Historical Overview of Kendall Square Research Corporation
Kendall Square Research Corporation (KSR) was founded in 1986 by Henry Burkhardt III and Steve Frank, another computer designer. KSR was the third computer company co-founded by Burkhardt—the first two having been Data General Corporation and Encore Computer Corporation. At KSR, Burkhardt chose a radical design for a supercomputer that would make the machines attractive for both government and university laboratories and for commercial users. The company delivered its first machine to the Oak Ridge National Laboratory in the fall of 1991, and Kendall Square Research reported that it reached break-even volume before the end of 1992.
From its beginning KSR funded its development and the design of its products through private placements of its equity and debt securities and an initial public stock offering in April 1992. A subsequent stock offering in April 1993 brought the total financing of the start-up to about $150 million—more than $80 million of which had come in two offerings of stock to the public. Shares in the first of the public offerings had more than doubled in price in less than 18 months. By October 1993, the market capitalization of the shares was about five times estimated 1993 sales and 45 times estimated 1993 earnings.1
From the time of its first sale, some critics questioned the criteria used by KSR to determine when revenue would be recognized. The company recognized revenue on product sales upon written customer acceptance of the product. Acceptance often occurred before the configuration of a particular system was finalized or any payment had been received from a customer. Kendall Square Research fully disclosed its policy on revenue recognition in its financial reports to shareholders. (See Exhibit 1 for excerpts from the 1992 Kendall Square Research Corporation Annual Report to Shareholders.)
Kendall Square Research2 recognized at its founding that a key to its success would be its ability to deliver more computing power at lower cost per computation than its competitors. Its approach was based on linking many low-cost minicomputers and dividing the computer task among them, rather than building larger mainframe computers or trying to improve programming and software to wring more performance from the computer architectures that were available. The concept of massively parallel processing had first surfaced in the 1980s, and several competitors felt that it had tremendous potential.
A large-scale parallel computer was potentially faster and less expensive than traditional mainframes or available supercomputers. In addition to lower cost per computation or transaction, large-scale parallel computers were theoretically scaleable; bigger and more powerful machines could be assembled merely by adding processors and the memory and input and output devices to support them. However, as processors were added, programming became more difficult and cumbersome.
Kendall Square Research Corporation and Supercomputing
Kendall Square's founders believed that a highly successful parallel computer could succeed only if it could take advantage of the library of applications and languages that were already standard in the mainframe and supercomputer worlds. Their goal was to develop a standards-based multi-user system that could run multiple applications simultaneously, so that a sufficiently powerful system could serve an entire technical or business enterprise. Their ideal system would even be able to run scientific and business applications at the same time.
The technology that was needed to develop such a computer did not exist in 1986 when KSR was founded. For six years Kendall Square scientists and engineers invested in chip development technology that they hoped would turn their insight into a working computer, By 1992 they were sure they had succeeded, They had developed a large scale parallel computer which combined the power of massively parallel, distributed memory machines with the familiar shared-memory programming environment of conventional mainframes. The KSR1 family of computers was designed for high performance computing requirements typical of scientific environments, for decision support and complex database query applications, and for on-line transaction-intensive environments such as automatic teller machines or airline reservation systems.
The KSR1 family of systems scaled from 16 to 1088 processors. A $975,000 model with 16 processors could handle 1,200 transactions per second, a measure of commercial computer speed. Scalability allowed users to add computer resources in incremental and cost-effective steps without changes in software and without performance degradation.
Therefore, a customer could add to an installation at a later date without the need to replace all software or to reprogram operating systems. While the scaling up of processors in massive parallel processing computer systems was not fully reliable, reports from users of KSR machines seemed to support the Company's contention that they scale up more effectively than those of some other competitors.
In 1993 supercomputers with massively parallel processing were more commonly found in scientific applications than in commercial applications. One reason for this was the tendency for supercomputers to crash as users pushed their limits. In scientific applications, users were often willing to sacrifice reliability for performance, but most commercial applications required computers to work reliably for months at a time.
