There's a conversation on cryptography@metzdowd.org shortly after the release of the Bitcoin paper that touches on this tangentially:
James A. Donald - Sun, 02 Nov 2008
For transferable proof of work tokens to have value, they must have
monetary value. To have monetary value, they must be transferred within
a very large network - for example a file trading network akin to
bittorrent.
To detect and reject a double spending event in a timely manner, one
must have most past transactions of the coins in the transaction, which,
naively implemented, requires each peer to have most past
transactions, or most past transactions that occurred recently. [...]
Satoshi Nakamoto - Sun, 02 Nov 2008
Long before the network gets anywhere near as large as that, it would be safe
for users to use Simplified Payment Verification (section 8) to check for
double spending, which only requires having the chain of block headers, or
about 12KB per day. Only people trying to create new coins would need to run
network nodes. At first, most users would run network nodes, but as the
network grows beyond a certain point, it would be left more and more to
specialists with server farms of specialized hardware. A server farm would
only need to have one node on the network and the rest of the LAN connects with
that one node.
From the word choice used here, a datacenter of machines using a single server to construct blocks is intended. However, many different parties using the same pool server is not contemplated.
Of course, that doesn't mean that pooled mining is wrong - just that it wasn't anticipated.
If you search for "satoshi" on this page, you can see some responses that tangentially relate to your question.
– morsecoder – 2016-02-15T21:17:21.943