The viability of open source AI faces its most serious test to date. Current trends suggest a narrow window for open models to maintain parity with closed systems. Nathan Lambert argues that the gap in compute and data access is widening. Practitioners must now decide if open-weights models can survive without massive corporate subsidies.