Supercomputers are powerful tools for doing complex calculations and a necessary resource in advanced scientific research. The University of Minnesota is among the nation's leading institutions that provide supercomputing services to their scientists. Its achievement in this field--supported, of course, by substantial public funding--is noteworthy and of significant benefit to the state.
But the University's approach to obtaining supercomputing services has prompted concern and even controversy. It involves a University-created private, for-profit corporation, known as the Minnesota Supercomputer Center (MSC), that operates with considerable secrecy. MSC provides services to private customers as well as to the University, and it claims that information about its finances and operations is a "trade secret," and, therefore, not subject to public examination.
In response to recent controversy, the Legislative Audit Commission directed us to evaluate certain aspects of supercomputing services at the University. Since MSC was unwilling to give us unfettered access to its records, the commission decided to defer asking for an examination of MSC's internal operations until it received a more general report on the relationship between the University and MSC. We based our evaluation on information we obtained from the University.
The commission requested that the evaluation focus on the following questions:
To answer these questions, we interviewed University officials, faculty, and staff, reviewed University records, and talked with national experts on supercomputers. We emphasize that we did not examine records at MSC.
Most of the information we obtained came from the Minnesota Supercomputer Institute. The Institute (MSI) and the Center (MSC) are often confused because of their similar names and because they occupy the same building. They are, in fact, separate and quite different organizations. MSI is a unit within the University. It is the organizational vehicle through which the University purchases computer time from MSC and allocates it among University users. MSI also provides support to the twenty-nine research fellows who are the University's major users of MSC's services.
We examined the structure and activities of the Institute, and we have several recommendations to improve its operations. However, the main focus of our concern is on MSC, rather than MSI.
Our overall conclusion is that the University has not achieved an adequate level of accountability in its approach to obtaining supercomputing services. The source of the problem is that the University created an organization--the Minnesota Supercomputer Center--whose legal status is uncertain. Standards for such a government-created, private organization are not well established, and the University has added to the confusion in the way it has interacted with MSC.
The University does not hold MSC to the same standards of accountability it applies to University departments and other organizations it creates. Of particular importance, it allows MSC to operate with extraordinary secrecy, and without the kind of outside oversight other units of the University, and other supercomputer centers across the nation, routinely experience.
On the other hand, the University does not treat MSC like a private company either. It does not have an arms-length, business-like relationship with MSC, but rather allows MSC to operate with benefits and privileges it does not provide other companies selling services to the University. The University has taken several steps during the course of this audit to clarify MSC's status. Despite these steps, we believe the University does not treat MSC as a truly independent corporation.
We emphasize, however, that while the structure of the relationship and MSC's ambiguous status make weak accountability more likely, they do not make it inevitable. The University could further strengthen accountability in its relationship with MSC if it wanted to, but the University does not think there is a significant problem. Based on what they told us, University officials are mostly satisfied with the relationship. They talked largely about the benefits it brings to the University and state.
The University created MSC in 1982 as a private, for-profit corporation. At the time, officials argued that a private company could sell services to private customers, take advantage of certain tax provisions unavailable to the University, and provide supercomputing services to the University at an economical price. They also said that the arrangement would offer opportunities for research cross-fertilization that would benefit both the University and private business.
At the request of the University, the University Foundation (a private, non-profit fund-raising organization) became MSC's principal owner. The Foundation controls five of nine MSC board positions, but it is the University's central administration that has directed MSC's development and provided it with essential financial support, not the Foundation.
In addition, despite MSC's incorporation as a private, for-profit company, it has always been financially dependent on the University and, ultimately, the state. MSC has repeatedly turned to the University and/or the Legislature for loan guarantees and other financial support in order to make large capital investments. For example, a University-leased Cray-1 computer was transferred to MSC in July 1983. In 1985, the University advanced MSC $4.775 million to purchase a Cray-2, stipulating that the money would be paid back in computer services by mid-1986. In December 1985, the University extended a $5 million line of credit to MSC, which MSC has drawn on several times between 1985 and 1992. The University, the state, and the city of Minneapolis financed the remodeling of a building for MSC in the Minneapolis High Technology Corridor. In addition, MSC has repeatedly gone to the University and the Legislature to ask for increases in financial support and preferential tax treatment.
Despite its history, officials at both MSC and the University argue that MSC should not be considered a public entity (not even a "quasi-public" entity). They argue that it should be accountable only as a private vendor of services to the University. But we found that the University does not, in fact, treat MSC like an independent business.
For example, the University pays MSC for services in a lump sum at the beginning of the contract year. The University has granted special purchasing privileges to MSC and not collected sales taxes on the transactions. The University has historically charged MSC less-than-market rental rates for space, and the University allowed MSC to pay its "rent" with supercomputer time that the University would have probably received anyway under a previous agreement. The University continues to charge MSC a less-than-market rental rate under the new lease agreement.
We also found that in 1987 a complex series of financial transactions allowed the University to relieve MSC of a $3.7 million obligation and provide MSC with $800,000 of additional capital without approval from the Board of Regents and without public disclosure. The transactions began when MSC made a "refund" of $3.7 million to the University. Though we cannot be sure without having access to MSC's records, we think MSC made the payment with money it borrowed from the University. We know that a few days before the payment was made, MSC borrowed $4.7 million from the University. Whatever the source of the money used to pay the University, the University deposited the payment in a central reserve account. Interestingly, several weeks later, MSC was able to pay off its debt to the University when the University transferred $4.5 million to MSC in return for a special issue of "preferred stock." In short, MSC was relieved of its $3.7 million obligation and provided with $800,000 in additional cash.
