On April 19, 2012, Altera reported that sales had dropped 16 percent compared to the previous quarter, and a massive 28 percent from a year ago.
Shut up shop. Go home. The party is over.
So what's happening on our All Programmable Planet?
Well, it's not the Perfect Storm that struck when the dot.com bubble burst, but you might be forgiven for thinking so. The facts are that sales missed what Altera forecasted and what Wall Street expected by around $37 million. Wall Street reacted by dropping the stock 6 percent on Friday, April 20 and a further 2 percent the following Monday, wiping close to $1 billion from the value of the company.
John Danne, Altera CEO, explained the miss by two factors. During March, orders from the communications customers dried up, and at the same time, Altera booked a lot of orders that it was unable to ship that month.
The communications companies told Altera that they didn't receive the equipment orders they expected from network operators. March is the last month of the financial year for many companies, and one way for them to "enhance" numbers in the final quarter is to commit to fewer orders, which may explain the phenomena. It's not unusual to find odd behavior near year-end, and I've seen companies desperate to spend money in March so that the budget isn't lost.
Altera's second problem was what is known as "product mix." FPGA companies have thousands of different line items if you count permutations of chip types and package options combined with speed and temperature grades. The companies typically hold most of their inventory in wafer (or slice) banks. This means that the wafer has been fully or partially fabricated, but can be diverted into different packages, or screened for speed in a couple of weeks or so. The trick is to balance the inventory against the expected orders. But when customers require parts that were not forecast, then the FPGA companies need longer to fulfill the orders. This is what happened to Altera.
So, back to my original question -- Is the sky falling? Not according to the folks at Altera. They forecast that sales will recover back to revenues similar to their December quarter. That would still put them around $100 million below a year ago, but at least it's going in the right direction. John Danne was also bullish about design wins the company is achieving, which typically turn into significant sales in two or three years' time.
What of Xilinx?
The folks at Xilinx must be living on a different planet -- well, at least they were last quarter. Sales increased $48 million, which is 9 percent higher than the number reported the previous quarter. How come? After all, they serve the same markets as Altera and many of the big companies use parts from both suppliers. This was partly explained by $18 million from end-of-life sales, but clearly Xilinx didn't see the drop-off in orders from the communications guys.
Just like Altera, Xilinx is very bullish about 28-nm design wins. Both companies claim to have the best combination of performance, capability, and power -- they can't both be right. Moshe Gavrielov, the Xilinx boss, claims that customers are flocking to FPGAs because many system designers can't justify building ASIC and ASSP products on the latest technology nodes. This is most obvious, he claimed, for wired communications where the number of suppliers is shrinking fast.
Is this your experience? Do you use FPGAs because there's no ASSP available? Are you using more FPGAs today than in previous designs? I'd love to hear from FPGA customers (as opposed to hearing from the vendors).