Andrew Samwick makes the case for infrastructure as the cornerstone of stimulus:

[T]here is no free lunch: the money we spend today is a loss to the Treasury, whether as “timely, temporary, and targeted” tax cuts that have no discernible impact; payments to delay bankruptcy for large, mismanaged entities, whether AIG or the Big 3; or the largest public works program since the interstate highway system. That loss to the Treasury must be made up at some future date, by later cohorts of taxpayers.

Fortunately, both of these problems can be overcome by focusing all new spending on investment rather than consumption and on public investment rather than private investment. By their nature, capital investments last for years or decades, so that there is a better chance that those who are paying for the spending are reaping its benefits. Public investment also meets the criterion that the spending goes for projects that are within the government’s responsibilities. Repairing roads today removes the need to repair them for a number of years. In 2005, the American Society of Civil Engineers released a report card in which it estimated that $1.6 trillion would be required over a five-year period to restore the nation’s physical infrastructure to good condition. If I had a target of $500 billion to spend, every dime would go for public infrastructure investments, and we’d still have quite a bit of work to do.

These are reasonable points. The counterbalancing consideration is that it takes time to get these projects going. My understanding is that there aren’t $500 billion worth of products that are “ready to go” within the relevant time frame. So we need to mix infrastructure investments with other stuff that acts faster like aid to state and local governments.