The "refund" transaction was classified confidential until we asked that it be made public. Apparently, the Legislature, the state Department of Finance, and the Board of Regents have not been aware of the transaction until now. In addition, the Board of Regents did not approve the University's $4.5 million purchase of preferred stock in MSC. In fact, the University was unable to provide us with any documentation to show who authorized the purchase. The refund and stock purchase transactions demonstrate why the relationship between the University and MSC lacks adequate accountability.
The University argues that it ensures accountability by controlling four MSC board appointments. But the University has allowed one of its positions to remain vacant for over a year, and, since April 1986, the University has used another of its positions to put MSC's president on the board. Therefore, currently the University has only two board appointees who are in any way independent of MSC management, and both are University senior vice presidents with many other demanding responsibilities. Since most information these board members receive about MSC's operations is considered a "trade secret," it is not provided to other University officials or to the Board of Regents for independent analysis. The only exception occurred earlier this year, when the Chair and Vice-Chair of the board were orally briefed on the proposed $32 million contract between MSC and the University. They were told the approximate effective rate the University would be paying under the proposed contract. But they were not allowed to share this information with other board members or subject it to independent analysis.
Both of the University vice presidents who are MSC board members told us that they are basically satisfied with the quality and quantity of information they receive about MSC's operations from management. They also strongly asserted to us that the information shows that the University "is getting a good deal" in its relationship with MSC. However, we are not convinced that they have had access to information that convincingly shows the University is receiving a good value for its service payments. For example, they were unaware of the rates that commercial clients are paying for the preferential services they receive. Also, the only analysis of University supercomputer rates that we saw is flawed.
Based on the limited information we had access to, we think assertions about the "good deal" the University is getting from MSC are open to question. But, we cannot publicly discuss the basis of our concern because the information involved is classified a "trade secret" by MSC and the University, and we are legally bound to respect that classification.
Based on the interviews we conducted, MSC appears to be the only national supercomputer center that refuses to disclose information about its finances and the services it provides. Additionally, almost all other centers use various forms of outside review to ensure that their resources are being used effectively.
We also observed that Minnesota's approach to obtaining supercomputing services has not facilitated interactions between University scientists and private industry users of the machines. Most of the research fellows we talked with noted that MSC prevents them from having contact with MSC private customers, even though the researchers are housed in the same building with MSC. In fact, MSC is currently building a wall that will further separate its operations from University researchers, even though the researchers have protested. Also, neither MSC or the University has facilitated the development of high-tech businesses that were part of the supercomputer initiative's original objectives and the purpose of the state's contribution of $5 million for the supercomputer center building.
Though we do not endorse the structure and arrangements the University has created, we do not call for the University to end its relationship with MSC. On the other hand, the University should not, in our view, maintain the relationship simply to keep MSC in business if there is a better alternative, especially one that would foster closer collaboration between supercomputer service providers and users. We urge the University to more objectively assess whether or not its approach to obtaining supercomputing services is the most cost-effective alternative and best suited to serve University researchers.
If the University maintains the existing arrangement, it should take steps to strengthen accountability and help ensure that public resources are being appropriately and efficiently used. Its first step should be to provide more information about what the University is getting in return for the large sum of public money it provides to MSC. It is inappropriate to ask the Regents, the Legislature, and the public to trust that the University is getting "a good deal." MSC's legitimate trade secrets can and should be protected. But the University has accepted and participated in a level of secrecy about MSC financial operations that is unnecessary and contrary to good public policy. At a minimum, Regents should be given information that would allow them to independently analyze the price of MSC's services and make comparisons with alternative providers.
In addition, the University should better use its positions on the MSC board. It certainly should not allow positions to go unfilled for long periods of time. Nor should it use one of its positions to appoint MSC's president to the board. The University should appoint to the MSC board at least one distinguished researcher familiar with supercomputing services. In our view, the only justification for MSC's existence is to provide services to University scientists. It makes sense, then, to have at least one University scientist involved in supercomputing on MSC's board. Currently, University scientists and even MSI research fellows are not allowed to attend MSC board meetings. In addition, the University should consider appointing a Regent to the board.
We also think that, as a condition of its support, the University should require MSC to undergo periodic review by an outside panel of supercomputer and scientific experts. Peer review is a widely-accepted professional practice in the public and private sectors. And, again, MSC's legitimate "trade secrets" can and should be protected.
As noted earlier, we also examined the operation of the Minnesota Supercomputer Institute (MSI). Our major concern is the way MSI allocates supercomputing resources among University researchers. We found that some large allocations of supercomputing time have not received adequate review, particularly allocations to MSI's director. Therefore, we recommend that the University strengthen accountability for the allocation of supercomputer time. Proposals that involve large amounts of time should receive objective review by experts able to judge their merit. Also, the percentage of total University supercomputing time awarded and used by University researchers should be routinely disclosed and circulated among University scientists.
Finally, we think it is important that the Legislature take action to define the status of private and "quasi-public" organizations created by government. So-called "public/private partnerships" have expanded in recent years, and many have undoubtedly produced considerable benefit. But uncertainty persists about how to make such partnerships, and the organizations they create, accountable.
We think the history of MSC shows the need for the Legislature to set standards and expectations for all organizations created by government, including organizations created by the University.
The ambiguous status and weak accountability associated with organizations like MSC create a significant potential for abuse. We think the Legislature should set minimum standards of openness and public accountability for these organizations, including disclosure of audited annual financial statements.