Is Altera the first of some 'big name' foundry deals for Intel?

It's already making devices for two programmable logic start ups and a network processor company. But while Achronix, Tabula and Netronome are accessing Intel's 22nm process, Altera will be on the 14nm node and taking advantage of the microprocessor giant's finFET technology. The fact that Altera is pushing ahead with process technology is, in itself, no surprise: both Altera and arch rival Xilinx have chased Moore's Law aggressively. What is surprising is that Altera has gone with Intel, having previously been linked inextricably with TSMC. A hint that Altera was looking beyond TSMC came in October 2012, when Brad Howe, senior vice president for R&D, told New Electronics that it may use other foundries in the future. "Although we don't have plans to switch," he said, "we continue to monitor the industry and there's a lot of movement in the foundry sector at the moment." Altera's president and ceo John Daane said: "Altera's fpgas using Intel 14nm technology will enable customers to design with the most advanced, highest performing fpgas in the industry. In addition, Altera gains a tremendous competitive advantage at the high end in that we are the only major fpga company with access to this technology." Achronix has a different view; it says it has a deal in place for Intel to manufacture fpgas on its 14nm process and claims work has already begun on 14nm products. Altera believes the performance and power profile of Intel's 14nm process is a good fit for some – but not all – planned products. Manufacturing choices, it says, will made to create product differentiation and to increase its competitive strength, as well as to meet performance, schedule and supply needs. "TSMC remains an important part of our future product development," Daane noted. "We look forward to continuing our close partnership to jointly develop technologies for next generation products." The move has attracted the attention of industry commentators. John Cooley from deepchip.com said: "Big Silicon let out a collective panicky 'gasp' when Intel said it's making 14nm FinFET chips for Altera. This means TSMC, GlobalFoundries, Samsung and UMC will now be facing Intel offering working 14nm silicon that's 24 to 48 months ahead of them – and it's cherry picking the lucrative fab deals. Ouch!" Respected analyst Gary Smith added: "This is very bad news for the big commercial fabs. Fpgas are the high volume, high margin cash cow designs that keep the line full and steady money coming in. This new Altera 14nm deal is the Intel foundry scooping up the gravy dollars away from TSMC." Meanwhile, Cooley cites sources that claim Cisco has signed a deal for Intel to make its networking chips and that Apple is in talks about foundry capacity. While TSMC will still have lead customers for its yet to be announced 14nm process, the loss of Altera will be significant. Apart from the volume – and hence revenue – implications, fpga manufacture is an important part of the ramping process for new processes. TSMC has experienced well documented problems in manufacturing 28nm chips – even with the help of Altera and Xilinx. The loss of Altera – at least for the 14nm node – won't help the foundry giant. But how might this affect the 'balance of power' between Altera and Xilinx? Both make strenuous efforts to get leading edge chips out ahead of each other. The reason? Their major customers are comms companies desperate to keep ahead of the rapidly moving market. If Altera can get 14nm chips into customers' hands before Xilinx, it could get a significant advantage. Cooley believes Altera may get 14nm samples by the end of 2013. "With 14nm silicon," he concluded, "Altera will get to leapfrog Xilinx, which will be stuck in TSMC at 20nm."