Many foreign aid agencies fund large-scale agricultural training for farmers in developing countries, but little rigorous research has been conducted on whether these programs are effective. We used a clustered randomized controlled trial to estimate the effectiveness of a U.S. government-funded farmer training program that trained more than 50,000 farmers throughout Armenia. Three years after farmers received training, training did not increase household income or consumption. Training also did not affect mediating outcomes, such as adoption of agricultural practices or changes in cultivation of crops, which suggests that longer-term impacts are unlikely to materialize. Many farmers lacked the financial means to invest in the types of practices that were the focus of the curricula, and farmers were also often unwilling to try new crops that have higher up-front costs even if they are much more profitable in the long run. Our findings highlight the challenges that even a well-implemented training program has in spurring behavioral change among farmers and the challenges of providing effective services when foreign aid agencies prioritize having a large programmatic footprint. These challenges were central to the lack of impacts of this particular program but are underplayed when foreign aid agencies decide whether to fund agricultural training programs.