EDGAR CORRESPONDENCE
March 10, 2006
Ms. Kathleen Collins
Accounting Branch Chief
Division of Corporation Finance
United States Securities and Exchange Commission
100 F Street, N.E.
Washington, D.C. 20549
Re:Cray Inc. — file No. 000-26820
Dear Ms. Collins:
We have been in correspondence with you concerning certain of our Exchange Act filings since your first letter dated August 18, 2005. This letter responds to your letter of February 3, 2006, in which you asked for further information following our most recent response letter of December 16, 2005. Our numbered responses correspond to the number of your comments and questions.
1. Question
We note your disclosures regarding Accounting for Income Taxes under Critical Accounting Policies and Estimates. Tell us how you considered the Interpretive Guidance in SEC Release 33-8350 for your Critical Accounting Policies disclosures. Please note that your discussion of critical accounting policies and estimates should serve as a supplement to the financial statement footnotes rather than reiterate the information included in the footnotes and should describe how estimates and related assumptions were derived, how accurate the estimates and assumptions have been in the past, and whether the estimates and assumptions are reasonably likely to change in the future. In this regard, your disclosure should include an analysis of specific sensitivity to change based on other likely outcomes. Your disclosure should also include quantitative as well as qualitative information when information is reasonably available.
Response:
We have noted the Interpretive Guidance for the MD&A as a whole and the portion related to critical accounting policies and estimates. We will enhance our disclosures in future filings, particularly with regard to critical accounting policies and estimates, starting with our 2005 Form 10-K. In the 2005 Form 10-K, among other things, we will enhance the disclosure on income taxes and our judgment with regard to the valuation allowance for deferred tax assets.
2. Question
We note your response to our Prior Comment No. 1, including your description of the positive and negative evidence relied on in determining that the majority of the valuation allowance should be reversed. Further we note that you “...believed that the positive evidence supporting the reversal of our valuation allowance outweighed the negative evidence and was more objective in nature than the negative evidence.” It is not clear why you believe the positive evidence outweighed the negative evidence and why you viewed the positive evidence as “more objective” in determining it was more likely than not that you would be able to realize your deferred tax assets. Explain why you believe the positive evidence outweighed the negative evidence and why the positive evidence is more objective. For example, why is your expectation of meeting certain revenue targets more objective than the risks associated with new product development or the risk that the U.S. Federal government would spend less on supercomputers? Please be as detailed in your response as possible to support your conclusions.
Response:
As noted in our December 16, 2005 response, we considered both positive and negative evidence in evaluating whether it was more likely than not that the valuation allowance on certain of our deferred tax assets (principally loss carry-forwards generated in the U.S.) should be reversed. Our policy is to consider actual operating performance, and in particular our most recent results of operations, to be the most objective and verifiable evidence of how we will perform in the future. We therefore accorded the most weight to our most recent performance. Clearly, past operating results are a measure of what has been accomplished while forecasts, plans and potential risks are best estimates of what is reasonably anticipated to occur, with the attendant risk that future performance may fail to match expectations.
Other evidence such as shorter term forecasts were given more weight than longer term financial models because they necessarily involve less dependence on future contingencies. Our management, when developing annual plans, considers key risk factors and their probability of occurrence when finalizing the plan. If a material adverse risk was considered probable of occurring, and was not considered in formulating the plan, it would not be an effective plan. In 2002 we achieved or surpassed the goals in our plan. In 2003 we overcame several potential risks, including the roll-out of the Cray X1 product, and outperformed our revenue plan by 12% and our operating profit plan by 54%. The same key senior managers developed and approved a profitable plan for 2004.
In order of assigned weighting, with past performance considered the most objective and verifiable, we considered the following four criteria to be the most important in making the deferred tax asset valuation determination under the requirements of SFAS No. 109:
| | • | | Actual 2003 performance; adjusted U.S. pre-tax income of $32.5 million, and actual 2002 performance; $7.0 million in adjusted pre-tax income. At the end of 2003, we had just completed eight consecutive profitable quarters and had achieved U.S. cumulative pretax income over a three-year period, excluding unusual items and goodwill amortization. See answer to question 4 in our December 16, 2005 response. |
| |
| | • | | The remaining time before first expiration of tax loss carry-forwards, 7 years in the future. |
| |
| | • | | Our financial ability to carry out a profitable future strategy. Our financial position was greatly improved from prior years, with $74 million in cash and short-term investments and almost no debt. |
| |
| | • | | Our 2004 financial plan with pre-tax income of $26 million which considered all known risks and opportunities that existed at that time; the plan was presented to and approved by our Board of Directors. |
In considering the negative evidence, the only item that was both objective and verifiable was the losses incurred prior to 2002. While considered important, as this was how the level of the tax loss carry-forwards was created, they were more than two years in the past and occurred when we were in the “development stage.”
