Testimony of Under Secretary William A. Reinsch
on High Performance Computer Export Policy

Mr. Chairman, thank you for the opportunity to testify before the Committee on export policy for high performance computers (HPCs).  My testimony will describe recent changes the Administration has made to adjust our computer export controls to ensure effective controls while reflecting the reality of rapid technological progress in the computing sector.  These adjustments were motivated by that progress as well as our belief that national security benefits if U.S. manufacturers remain the world's leaders in this technology.

On July 1, President Clinton unveiled new export controls on High Performance Computers and semiconductors to ensure that control levels accurately reflect advances in microprocessor technology and that regulations accurately reflect any risk posed by HPC exports.  The revised controls maintain the four country groups announced in 1995, but changes the countries and levels as follows:

First, Brazil, the Czech Republic, Hungary, and Poland were moved from Tier Two to Tier One.   This reflects both their strong nonproliferation records and their close economic and security ties with the United States. 

Second, the control level for Tier Two countries, countries that pose little or no security   or proliferation risk, was raised from 10,000 to 20,000 MTOPS. 

Third, the two-level system we established in 1995 for civilian and military end-users was maintained for Tier Three countries, with individual license levels raised from 7,000 to 12,300 MTOPS for civilian end users and from 2,000 to 6,500 MTOPS for military end users. 

Finally, prior notice of exports to Tier Three countries of systems above 2,000 MTOPS is required by law.  The President's announcement would raise the notification floor to 6,500 MTOPS.  As you know, the latter change, by law, cannot take effect until the conclusion of a six month waiting period, which will occur on January 23rd.  (The President decide to delay the implementation of the change for military end users to coincide with the adjustment to the level for the notifications.)  At the time of his announcement, however, the President indicated that he was prepared to implement this change immediately and urged Congress to reduce the six-month waiting period to thirty days.  Unfortunately, Congress has not yet acted on that request.

The President has also directed the relevant agencies to continue to review these levels every six months to determine if further adjustments are warranted.  We believe this commitment is as important as the changes we have made, and we intend to meet the schedule he has laid out.  

The role of the Commerce Department in developing these recommendations was to collect and analyze information on computer technology, the high performance computer market, and the capabilities of foreign producers.  Our work was done in partnership with other agencies, particularly the Department of Defense and the Intelligence community.  In examining these issues, we built on the work done in 1995 and in 1998.  I would like to briefly describe some of what we found.   


Computers were once large, bulky items, expensive and difficult to install and maintain.   Only a few suppliers in the U.S. and Japan existed, and many systems were custom built.  Today, the situation has changed radically.  The engine of change is the microprocessor, which provides the basis of computing power. Improvements in microprocessor design and manufacturing mean that manufacturers now sell single chips with speeds over 2000 MTOPS, the current threshold for notification and post-shipment inspection under the NDAA.  These microprocessors, sold in the tens of thousands, power ordinary PCs and laptops which under current law must be treated as high performance computers.

These chips and the related motherboards are commodity items which are shipped to thousands of distributors around the world.  The computers which use these items include servers and workstations, which require little support to install or operate.  The affordability and widespread availability of these computer products permit foreign end-users to modify commodity computers into high performance systems.  The result is that high performance computers are smaller, cheaper, simpler to install and maintain and more reliable than ever before.

Although overshadowed by advances in microprocessor capabilities, recent gains in network technology have also had a significant effect on high performance computing.  Advances in software, interconnects and parallel processing mean that the performance levels once associated with giant machines can now be obtained by smaller and less expensive computers. While there is still an advantage in using one HPC instead of a group of desktop computers, software and connection capability is opening the way to an increase in both use and availability of very high performance computing based on PCs.  Applications for this type of computing include financial analysis, web servers, data warehousing and data mining and have greatly increased demand for parallel processing systems.  The affordability and widespread availability of these computer products permit foreign end-users to configure these mass market products into high performance computer systems.




The number of firms capable of producing computers able to perform above 7000 MTOPS has greatly expanded in the last few years as global demand for computers has increased rapidly, in part driven by the growth of the Internet.  Previously, only the U.S. and Japan had firms capable of producing HPCs for the commercial market.  Other countries, primarily in Europe, had the technological capability to produce HPCs but could not compete effectively.  This is no longer the case.  While the U.S. and Japan remain the two countries with the ability to produce high-end HPCs (vector or massively parallel processor with speeds over 20,000 MTOPS) in commercial quantities, the technology for producing HPCS using four or eight chips is widely diffused around the world.


We found 170 computer manufacturers located around the world who account for three quarters of global computer production.  More than 120 of these manufacturers are outside of the United States.  28 are in countries which are not members of the Wassenaar Arrangement, which provides the basis for multilateral controls on HPCs - computers are not controlled by the  nonproliferation regimes.  The technologies and the production lines these companies use to build PCs and servers can also be used to build computers with speeds over 7,000 MTOPS.  Many of these producers are located in countries that do not observe U.S. or multilateral restrictions on computer exports.  Beyond that, the policies of most of our major trading partners in Europe and elsewhere do not restrict exports of Wassenaar controlled dual-use items to countries like China.  If the U.S. requires its firms to obtain licenses for computers which can be easily produced by non-US suppliers, these foreign suppliers can -- and will -- fill the demand.  They may not do so now because they cannot compete with us economically, but if we remove our companies from a market, they will step in promptly.




