The narrative that powered trillions in AI capex investment just cracked. MRB Partners economist Prajakta Bhide published research showing that AI's contribution to U.S. GDP growth in 2025 wasn't the economy's primary engine at all—it was second place to consumer spending. More critically, when adjusted for imported semiconductors and equipment, AI's actual GDP impact drops by more than half, from an assumed 90 basis points to 40-50 basis points. This shifts the fundamental ROI calculation for every enterprise board and policy maker betting on AI infrastructure.

The economic story that dominated 2025 just shifted beneath everyone's feet. For the past year, the AI boom's narrative was simple and powerful: artificial intelligence spending was saving the U.S. economy. Valuations soared, debt issuance hit records to finance data centers, and economists credited AI capex as the bulwark against stagnation. That assumption shaped trillion-dollar investment decisions across enterprise, government and venture capital.

Then on January 8, MRB Partners published research that punctures the entire premise. Prajakta Bhide, the firm's U.S. economic strategist, dug into the actual GDP numbers from 2025 and found something the market had largely overlooked: AI-related capital expenditures were the second driver of growth, not the first. More importantly, consumption was the primary engine—exactly what usually happens during economic expansion.

But here's the inflection point that matters for capital allocation: when Bhide adjusted for imports—the computers, semiconductors, and telecom equipment that power AI systems—the actual domestic GDP contribution of AI collapsed. Without import adjustment, AI looked like it added about 90 basis points (0.9%) to real GDP growth between Q1 and Q3 2025, roughly 40% of total growth. With the adjustment? Forty to fifty basis points. Twenty to twenty-five percent of growth.

That's not a technical quibble. That's a 50% reduction in the economic case for the capex wave that's defining the decade. "The GDP value of AI is smaller than it might appear given that a lot of high-tech equipment is imported," Bhide explained. She's not wrong. The semiconductors power AI data centers. They come from Taiwan (TSMC), the Netherlands (ASML), South Korea. The multiplier effect American policymakers and investors assumed—that AI spending would generate U.S. jobs, tax revenue, and economic momentum—is cut substantially when the core components leave the country before they ever arrive at the data center.

This finding arrived with chorus confirmation. Bespoke Investment Group published similar analysis in December, showing that in Q2 and Q3 2025, AI-linked categories represented just 15% of quarterly GDP growth, with their overall GDP share below 5%. The singular narrative that dominated 2025—AI or bust—was built largely on Q1 data, when a one-time anomaly created "vastly overstated" perceptions of AI's economic role, according to Bespoke's analysis.

What actually drove 2025 growth? Consumers. Bhide found consumption remains resilient despite slower income growth and wealth concentration among top earners. She expects that resilience to continue into 2026 with support from fiscal stimulus and further AI investments, though she's watching job creation and productivity metrics closely. Without the AI boom, growth would have been "decent, above 1.5%" due to solid consumer spending. The narrative shift: AI matters, but it's not the difference between expansion and stagnation.

The implications hit different audiences with different urgency. For investors, this recalibration means the capex ROI case for AI infrastructure pivots from domestic GDP multipliers to productivity gains and competitive positioning. A $5 billion data center still makes sense if it generates proprietary AI capability and market share—but not if the investment thesis rested on GDP contribution flowing back to shareholders through economic growth and tax revenue. The 6-month window before 2026 capex budgets lock is critical. Enterprises and governments need models that reflect actual import-adjusted impact, not the narrative consensus that dominated 2025.

For policymakers, the shift is sharper. AI infrastructure investment creates jobs in construction and operations, but the semiconductor dependency means the economic multiplier is distributed internationally. That changes ROI expectations for government-backed data center investments and shifts the timeline for when domestic manufacturing policy (fab subsidies, supply chain resilience) becomes urgent rather than aspirational.

For professionals and builders, the timing window opens differently. The narrative correction doesn't slow AI adoption—it clarifies it. Companies should build AI capability for competitive advantage and productivity gains, not because they're riding an inevitable economic wave. That distinction matters for talent strategy, skill investment, and how startups position themselves to acquire customers.

The inflection point is clear: AI spending wasn't the savior narrative it appeared to be, and correcting that assumption changes everything about how enterprises, investors, and governments should evaluate capex decisions over the next 18 months. For investors, the ROI case shifts from GDP multipliers to competitive positioning and productivity. For decision-makers, the 6-month window before budget cycles lock is the critical moment to model AI investment against actual import-adjusted economic impact. For builders and professionals, the narrative correction actually clarifies the right reason to invest in AI—not because the economy demands it, but because the technology delivers competitive advantage. Watch quarterly productivity statistics and capex announcement trends in Q1 2026 to see how quickly the market absorbs this correction.