TSMC Reportedly Rejects $1B Offers From Qualcomm, Apple
Last month, TSMC's CEO Morris Chang made waves when he suggested that it could make sense for the company to dedicate fabs to particular customers. Fresh reports today, however, suggest that both Apple and TSMC made major bids for exclusivity on future TSMC production facilities, and both were rejected. Does this mean Chang's earlier comments were misinterpreted?
No -- but it does demonstrate the difference between what Chang/TSMC is willing to contemplate and what companies like Qualcomm and Apple want. TSMC may be willing to commit full fabs to customer-specific production, but the company isn't going to sign a piece of paper that gives a customer direct control over what happens to those fabs in perpetuity. Chief Financial Officer Lora Ho told Bloomberg: "Dedicating one facility to a single product or customer creates the risk of a fabrication plant becoming a burden if the product, client or technology changes. You have to be careful. Once that product migrates, what are going to do with that dedicated fab?" said Ho. "We would like to keep the flexibility."
TSMC's overall manufacturing capacity continues to grow briskly year-to-year. Forecast vs. actual capacities above
Frankly, we're not surprised TSMC turned down the offers. $1B might sound like a lot of money, but it's significantly less than the cost of building a new fab. Two billion -- and a lot of paperwork -- is more likely a minimum figure to turn TSMC's head. Ownership of the fab, its technology roadmap, and who controls its use are going to be absolutely critical factors to any agreement, the last thing TSMC is going to do is commit to building a facility only to have it sit idle or turn sour due to poor decision from its partner.
We still think Chang was serious when he described potentially partnering with a company like Apple, because it still may make sense in the long term. The question of how to build an agreement that's mutually beneficial is a thorny one, but not impossible. Given the difficulty of continuing to ramp lower process nodes, partnering up in an attempt to mimic Intel's model still may be the best way for companies to secure exclusive rights and cutting-edge products.
As we've previously noted, any deal between TSMC and Qualcomm/Apple would be a major shake-up for the foundry industry. Samsung recently announced a $4B investment in its Austin facility, where it currently builds Apple products. Apple is easily Samsung's biggest customer, accounting for the overwhelming majority of its foundry business. Any shift to TSMC would leave the South Korean manufacturer scrambling for other customers. The love/hate relationship between Samsung and Apple defies explanation, but Samsung isn't the only company that would be impacted by such a move. Any shift in the status quo could have repercussions for Qualcomm, GlobalFoundries, Nvidia, AMD, and even Intel. The latter has no interest in entering the pure-play foundry business, but smartphone manufacturers might shift towards x86 processors if their ability to buy cutting-edge ARM products is constrained.
Despite the Bloomberg report, we don't believe we've heard the last of this issue. Stay tuned.
No -- but it does demonstrate the difference between what Chang/TSMC is willing to contemplate and what companies like Qualcomm and Apple want. TSMC may be willing to commit full fabs to customer-specific production, but the company isn't going to sign a piece of paper that gives a customer direct control over what happens to those fabs in perpetuity. Chief Financial Officer Lora Ho told Bloomberg: "Dedicating one facility to a single product or customer creates the risk of a fabrication plant becoming a burden if the product, client or technology changes. You have to be careful. Once that product migrates, what are going to do with that dedicated fab?" said Ho. "We would like to keep the flexibility."
TSMC's overall manufacturing capacity continues to grow briskly year-to-year. Forecast vs. actual capacities above
Frankly, we're not surprised TSMC turned down the offers. $1B might sound like a lot of money, but it's significantly less than the cost of building a new fab. Two billion -- and a lot of paperwork -- is more likely a minimum figure to turn TSMC's head. Ownership of the fab, its technology roadmap, and who controls its use are going to be absolutely critical factors to any agreement, the last thing TSMC is going to do is commit to building a facility only to have it sit idle or turn sour due to poor decision from its partner.
We still think Chang was serious when he described potentially partnering with a company like Apple, because it still may make sense in the long term. The question of how to build an agreement that's mutually beneficial is a thorny one, but not impossible. Given the difficulty of continuing to ramp lower process nodes, partnering up in an attempt to mimic Intel's model still may be the best way for companies to secure exclusive rights and cutting-edge products.
As we've previously noted, any deal between TSMC and Qualcomm/Apple would be a major shake-up for the foundry industry. Samsung recently announced a $4B investment in its Austin facility, where it currently builds Apple products. Apple is easily Samsung's biggest customer, accounting for the overwhelming majority of its foundry business. Any shift to TSMC would leave the South Korean manufacturer scrambling for other customers. The love/hate relationship between Samsung and Apple defies explanation, but Samsung isn't the only company that would be impacted by such a move. Any shift in the status quo could have repercussions for Qualcomm, GlobalFoundries, Nvidia, AMD, and even Intel. The latter has no interest in entering the pure-play foundry business, but smartphone manufacturers might shift towards x86 processors if their ability to buy cutting-edge ARM products is constrained.
Despite the Bloomberg report, we don't believe we've heard the last of this issue. Stay tuned.
