The system-on-chip market was a cattle-call barely a decade ago but has thinned to just three major providers now, and that's best for everyone -- but especially Teradyne, says Barclays analyst CJ Muse. He also addresses whether there's any hidden upside to the Samsung-AMAT settlement.
Handicapping the SoC market: Three's not a crowd
Now that the market for system-on-chip (SOC) test is down to three main suppliers (Teradyne, proposed LTX-Verigy merger, and Advantest), Barclays' CJ Muse handicaps the field. Advantest's key customers are Intel and AMD (plus some Toshiba business), but though its T2000 has expanded it's still lagging TER and VRGY in performance and roadmap, he says. A merged LTXC/VRGY means no more single-platform strategy: look for the Diamond to overshadow the V101, and the 93K to win out over the Sapphire, he says. Teradyne, meanwhile, has emerged as the company with the most complete SoC test portfolio; and as with any merger, some affected LTXC/VRGY customers (especially those using to-be-discontinued products) could be up for grabs.
TER has pulled away from the pack over the past five years to grab ~50% share in 2010, thanks in part to mix (less CPU/GPU spending hurts VRGY and Advantest) but also by gobbling up other's business. TER is holding its own at digital powerhouses Qualcomm and Broadcom, is making inroads into LTX "strongholds" TXN and ST, and Muse says its Eagle Test division has "incremental gains" for core analog at Monolithic Power, On Semi, and Linear. TER also has a strategic focus on next-gen products centering on mobility (getting more out of chip design or test functionality) and power efficiency (exploring areas such as LED or different storage techniques) that should maintain its ~50% share, he believes.
A market that just a decade ago had a dozen jostling competitors now has three players with ≥95% market share (two with ≥80%), and that makes the market more rational, Muse writes. Socket wins and competition will still persist, but more rational behavior from management should keep up the ASP pressure and better support margins for everyone.
UPDATE 12/6: Advantest has submitted an unsolicited $12.15/share (~$700M) bid for VRGY, which said it would initially reject in favor of going forward with LTXC -- however, it coyly noted that "the Advantest proposal might lead to a superior transaction" (read: pay us more) and so will "engage in discussions." Muse's take: Advantest lags "severely" behind all three SoC test players (TER, VRGY, LTXC) and probably sees this is as its "last shot at obtaining core digital and/or analog technology" from either VRGY or LTXC. He sides with the VRGY-LTXC as the better strategic combination ("LTXC's low-end digital and analog is precisely what VRGY is missing"), but adding Advantest to the mix clearly puts that deal at risk. (Investors seem happy just to have a second suitor; VRGY stock is currently up 40% after ATE's offer, wiping out what was a 30% premium to VRGY's Friday stock price; LTXC is down 12%.)
AMAT-Samsung: Back to just friends?
Speculation that AMAT's settlement with Samsung over an IP-leak dispute featuring reduced tool pricing could float its market share at its key customer might not be quite accurate, says CJ Muse from Barclays. While the supplier marches on with new chip tool introductions, its "gradual progress in organic growth" puts it at roughly 19% market share through 2009 for wafer fab equipment -- that's below the 22% it held in 2004.
As for AMAT's newest etch products, Muse notes they "are intended to be useful in SADP/ LELE etch (for now) and in TSV silicon etch (in 2012)," and the Centris "may have helped performance at Toshiba, Samsung, and Micron" (AMAT said five customers have or will soon take shipment). Still, he sees these new tools as "more evolutionary than revolutionary, and aimed at recapturing share lost in silicon etch when deep trench architecture converted to stack in DRAM." And taking share won't be easy in an etch market that includes not just familiar foes Lam Research and TEL but also Hitachi, China's AMEC, and Mattson (for noncritical etch), he points out.