While Moore's Law promises to double computing capacity and reduce costs by a factor of 2 every 2-3 years, both public-sector and private-sector IT departments seem to be demonstrating the opposite: spiraling costs to maintain break-even functionality. Why this disconnect between theory and practice? According to an October 2001 article in CIO magazine, a large part of the reason may be that while hardware is getting better, faster, and cheaper, proprietary software is nullifying all these gains, and then some. Indeed, according to their analysis, $78 billion dollars of IT spend per year (nearly $20B per quarter) is being wasted because of "bad software". For comparison, $78B is greater than the entire federal IT budget, and represents between 16% and 45% of the entire reported earnings of the S&P500. What can possibly resolve this crisis? While CIO magazine and numerous other reports suggest that Linux and Open Source software represent a radical and effective remedy to the broken model that has effectively halted all growth in the technology industry today, these conclusions are too simple and too hasty. Of course it is easy to show that on a case-by-case basis System X running Linux on Intel delivers 6x or 100x better price/performance than System Y running on a proprietary RISC/Unix system, but these comparisons are meaningless without evaluating whether System X can scale to do everything System Y was intended to deliver in the first place. The comparison must be apples to apples. Recently a number of investment banks on Wall Street have migrated major trading systems (which transact trillions of dollars per year and which earn hundreds of millions of dollars per year) from proprietary RISC/Unix systems to Linux on Intel. These migrations were not simple replacements of file servers, web servers, and other edge-computing resources. These migrations involved a comprehensive migration plan, enabled by a sound enterprise architecture, and fulfilled by the first true enterprise Linux platform: Red Hat Linux Advanced Server. With the Common Operating Environment certification of Red Hat Linux Advanced Server by the Defense Information Systems Agency (DISA), Red Hat Linux Advanced Server has demonstrated the proven conformance to industry standards making the first enterprise Linux platform ready not only for enterprise deployment in the DoD but also across government In this talk, Michael Tiemann will show why it is necessary to stop thinking of Linux as another operating system and start thinking of Linux as an architecture. He will describe how this architectural thinking, the open source community, and a new approach from Morgan Stanley RISC/Unix systems Red Hat helped accelerate an enterprise migration from proprietary to Linux on INtel by three years, demonstrating "100 million reasons why architecture matters."