I will speak generally to the issue of computer exports and national security, recognizing that the Department of Defense is the U.S. Government agency best situated to provide the Committee with the Administration’s perspective on this issue.  I urge the Committee to invite DoD to discuss this issue with them in depth at the Committee’s earliest convenience.  I will note as a starting point, Deputy Secretary of Defense Hamre’s statements at the time of the President’s announcement that DoD’s national security concerns were taken very seriously during the interagency discussions on this issue, that every one of Defense’s concerns were accommodated, and that DoD is satisfied that it can continue to protect the U.S. with these changes.

The President announced the new control levels after a careful interagency review, involving the Departments of Defense, State, Energy, and in consultation with the Intelligence Community.  Agencies considered a range of options and took an in-depth look at the technology available in the U.S. and overseas.  The focus or our work has been to determine what we can realistically control.  Computers and computer chips which are relatively inexpensive, manufactured in the millions and sold through thousands of outlets around the world are readily available to any purchaser and cannot effectively be controlled.  We think it is better for our national security to focus our energy and resources on critical items that we can control.

What we have concluded is that computers at all levels, including PCs, have potential military applications.  Our own military production illustrates this.  The weapons found in the U.S. arsenal today were built with computers whose performance was below 1000 MTOPS -- in many cases only 500 MTOPS.  The level of computational power used to develop all the bombs in the current U.S. nuclear arsenal, for example, is less than that found today in many workstations.  Our conclusion is that the amount of computing power needed to design and manufacture modern weapons, once you get over a few hundred MTOPS, is not a critical variable.  Other factors -- software design, access to sophisticated manufacturing techniques, experience and test data -- are more important.

For this reason, our 1999 review focused not only on the potential military applications but also on the issues of controllability and technological leadership.  Our conclusion was that the national security interests of the United States are best served by ensuring that the United States retains its technological advantage in the design, development, and production of microprocessors and computers.  As some 50 to 60 percent of revenues for these sectors are generated through exports, the ability to compete internationally is an essential ingredient to the continued viability of these industries.  Today, no nation can remain at the leading edge of technology unless it participates in the global market. The continued generation of revenues from exports provides the needed capital for research and development which keeps these industries at the cutting edge of technology.  This, in turn, ensures that the U.S. military and U.S. defense-related industries will continue to have access to the computer technology they need to maintain our military advantage.   

NDAA Requirements

To date, we have received a little more than 2500 notifications under the NDAA process.   Of this amount, 205, or 8 percent were converted to licenses.  Of those 205 licenses, nine, or less than one half of one percent, have been denied.  A number of other cases were returned to exporters to allow them to assemble the documents needed for the lengthy process of license review, and roughly one hundred and twenty notifications are pending at this time.

To be candid, there are problems with the NDAA process.  In particular, the 2000 MTOPS level  for notification and verification.  Rapid advances in microprocessor and computer performance make the 2000 MTOPS unworkable, and threaten to clog the system with thousands of notifications to which no one is going to object.  This problem will only grow as microprocessor speeds continue to increase.

The 2000 MTOPS level is forcing the Department of Commerce to use scarce investigative resources to conduct a rapidly growing number of post-shipment verifications in Tier Three countries which detract from more important enforcement efforts.  The fact that we are required by law to visit all such computers, including PCs and laptops, forces us to spend time on those that don't matter.  If our investigators were permitted to exercise their discretion, we could focus our efforts directly on any problematic locations we might identify.   Adjusting the MTOPS level upwards for these required inspections will be helpful, but it will only be temporary as chip speed increases and commodity level PCs and laptops grow increasingly powerful.

This matter will become much more important in upcoming weeks, as several manufacturers place tens of thousands of chips with speeds greater than 2000 MTOPS on the market.  We expect to see single chips with speeds greater than 3000 MTOPS enter mass production before the end of the year.  In particular, I would call the attention of the Committee to the decision of at least one U.S. computer company not to market laptops and PCs with performance above 2000 MTOPS in Tier Three at this time, because of NDAA requirements.  It is not possible for PCs and laptops sold in the thousands in retail stores to wait ten days for interagency review.  If these companies sell these commodity computers in Tier III, the NDAA review system would be overwhelmed.  I view this as only a temporary grace period.  More to the point, we must be clear that the NDAA notification process does nothing to prevent militaries in Tier Three countries from obtaining single chip PCs and laptops with performance above 2000 MTOPS if they want them, since these are inexpensive, highly portable and widely available from retail outlets around the world.  


Computer export policy has broad ramifications for the health of the entire American economy and for U.S. national security.  The U.S. is the world leader in information technology and in computer manufacturing, and the industry's future depends on exports.  Ill-advised export controls could put this vital sector at risk with little benefit to national security.  In the short term, U.S. companies would lose sales opportunities to foreign firms.  In the long term, U.S. companies would suffer a significant and perhaps irreversible loss of market share in foreign markets.  The loss of export revenue would adversely affect the ability of U.S. computer companies to fund research and development on next generation technologies, thereby harming their ability to maintain product lines on the cutting edge of technology, including products with strategic military applications.

High performance computers are only one example, although a salient one, of the eroding effect of economic globalization on export controls.  Our approach to this sector is an example of our strategy for enhancing national security in light of the globalization and commercialization of the industrial base which supports our military.  Experts at the Defense Department argue that the US can achieve a net security gain if it properly exploits the globalization and commercialization trends.  Conversely, a strategy dominated by risk-avoidance, or the absence of a clear strategy altogether, will lead to a loss of security, due to gains made by potential adversaries as they take advantage of the global availability of militarily-useful technology.