Other potential negative evidence was much less verifiable:
| | • | | While there is always a risk that customers, including the U.S. government, may spend less on high performance computing, most external forecasters predicted continued growth in government spending. At the time, we were not aware of any significant changes in the U.S. government appropriations targeted to high performance computing for fiscal year 2004. |
| |
| | • | | Although there are always risks associated with product development, our development reviews at the time of the decision did not foresee a significant variance in schedule from what was inherent in the 2004 plan. Our judgment, based on past experience, was that the upgrade from the Cray X1 system to the Cray X1E system was not a significant technical upgrade and would be accomplished on schedule. The Red Storm development program during 2002 and through 2003 remained substantially on schedule. About 20% of our 2004 plan, or $60 million, was predicated on new products which were to be sold. We actually needed no revenue from this source to be profitable in 2004. We believed at the time that customers would continue to buy the Cray X1 system, particularly internationally, and then upgrade later to the Cray X1E system, because this was believed at the time to be relatively simple and cost effective for the customer. Our expectations were not realized because of a subsequent uncharacteristic defect by our ASIC fabrication vendor that we did not detect until too late; this caused approximately four-months delay in the availability of the chip and adversely impacted our revenue. However, our judgment was based on the best information available at the time. |
| |
| | • | | Risks associated with our competitors’ potential impact on revenue and margins were considered in the 2004 plan. Overall gross product margins were expected to decrease by about 6%. Part of this was due to an expected change in product mix but also the expected margins on Cray X1 sales were forecasted to decline by about 5%. Competitive risks were considered to be offset by Cray building a stronger sales team, particularly internationally, and competing for more deals. During 2003, Cray hired a new sales and marketing executive, Peter Ungaro, and he hired a new European sales leader. Five marketing and fourteen new sales people were added in 2003 to a relatively small base. We anticipated that product revenue from our international subsidiaries would more than double in 2004. |
The following responds to your question as to why revenue targets were more verifiable than the impact of identified risks associated with new product development or the risk that the U.S. government would spend less on supercomputers. We did not mean to convey that our revenue targets were more objective than the risks. As indicated above, annual plans including our 2004 plan are much more than arbitrary sales targets. They incorporate all known opportunities and risks as identified by our sales force and management team. Our plan was to increase revenue to $300 million, of which $191 million was for products, excluding development project revenue of $58 million. This was a 23% increase over the $155 million in 2003. About $90 million of the 2004 total product revenue expectation was in backlog (note that $10 million of our $100 million backlog was for revenue expected to be recognized after 2004) and $51 million was expected from service, where forecast risk was minimal. The remaining forecast, $159 million, was expected to come from new bookings. The new bookings/revenue requirement was believed to be reasonable based on a
listing by customer of significant 2004 opportunities used in preparing the plan. During 2003 about $125 million in new product revenue came from bookings in that same year. Further, we expected to introduce a second product in 2004, the Cray XT3 system, and international had a richer pipeline of opportunities.
In our judgment, the positive evidence as a whole was more objective and verifiable and outweighed the less verifiable negative evidence at the time of the decision. We therefore concluded that it was more likely than not that we would be able to realize our deferred tax assets.
3. Question
Your response appears to suggest that your analysis relied heavily upon your 2004 revenue forecast. Tell us why revenues, and not taxable income, were the focus of your analysis. Also with regard to your revenue projections, address the following:
Response:
We did not intend to suggest we placed an undue heavy reliance on the 2004 revenue plan. As described in our previous responses to your comments, we believed that the most objective and verifiable information, and the most important, was our actual pretax income results in 2003 and 2002. The 2004 plan was used more to validate that we anticipated a continuation of positive taxable income, which we believed would continue in ensuing years.
We considered taxable income, not revenue, in our analysis for determining whether it was more likely than not that we would generate sufficient income before taxes to reverse our previously established valuation allowance. However, we believe that our revenue performance is the most significant driver of our taxable income. As noted in our response to Comment No. 2 above, we considered the impact of our research and development efforts (expense-driven) in our evaluation of positive and negative evidence. We believe that our operating expenses (research and development, sales and marketing, and general and administrative costs) are generally more predictable and more controllable. Any shortfall in our plan would likely come from missing our product revenue target and associated gross margin target, and as such, we believe that a focus on revenue was and is appropriate.
a. We note that your average sales cycle ranges between six months and two years. We further note your practice of deferring revenues until customer acceptance. Explain how you were able to conclude that you would achieve your revenue target of $191 million of product sales in fiscal 2004 when there was only $100 million in backlog at the end of 2003. In this regard, tell us how you believed you would be able to successfully achieve acceptance on the $100 million in backlog, and also sell, deliver and achieve acceptance on an additional $91 million in product sales in less than one year.
