AMD just announced their 2Q financial results, and there are several items of note.
32nm APU supply has improved. SemiWiki contributors have regularly discussed the importance of a close partnership between design and foundry at leading process nodes. DFM, and significantly, DFY analysis steps are now an absolute must. AMD has the luxury of being the principal customer for GLOBALFOUNDRIES’ 32nm SOI process, and can work closely with them on design optimizations. After a somewhat delayed ramp, the ongoing collaboration between AMD and GF appears to have resolved their 32nm design yield issues.
AMD’s revenue for the quarter fell $180M (~12%) below last quarter’s forecast. The cause was attributed primarily to the 32nm APU Llano (A-family) part number, for which a substantial inventory build did not materialize in corresponding sales in the quarter. Correspondingly, the Accounts Receivable amount payable to GF on the balance sheet increased significantly, as well.
An analyst on the financial call (Stacy Ragson, Sanford Bernstein) asked an insightful question, “AMD signed a take-or-pay wafer sourcing agreement with GF for 2012 – is the inventory buildup a consequence of this commitment, which exceeds current sales volumes? Won’t this require aggressive discounting, and a continued decline in microprocessor ASP?” Unfortunately, management’s answer wasn’t very satisfying.
The key lesson is that design firms need to forecast and negotiate appropriate wafer supply agreements with their foundries. That’s not a new phenomenon, of course, but the transition from 45/40nm to 32/28nm (and soon 22/20nm) has raised the importance. The foundry’s R&D and capital investments required for each new process transition are driven by the revenue projections from these agreements with their customers. AMD may continue to suffer from their 2012 wafer sourcing commitment to GF that exceeds forecasted sales.
Another comment from the AMD CFO was important – “discretionary expense cost controls were put in place late in the quarter”, to improve financial results. In our industry, that typically implies two steps that are typically taken. First, any pending (high expense) tapeouts were deferred past the end of the quarter – but, that’s only a temporary financial move, and the expense will now show up this quarter. Second, and more significantly, hiring activities for future growth (higher-risk) projects are typically “frozen”. Surprisingly, AMD’s headcount during this tough financial quarter rose substantially, with ~4% growth (11,737 vs. 11,265) – that does not seem consistent with stringent expense controls.
If you’ve been in this business for a while, you know belt-tightening steps are typically taken during tough business cycles. The AMD comments on the financial call did not provide much detail on the expense discipline they were pursuing. Rather, their expectation is to “reduce the inventory glut through growth, ensuring that the OEM’s and channel distributors have the parts (and CPU motherboards) they need, while maintaining ASP’s and margins” – that’s a pretty tall order.
There were very few comments on AMD’s plans in the (lucrative) x86 server market – the focus during the financial call was primarily on pushing desktop/laptop Fusion APU parts through OEM and channel sales. Although this was a financial summary, not a product roadmap presentation, that lack of discussion given to the server family was telling.
The (Pretty) Ugly
The worst aspect of the AMD results perhaps wasn’t the disappointing sales, the rising APU inventory, or the high expectations that desktop and laptop sales will pick up with the upcoming Windows8 release.
In January 2011, AMD released their current CEO, with the Board-of-Directors providing the justification that “new leadership was needed, to provide a convincing strategy in the mobile tablet and smartphone space”. Alas, since that time, there has been no communication of that strategy – if we are to assume that this was indeed the reason for Mr. Meyer’s departure, the actions taken since are inconsistent with the actions 18 months ago.
For their part, Intel has communicated (some of) their strategic plans across the entire product space, especially in the high-end server and low-end mobile markets, which are the sources of financial growth for the foreseeable future. They may or may not be successful in the very low-end, but their current strategy is clear.
When a company like AMD is competing against a dominant, proverbial “800-pound gorilla”, the most important characteristic to maintain is credibility – “do what you say you’re going to do, communicate a roadmap that is within your resources and capabilities, and execute to the product plans”. In the Valley, there have been several recent examples of CEO changes, which occurred as a direct result of diminished credibility of (or outright confusion about) the company’s communicated forecasts. Although AMD certainly has had to address short-term issues – e.g., 32nm yield ramp at GF – the industry is still waiting for a strategy from AMD that is consistent with what they indicated was a crucial priority last year, critical enough to pursue a change in leadership. That’s the ugliest news of all.
Just my $0.02.