AMD has just released its quarterly earnings report, and it turned out to be slightly better than expected by both analysts and yours truly: revenue came in at $1.63B and the non-GAAP operating loss decrease substantially, from $363M in Q2 to $148M in Q3.

Despite that, net loss figures were worse than expected by some, at no less than $396M (versus $600M in Q2). But this was partially due to AMD converting its Spansion stake into common stock: because Spansion's value went down further since last quarter, they had to endure a $57M charge related to this operation. All business segments improved, although none have yet returned to profitability; the GPU business, for example, has gone from a $50M operating loss to only a $3M loss.

Gross margins were 41%, a substantial improvement from Q2; ASPs were slightly up, primarily because of an improved product mix. However, much of the margin improvement simply came from higher fab utilization: for companies like AMD and Intel which have their own factories, gross margins are not calculated based on a wafer costs, but rather based on the costs of running the fabs.

On the ex-ATI front, "Acquistion Related Charges" seem to be stuck at roughly $76M/quarter, with AMD predicting more of the same again for 4Q and no word on how many quarters after that. This puts a slight damper on the still good news that GPG (Graphics Product Group) and CE revenues were up substantially and nearly at break even on their own. . . if it weren't for the $76M quarterly charge drizzling on the parade. Still, the bullish optimism in both existing mid-range HD 2xxx sales and forward-looking expectations for 55nm RV670 adoption was clearly evident in the executive team's reports.

AMD's cash position decreased by only $66M; however, this includes short-term investments and marketable securities. Thus, AMD's stake in Spansion moved on the balance sheet, from being a line of its own to becoming part of that figure. This makes that number a rather weak indication of sustainable cashflow; instead, it might be more interesting to look at Shareholder's Equity (which corresponds to liabilities minus assets), which decreased by $347M from $4.47B to $4.12B. Even this is sadly far from a perfect indicator of financial health, however.

Overall, AMD's quarter was a solid improvement from the previous two quarters, and the prospects of ramping up 45nm in the first half of 2008 (rather than 2009, as some had previously thought!) should give a good foundation for the future, making the gigantic operating losses of Q1 and Q2 a thing of the past. Fab30 (90nm/200mm) will shut down by the end of the year to be converted into Fab38, and this should keep margins in check as all production migrates to 65nm - however, it is hard to predict how Intel's own 45nm migration will affect things.

While the worst may be over in the short-term, many challenges still lie ahead for AMD, including how to find the cash necessary to finance its 45nm process migration and ramp and keeping the balance sheet in safe mode. Nehalem poses a clear challenge to its server business, while Phenom seems unable to beat Intel's 45nm Yorkfifeld (even on a per-clock basis) in single-threaded and integer workloads. But AMD's technical position will arguably remain better than at Conroe's release last year, and thus everything depends on execution now.