The business market for supercomputers had been estimated to be about $31 billion—far larger than the scientific and research market, which had been estimated to be about $4 billion.3 In 1992, Kendall Square signed contracts with Neodata, a direct-mail company working with Electronic Data Systems (EDS), and AMR, the parent company of American Airlines, to work together to develop systems using KSR computers.
Revenue Recognition at Kendall Square Research Corp.
Critics of the revenue recognition practices used by KSR claimed that the company was far too liberal in what it called a sale of a machine or system. They cited cases in which laboratories had ordered or received equipment which was subsequently accepted but for which there was no prospective funding, or for which research grants had been requested but not yet granted. In such cases there was no assurance that the research grant would be received and, even if it was, there could be significant delays before the customer would pay KSR.
Other analysts thought that the liberal accounting practices used by KSR were not uncommon in the computer industry, but such practices were less visible in the financial statements of more mature companies such as IBM or Digital Equipment Corporation. These analysts pointed out that terms requiring payment in six or nine months after a customer had received a grant from a government agency—often very slow payers—were not unusual. Such terms might be appropriate for the typical university or research laboratory that were the base of early KSR users.
Terms of Kendall's supercomputer deliveries were often extremely attractive to researchers. For example, Edward Lazowska, chairman of the department of computer science at the University of Washington, said that it has two KSR1 computers with a total of 60 processors. He said, "We paid $1 million for half of it, and the rest was a loan of equipment against future grants." Kendall . . . treated the entire shipment as a sale.
William Goddard, a physicist at the California Institute of Technology, said he paid for 32 processors in September 1992 using a National Science Foundation grant. He liked it so much that last spring he wanted to double the size of the system and applied for more grant money. Kendall shipped him the computer immediately, but he and Kendall are still waiting for the grant award. "I'm committed to buying it," he said. "I am applying for two very large grants,"4
Other contracts included one for $1.5 million that could be canceled if the second of two phases was not completed. Still others involved distributors that had the right to cancel the contract if a customer could not be found.
A second sales practice added to the concerns of critics of KSR's accounting policies. Some users purchased machines with fewer processors than were eventually thought to be needed. In these cases the processors purchased were booked as revenue and accounts receivable, but KSR had delivered additional processors to the customer for installation and use by the customer. The
additional processors remained on KSR's balance sheet as inventory pending the customer's decision to keep the processors, apply for grants to buy them, or to otherwise make arrangements to pay for the loaned processors.
As a result of these practices, revenues recognized had exceeded cash collected from customers by significant amounts. Some critics thought this indicated that revenue was being recognized too early. Other observers thought this was normal for a new, growing company in the process of ramping up sales.
Of additional concern was the practice of KSR giving research grants to some users of its systems. Such grants had been reported to range between $5,000 and $50,000 or more. Henry Burkhardt was reported to have agreed that grants ". . . sometimes do run into substantial amounts, but are legitimate expenses to foster software development, not disguised sales incentives.. . ."5
In August 1993, The Wall Street Journal reported:
WALTHAM, Mass Kendall Square Research Inc, is "quite comfortable" with analysts' estimates that it will earn 45 cents to 50 cents a share for the year, President Henry Burkhardt said.
Mr. Burkhardt also said he is comfortable with estimates that revenue for the year will top $60 million, up from $20.7 million. And he said he is comfortable with third-quarter earnings estimates of 9 cents to 13 cents a share. For the year-earlier quarter, Kendall Square reported a net loss of 29 cents a share on revenue of $5.3 million.6
1. Evaluate the revenue recognition policies used by Kendall Square Research.
Do they conform to generally accepted accounting principles with respect to
a. The timing of revenue recognition?
b. The amount of revenue recognized?
c. The matching of costs and expenses to revenue?
2. The rapid growth in sales expected by Kendall Square Research managers will strain the company's ability to finance its expansion. How, if at all, should this fact affect the ways in which management chooses accounting principles and a reporting strategy