For the last 50 years, the United States has implemented various types of export controls aimed at restricting the trade of strategic products, including high performance computers (HPCs). The origins of today's high-technology export controls can be traced to the Cold War era and efforts by the United States to restrict Communist-bloc nations' access to these strategic products. While these controls proved successful, U.S. policymakers now face the challenge of developing a new framework that can keep pace not only with a more politically uncertain global community, but also with the explosive growth of information technology (IT).

As computers have become faster and more powerful, export controls have not been adjusted to match these changing realities. Soon, many commercial mass-market computers, even some laptops, will be powerful enough to fall under Department of Commerce (DOC) controls originally drafted to restrict export of a relatively small number of top-end HPCs.

In the old economy, where the United States dominated the markets for HPC's and where the production of these items was limited, it was feasible to control these exports. However, in the new global economy, where many or most computers will soon be HPCs (or can be configured to be HPCs), and are produced in many countries around the world, it will become increasingly difficult for the United States to control who gets access to these products.

In the aftermath of the Cox Report, the Progressive Policy Institute (PPI) shares the belief that renewed attention should be given to export controls, particularly in the case of China. Controls must restrict the export of HPCs that pose dual-use threats and are unavailable from anywhere other than the United States and its closest allies. But, we should not overreact and allow reports of Chinese espionage to unduly influence export controls, particularly on computing products that will be available throughout the world regardless of U.S. export controls.

Computer export controls must be targeted if they are to effectively protect our national security and our computer industry's competitive advantage. As a result, PPI recommends that the federal government raise the criteria defining high-performance computers and establish a sliding index that allows export controls to keep pace with the constantly increasing levels of computer processing power. Government export policy needs to reflect the speed and innovativeness of the overall economy and not retard the growth and competitiveness of an industry that is driving the transformation to a digital economy.

Under an indexed regime, the United States government can avoid expending scarce resources on futile attempts to prevent the spread of commonly available mass- market products. By selectively focusing on certain countries and the small number of truly high-performance computers, the United States can more effectively control the spread of dual-use technology and reduce security risks.

Background & Current Policy

Since their introduction over 20 years ago, the speed and power of personal computers has grown exponentially, with no end of that growth in sight. As the technology developed, the U.S. government implemented export control regimes to block the transfer of the most powerful "supercomputers" to certain countries. In 1988, the U.S. Department of Commerce (DOC) defined "supercomputers" as any computer operating at 195 millions of theoretical operations per second or MTOPS, which measures a computer's highest theoretical performance if it is operating flawlessly, or greater. The DOC also required licenses for the exportation of computers operating at or above 12.5 MTOPS. Under these levels, the standard 386 MHZ-chip PC of the early 1990's was considered a "supercomputer." These restrictions remained relatively intact until February 1994, when the supercomputer threshold was raised from 195 to 1,500 MTOPS, and the license requirement for most exports was increased from 12.5 to 500 MTOPS.

It was not until the creation of the current controls in January 1996 that the government began considering the long-term projected availability of HPC's. Relying on a 1995 Stanford University study commissioned by the U.S. Commere and Defense Departments, DOC established the concept of "forward looking foreign availability." Among other things, this concept relied on the Stanford report's projection that the spread of U.S. manufactured computer technology up to 7,000 MTOPS would be worldwide and uncontrollable by 1997. Anticipating the report's projections-- which proved accurate--DOC based its controls on the level of computer performance that would be available in 18 to 24 months, rather than at the time the regulation was conceived.

These forward looking regulations--which, with some alterations, still govern HPC exports--were established by amendment of the Export Administration Regulations by the Clinton Administration in 1996. They identify four groups of countries for export controls on computers. To export to the first group, or Tier I, no licence is required for even the most powerful supercomputer. This tier comprises the United States' 28 closest allies and therefore poses little security risk to this country.

Exports to Tier II countries--106 trading partners with mixed proliferation and export control records (e.g. South Korea and South Africa)--require licences for computers operating at levels above10,000 MTOPS. Record keeping is required for exporting computers that operate below 10,000 MTOPS.

Tier III countries are defined as countries that pose a proliferation, diversion, or other form of security risk (e.g. China, Russia, and Israel). Originally, exports of computers operating at up to 7,000 MTOPS to civilian end-users in the fifty Tier III countries required no license. However, licences were required for sales to military end- users of computers operating above 2,000 MTOPS. These controls were tightened in 1997 when Congress, through amendments to the National Defense Authorization Act (NDAA), required exporters to notify the DOC of computer exports between 2,000 and 7,000 MTOPS to any end-user in a Tier III country. Under the advanced notification rules, any agency involved in the export control system, such as DOC, DOD, State, and Energy, has 10 days to object to a sale without a licence. An objection by one of these agencies necessitates a license.

