growth. The results are sobering. While companies poured massive capital into AI infrastructure, training, and deployment, the economic returns haven't materialized at scale.This doesn't mean AI isn't being used. Everyone from lawyers to software engineers to customer service teams has integrated AI tools into their workflows. But there's a vast gap between individual productivity gains and economy-wide transformation. What we're seeing is a classic case of "what gets measured gets managed" - and right now, AI's benefits are incredibly hard to measure.I've lived through hype cycles before. During my startup days, every company was adding blockchain to their pitch deck whether it made sense or not. This feels similar, but bigger. The difference is that AI actually works for many applications - it's just not clear if it works well enough to justify the investment levels we're seeing.The report points to several explanations for the disconnect. First, genuine productivity gains from AI are concentrated in specific sectors and roles, not evenly distributed across the economy. A programmer using GitHub Copilot might be 30% more productive, but that doesn't translate to 30% GDP growth in the tech sector.Second, many companies are still in the experimentation phase. They're spending on AI infrastructure and consulting but haven't yet changed their core business processes. That shows up as cost without corresponding revenue or output gains.Third - and this is the uncomfortable part - some of the productivity claims might be overstated. When companies announce they're "using AI," it often means they've deployed tools that employees ignore or use minimally. The gap between deployment and actual usage is massive.The technology is genuinely impressive. Large language models can do things that seemed impossible five years ago. But impressive demos don't automatically translate to measurable economic value. The question Goldman is asking is the right one: where's the actual return on all this investment?This matters for investors and companies making huge AI bets. If Goldman is right that economic impact is minimal so far, we're likely in for a reckoning. Either the productivity gains will start showing up in the data over the next few years, or we'll see a significant pullback in AI spending as CFOs start asking harder questions about ROI.My take: AI will have real economic impact, but the timeline was always longer than the hype suggested. The companies that will win are those focused on specific, measurable use cases - not the ones slapping on everything and hoping for magic.