I study the effects of innovation policies in product-line markets, using a dynamic oligopoly model with endogenous pricing, product-line adjustment, and R&D. I apply the model to the desktop GPU market, a duopoly between NVIDIA and AMD. I find that a dollar of R&D generates about $9.0 of within-market surplus net of that dollar, with the firms capturing $3.0 and consumers the rest. But the return to public support is far more modest. I simulate three stylized interventions: an R&D credit and two product rebates, one at the cutting edge and one on the mass market. Only the R&D credit induces a material increase in private R&D, but it returns only $0.51 of surplus per public dollar because additionality is low: most of its outlay reimburses research the firms would have conducted anyway. The rebate on cutting-edge products returns only $0.23 per public dollar and sets back the technology frontier. Subsidizing both firms, it ignites a price war at the top of the line, benefiting consumers but undercutting the reason firms advance the frontier. The mass-market rebate sets off the same war lower down the product line but expands adoption enough that R&D survives, returning $1.09 per public dollar. These results suggest that innovation policy in product-line markets is a more intricate design problem than in markets with simpler structure. Each instrument induces a distinct equilibrium response in prices, qualities, and R&D. The static and dynamic components of that response can offset or reinforce each other, attenuating or even reversing the intended effect.