Paul Graham's article is intriguing. Of course Microsoft isn't dead; as Joel Spolsky likes to remind us, they have enough money (40 billion or so) in the bank to continue for years, decades even. They can buy any talent or startup they want. They can throw a hundred programmers at a problem faster than you can say absorb-and-extend. But are they dead as a threat? He makes a good case.
Software can be split into: the OS, business apps, games, email, entertainment, and e-commerce. The last four have moved off the PC. Games are on consoles or on-line (World of Warcraft). EMail and entertainment are on google, youtube, and friends. E-commerce (travel, ebay, banking) is browser-based. The only thing left are business apps and the OS.
Microsoft Office rules biz apps, and given the huge migration costs (data migration and training), that isn't about to go away. CAD/CAM and other specialty software will stay on the PC for now.
As for the OS, it really depends on the others. If you can do games, email, etc on some other OS, why buy Windows?
The fly in Paul Graham's argument is price. Apple used to love ride the demand curve. They'd bring out Macs at $4000 and let the die-hard fans buy. Then they'd lower the price and well-off firms would pony up. Finally the price would drop to its competitive mass-market price. What's to stop Microsoft doing the same, and dropping the price of Windows to $10 and Office to $40? At that price, Linux's warts start to look more troubling.
Another issue in Microsoft's favour is the PC itself. With prices down below $500, the main rationale for thin clients goes away. If you have a 3G CPU with 2G of RAM, it makes little sense to use it only as a glorified terminal. By its existence, uses will be found for it. Therefore client-side sofware won't disappear. Therefore Microsoft's "home turf" isn't going away. Doesn't mean they'll stay on top, but as the Hotmail founder said when bought out for $460 million bucks -- "when you have millions of customers, it's too big a lead for anyone else to catch up".
Privacy is the killer of on-line apps. A few bad news stories about data theft or government back doors into Google code, and the whole web app things gets riskier.
Update: Fouad's new blog Developer's Vista covers this topic. Also, this article suggests XBox is a 5 Billion dollar money loser and a "disastrous endeavour". Good thing they can afford it.