Response:
Our sales cycle begins long before the products are available and released for shipment to customers. For some customers it begins during the early stages of product concept. Our tentative product roadmap and plans are shared as part of the process. As such, we have early discussions with customers about our planned product introductions including technical specification targets, estimated pricing, and timing of availability. So at any time we have customers in various stages of the sales process, even before a product is available to ship; some are in contract negotiations; some are in award phases; others are in RFP and benchmarking phases; and some may also be in system exploration phases. The award phase to contract is typically three months or less. Depending on where the customer is in the RFP and benchmarking phase, the time to contract completion can be as short as three months or as long as twelve months. For the Cray X1 system, contract to revenue recognition time was relatively short, as little as two to four months, in many cases. At the time we were evaluating whether to reverse the tax allowance we had customers in various stages of the sales process, with several significant opportunities in the award and late benchmarking phases in addition to our backlog. For example, the following customer opportunities were being pursued at the time of judgment: we signed an agreement for $43 million with the Korea Meteorological Administration (“KMA”) in late April; were notified of a $10 million award from a major research center, Pittsburgh Supercomputing Center, in May; and also in May, we received another $12 million procurement award for another major research center, Oakridge National Laboratory, awaiting final Department of Energy approval. In addition, we were preparing a $20 million bid for an opportunity in China.
Although, as it turned out, except for the China transaction, we received all these orders in 2004, we were able to recognize revenue only on the Oakridge transaction in 2004 due to a combination of unanticipated events including delayed orders, contract clauses and product development problems.
We define backlog based on the awarded contract values and the expected revenue recognition (which generally includes acceptance clauses) dates. This does include judgment as to the timing of expected revenue recognition. As noted in our response above to Comment No. 2, in order for us to achieve our plan of $191 million in product revenue, we had to book an additional $159 million in 2004 and recognize revenue on these bookings (note that the $91 million in your question should be $159 million because that backlog number included development project backlog). We believed that this “bookings to revenue rate” was reasonable given our prior performance in 2003 wherein we booked and recognized revenue on about $125 million, particularly when considering the pipeline of later-stage opportunities we had in early 2004.
While our plan was to achieve $191 million in product revenue excluding development project revenue ($300 million total revenue), we would have been profitable with revenue well below those numbers. Our plan was for $26 million in pre-tax profit in 2004. Our Cray X1 product gross margin was expected to be about 45%. Based on this, the breakeven total revenue for Cray would have been approximately $242 million ($300 in total revenue less $58 million, or $26 million pre-tax profit divided by 45% Cray X1 margin). Product revenue could then be as low as $133 million ($191 million less $58 million), and below the 2003 product revenue level of $155 million. Required new bookings and revenue to achieve breakeven would be about $101 million ($159 million less $58 million), well below the $125 million achieved in 2003.
As noted above in response to Comment 2, we considered the 2004 plan as less objective and verifiable evidence than our actual pre-tax operating results in 2002 and 2003.
b. We note your 2004 revenue target for the Cray X1 system was $100 million. We further note your statement that ‘following the first quarter of 2004, U.S. government purchases shifted unexpectedly away from Cray X1 systems...’ Tell us why you believed you would experience such strong sales from a product that was being phased out — to be replaced by an upgraded version with nearly triple the performance within a few months time.
Response:
A significant portion of our product backlog at December 31, 2003 was for Cray X1 systems, which were expected to be recognized as revenue in 2004. Several of the large expected orders in the first half of 2004 were bid as Cray X1 systems: KMA, a $43 million total order was anticipated to be a Cray X1 system initially, then upgraded to Cray X1E system, of which $12 million was anticipated in 2004 revenue; Oakridge National Laboratory was expected to purchase a $12 plus million Cray X1 (we believed $25 million was possible for the year) was delayed but booked late in the year; an anticipated transaction in China of $20 million, was later lost. We were also anticipating orders from various Department of Defense sites. All but the Chinese opportunity were later booked in 2004, but only the revenue for the Oakridge National Laboratory transaction was recognized in 2004.
