TLDR: Census data is complicated.
The simplest actual answer to your question: If you type in "united states" on the front page of American FactFinder, you can get to the "Community Facts" page for the US, which has direct links to several kinds of data at the national level.
Probably way more than you really wanted to know:
More generally, Census data is a very simple phrase behind which lies a ton of variety, nuance and detail. I've spent several years now working on various projects to make Census data easier to use, and it's harder than you would guess. Even a straightforward concept like "population" has multiple answers from the US Census: every ten years the Decennial census comes up with one count; annually, the American Community Survey (ACS) estimates the population using a different methodology, and every month, the Population Estimates program produces its own numbers. Oh, and also monthly the Census Bureau and the Bureau of Labor Statistics update the Current Population Survey (CPS) which has its own total population number.
Is there no straightforward to say "I want data X about region Y, grouped by subregion Z, give it to me now, please and thank you"? Importantly, Y may be "the entire United States", and ideally, X may include multiple data points.
For the American Community Survey alone, there are several hundred table variants, and several of the variants come in two forms, and some of them come in about 20 forms because they break down figures according to race or hispanic status. (In total, there are over 1400 tables in the ACS). Because the raw data can't be released for privacy protections, the Census needs to presumptively tabulate all the different kinds of cross-referenced data people might want. (And some of the data you may want is collected in the CPS or other programs.)
the Census Bureau data is tabulated into "summary levels" which represent different classes of geography. The ACS Glossary has a summary level list has 188 different kinds of value for Y, with hundreds of thousands of actual values. And there are more summary levels than on that list. The Missouri Census Data Center (MCDC) has delved into summary levels and tried to compile a master list which has 220 of them.
These lists are complicated for technical reasons (mostly to do with the next section) but even if you think in terms of simple shapes-on-a-map, there are about 30 basic geographies.
Also, for the US and for states, there's a separate concept ("geographic components") to do with "give me the total of data X for just the people living in "urban areas", or "not in a metropolitan or micropolitan statistical area". There are about 20 of these components.
grouped by subregion Z
only some Census geographies are by definition properly contained by specific other geographies. Cities (sumlevel 160, formally known as "places") cross census tract and county boundaries, adding a bit of complexity to one commonly desired grouping. And in some parts of the country, locally prevailing government systems make "places" a less appropriate way to break things up: things are different in many ways in New England, especially.
Oh, and "the census year from which I want the data." Census maps change every ten years, so data is not always comparable across Census releases. How you do your comparisons when that happens depends sometimes on whether you're looking at statistical areas like census tracts, or legal areas like states and counties. And the questions change too, not just the maps. In 2000, the Census Bureau allowed people to select more than one race, complicating comparison of new data with data collected before then.
Third Party Projects
A lot of people have worked on third-party projects to make Census data easier to use. I invite you to check out a project I've worked on called Census Reporter. It's limited to the American Community Survey for the foreseeable future. I'm proud of what we've done, and we get great feedback, but I know for a fact we have a long way to go to make it as easy to find tables as you (and I!) want.
Besides Census Reporter, there's a commercial project called Social Explorer which has done a lot of work to make the data more directly cross-comparable. You may be able to access it through your local library.
There are some other great projects which also take on these challenges, but which tend to be aimed at researchers who are more deeply invested. For example, the aforementioned MCDC, NHGIS, IPUMS, Census.IRE.org (which I also helped to make), and I'm sure more that I'm forgetting.