These amendments also allow the President to revise the MTOP threshold for Tier III countries. Such revisions would take effect 180 days after the President submits a report to Congress justifying the revisions. Tier III countries are among the fastest growing computer markets in the world, and the regulations concerning HPC exports to these countries is at the heart of the current export control debate.

Finally, there are seven Tier IV countries (Cuba, Iran, Libya, Iraq, North Korea, Sudan, and Syria) which have been classified as terrorist states. All computer exports to these countries are prohibited.

An Outdated System

The export control system was originally developed to restrict the transfer of America's most sophisticated supercomputers to dangerous end-users. However, technological innovation continually redefines "sophisticated," and has made existing MTOPS thresholds obsolete. As many in the computer industry are fond of saying, "yesterday's supercomputer is today's laptop," and Gordon Moore, the former CEO of Intel, has claimed that the power of microprocessors (chips) doubles every 18 months. The accuracy of Moore's observation has earned it the title, "Moore's Law," as it has held true for thirty years. If this rate of advancement continues, as is expected, many mass-market computer products will soon hit the MTOPS ceiling and be subject to unnecessarily restrictive controls.

Indeed, the current export regime will face a major challenge at the end of this year when Intel Corp. is expected to release its Pentium III Xeon chip which operates at 1,283 MTOPS. When the Pentium III becomes available, many standard products that link two or more chips will be catapulted above the 2,000 MTOPS level and into "supercomputer" status. For example, small business e-mail servers, which typically link two, four, or eight chips, will operate at 2,566 MTOPS with two chips, and therefore carry DOC notification requirements if exported to any Tier III country. In addition to servers--other high-tech commodities--such as laptops and even children's video games, will have processing speeds above 2,000 MTOPS. The processing power of these computers is a far cry from that of the most powerful supercomputers like IBM's Pacific Blue, which operates at a rate of 1.6 million MTOPS, or 800 times faster than the slowest government-designated "supercomputers."

If left unchanged, the current export control regime will run head-on into these advances in technology and increase the notification and licensing review caseload. The potential then arises for an overburdened system to more slowly process applications and in the process retard the growth of an industry that is critical to America's overall economy growth.

Global Markets and the Effects of Foreign Competition

Equally important as "Moore's Law" is its corollary that the cost of computing drops nearly 25 percent per year. Coupled with the Internet's dramatic growth and the convergence between telecommunications and computers, these falling costs have created a rapidly expanding global market for HPCs. An estimated 80 million personal computers were sold worldwide in 1997, with a significant percent being sold outside the United States. Also, of the over 2.5 million servers and workstations sold in 1997, more than 1.6 million were sold outside the United States.¹

While U.S. computer companies remain world leaders, an increasing number of foreign competitors have emerged. In 1997, 15 of the top 25 personal computer manufacturers and 14 of the top 25 workstation manufacturers were not U.S. companies. Indeed, if U.S. HPC export controls are not raised, foreign competitors will take market share from U.S. companies. More importantly, many of these companies are located in countries that either do not belong to (e.g. Taiwan, Hong Kong), or do not strictly enforce (e.g. Germany), multilateral or unilateral export control regimes.

The computer industry is a key driver of the country's booming economy. It is not only one of the nation's largest industries, but also the top merchandise exporter.² However, if unrealistic computer export controls impede foreign sales, some of the more than 4.3 million jobs in the American high-tech industry would be threatened.³ Such losses would be especially damaging since those jobs pay, on average, 77 percent more than jobs in other sectors.⁴

National Security

Unrealistic export controls may also threaten our peace and prosperity by limiting the technology available to our military and intelligence communities. A high-tech military for the 21st century depends on successful American high-tech firms funding R&D to develop of next generation products and technologies. The military and national security communities' dependence on technologies developed in the private sector is now greater than ever as the government is forgoing the lengthy contracting process and procuring directly from civilian markets. Indeed, while controls on bona-fide high-end supercomputers will serve our national security interests, export restrictions on mass- market computers that restrict innovation in the computer industry could reduce the lead enjoyed by U.S. intelligence and weaponry systems.

Advancements in technologies known as "clustering" and "parallel processing" also pose new challenges to our export control regime. These practices allow organizations to achieve high computing power without traditional supercomputers by connecting many lower-level machines together to perform tasks at high speeds. Often, the individual computers comprising these networks can be "off the shelf" equipment which fall below current MTOPS thresholds, yet their aggregate power equals or exceeds that of highly sophisticated supercomputers subject to export controls. While the extent of such networks is still not fully known, examples like the NT Cluster at the National Center for Supercomputing Applications illustrate just how effectively they work.

Clustering and parallel processing, along with rapidly advancing chip technology, growing foreign availability, and manufacturing, are each examples of how difficult it is becoming to control low-end computers and why redirecting the focus of export controls is so essential.

Export Policy in the New Economy