We believed at the time that customers would continue to buy Cray X1 systems, particularly internationally, and that some of our current Cray X1 system customers would further expand their systems, then upgrade later to the Cray X1E systems. We believed customers would upgrade to the Cray X1E because this was believed to be an easy upgrade performed at customer sites which would be cost effective for the customer. We lowered the effective price of the Cray X1 system to make the initial purchase with a later upgrade to the Cray X1E system more enticing; the impact of this pricing strategy was reflected in our plan. This belief subsequently proved incorrect, but it was made based on the best information available at the time. Our management had nothing to gain and much to be lost (in credibility and performance bonuses) in signing up for an unrealistic plan. Management believed the plan was achievable at the time it was adopted and approved by the Board. No bonuses were paid to senior management as a result of 2004’s poor performance compared to plan.
c. As noted above, it appears that your 2004 revenue projection of $300 million relied heavily on sales of the Cray XT3 systems and Cray X1E systems and upgrades. We further note that these products were not
expected to be released until the second half of 2004. Again, given the long sales cycle and your practice of deferring revenue until customer acceptance, explain how you determined that you would be able to sell, deliver and achieve customer acceptance on sales of these products in six months or less.
Response:
As mentioned above, our sales cycle often begins long before products are shipping in the market. As such, we have discussions with customers about our planned product introductions and upgrades that include technical specifications targets, estimated pricing, and expected time of availability.
Outside of the Red Storm project revenue associated with Sandia, anticipated to be about $46 million in 2004, we expected $30 million in the Red Storm derivative product or the Cray XT3 in 2004. We were well along in the sales cycle to be awarded a $10 million order from Pittsburgh Supercomputing Center. Oakridge National Laboratory was working toward obtaining funding to purchase a large system ($16 million) with requirements similar to the Cray XT3 specification. The market for massively parallel supercomputers was well established and large, and systems existed in the market that were similar, but deemed inferior, to the Cray XT3 system. At the time of our plan and deferred tax asset valuation judgment, we believed that the project was on track, and that we could get sufficient orders and ship prior to the fourth quarter in order to attain customer acceptances by year-end.
Our judgment was that the upgrade from the Cray X1 to the Cray X1E was not a significant technical upgrade (the upgrade required a new CPU module that could be inserted at customer sites) and should be accomplished on schedule. The upgrade required very little software changes. The success in 2003 of the Cray X1 line gave us confidence at the time of the plan that we could meet our combined Cray X1/X1E goal of $152 million. Cray X1 sales in 2003 were approximately $144 million. See answer to Question 3b above for more information regarding customer opportunities.
While we publicly acknowledged the risk of product development delays or product acceptance, we believed that our product development activities were progressing reasonably to meet our planned product introduction. We also believed that we would be able to achieve product acceptance, and would therefore be able to recognize revenue in a timely manner. Our prior experience with the Cray X1 product was that we were able to obtain acceptance of our products in a timely manner (2 to 4 months) and we believed that this experience would continue.
d. On page 18 of your Form 10K for fiscal 2003 you indicate that you expected to offer favorable pricing on sales of the Cray X1 in 2004. How did this impact your analysis — specifically your projected margins and ultimately, your projected taxable income?
Response:
We incorporated into our 2004 plan the effect on product margins of offering favorable pricing. Our planned total gross margin for 2004 was just under 36%, which was below the reported 42% overall margin we experienced in 2003. In 2003 the average margin on the Cray X1 was about 50%. Our planned 2004 margin was 45% for the Cray X1. In addition, we took into consideration some higher margin deals that were already in backlog, as well as expected new orders.
Consistent with the guidance of paragraph 103 of SFAS No. 109, we believe that the positive evidence (positive taxable income for two years, planned and expected taxable income for 2004 supported by reasonable risk adjusted assumptions) was of sufficient quantity and quality to counteract the negative evidence (historical losses during the development stage). Accordingly, we believe based on the weight of all available evidence at the time, the reversal of the deferred tax valuation allowance was appropriate.
Please contact the undersigned if you require additional information or desire to discuss further any of the foregoing responses (phone: 206-701-2168; facsimile: 206-701-2218; email: brianh@cray.com).
| | | | | |
| | Yours truly,
| |
| | /s/ Brian C. Henry | |
| | Brian C. Henry | |
| | Executive Vice President and Chief Financial Officer |
| |
| | | |
| cc. | | Peter J. Ungaro, Chief Executive Officer and President
Kenneth W. Johnson, Senior Vice President and General Counsel
Cray Inc. Audit Committee
L. John Stevenson, Stoel Rives LLP
Andrew L. Read, Deloitte & Touche LLP
David C. Lee, Peterson Sullivan PLLC |
