Specialist
Former senior executive at Qualcomm Inc
Agenda
- Impact of Intel (NASDAQ: INTC), Samsung (KRX: 006400), GlobalFoundries (NASDAQ: GFS) and TSMC’s (TWSE: 2330) capacity expansion on chip pricing, wafer allocation and risk of oversupply
- To what extent is friend-shoring taking place when conducting supply chain management?
- Emergence of a competitive semiconductor supply chain limited to the US, Europe, Korea, Japan and Southeast Asia
- Response of fabless manufacturers and foundry customers – such as Qualcomm (NASDAQ: QCOM) and Apple (NASDAQ: AAPL) – to evolving geopolitical relations between the west and Greater China
Questions
1.
Could you give an overview of the foundry procurement market environment, identifying 2-3 trends or drivers we should monitor?
2.
What are the specific impacts of the supply-demand imbalance you described on a company such as Qualcomm looking for foundry capacity? Is it just that the company has all these wafer starts lined up and it doesn’t know where the customers are going? Will that change its procurement behaviour at all?
3.
How much would the chip prices typically come down by for a company such as Qualcomm when there is a demand shortage or oversupply?
4.
In terms of the oversupply or demand shortage, I hear different things in different sectors – specialists in the automotive sector speak to a supply shortage. How long do you expect to be in this oversupply situation? Why do you feel there is oversupply when others feel there’s undersupply?
5.
What are the key purchasing criteria when a fabless manufacturer or foundry customer such as Qualcomm or Apple is procuring foundry capacity?
6.
How important are more obscure factors, such as a foundry’s geographical location, the lead time or the particular capability at a certain process node?
7.
You said customers don’t really show a preference for where Qualcomm or another player might source their foundry capacity from. If we’re thinking about a European 5G hardware company such as Ericsson, have you seen any customers say that they don’t want you using foundry capacity in X location because of Y government regulation, or similar? There has been some trouble with Huawei in Europe.
8.
In terms of single- and dual-sourcing foundry capacity, is there an ideal number of foundries to work with? Is it a more-diversified-the-better-type attitude, or is it easier and cheaper to work with one supplier?
9.
What might make a company such as Qualcomm or Apple look to dual source again? Would it be losing access to some capacity because of some exogenous shock, such as a natural disaster or serious deterioration in geopolitical relations, or is there a commercial reason?
10.
How might a Qualcomm or Apple respond to losing access to capacity in Taiwan, for example, and over what timescale could they make up for any losses?
11.
If Qualcomm were to take one year to react to an extreme event, meaning it lost its capacity somewhere, what revenue hit might that cause the company?
12.
Could a fabless company such as Qualcomm or Broadcom take any measures now to diversify where they get their foundry capacity from and lessen the impact of any extreme event?
13.
Are there any risks associated with long-term contracts? A long-term manufacturing partnership, worth USD 4bn and running until 2028, was announced between Qualcomm and GlobalFoundries in August 2022. What would happen if GlobalFoundries couldn’t fulfil its commitments?
14.
What would happen if a foundry couldn’t fulfil its commitment in a long-term contract? Given the IP is in multiple places and the deal is off, would a player such as Qualcomm just go somewhere else?
15.
Are there any benefits of signing a long-term contract? I appreciate you said that USD 4bn of wafers over six years isn’t that much for Qualcomm, but why sign a contract until 2028 if it wasn’t favourable for the company?
16.
When partnerships arise between fabless manufacturers and foundries, how does paying for it work? Would there be a big down payment at the beginning and then the fab gets built in two years’ time, or would a GlobalFoundries say, “We’re building this fab here, do you want some wafers in two years?” When is the agreement for the wafer allocation made and money exchanged for that to happen?
17.
How much of an impact do incentives have on where Qualcomm or other fabless players or foundry customers look to source their foundry capacity from? The company said it’s open to working with foundries in Europe if the incentive programmes were attractive. What makes an incentive programme attractive to a Qualcomm?
18.
There have been some big regulatory announcements recently, such as the US CHIPS [Creating Helpful Incentives to Produce Semiconductors] and Science Act and the European Chips Act. The Qualcomm-GlobalFoundries deal was announced a couple of days after the US CHIPS and Science bill passed. How do regulations such as these change the thinking behind procurement for a company such as Qualcomm or Apple, or do they not change at all, and it’s more for the fabs themselves?
19.
What would need to be in place besides incentives or regulation to make locations such as the US or Europe more attractive for foundry semiconductor manufacturing to a company such as Qualcomm?
20.
We’re seeing hundreds of billions of dollars of foundry expansion across Europe and the US. Do you feel the foundry capacity coming online towards the middle of the decade will be less attractive than foundry capacity in Asia in perpetuity?
21.
On the lagging-edge side where dual-sourcing is more common, how would you pick where to put the biggest wafer allocation at a fab that has the same capabilities for technology, pricing and so on?
22.
When making decisions on wafer allocations in the three or four fabs we discussed, how do you forecast so you can be aligned with demand as closely as possible and not end up in an oversupply situation?
23.
In terms of the six months’ forecasting, you said there was an oversupply situation for some leading-edge chips. How well can you execute forecasting to mitigate or avoid oversupply?
24.
You said you trimmed the excess and obsolete by USD 100m during your time at Qualcomm. Is it prudent to always run an excess and obsolete, and how big would that be in revenue terms?
25.
Are there ever cases of a company losing a customer because of poor demand planning, or being unable to have access to the capacity they need or get a chip to them in a timely manner?
26.
If GlobalFoundries couldn’t provide Qualcomm with the chips for whatever reason, how would Qualcomm respond to that in a situation where capacity was tight? Would it be a matter of losing that business and the customer using another competitor, or can the company take measures to mitigate this situation?
27.
Smartphone chips are higher margin than most auto chips, but many auto OEMs would obviously want those chips as quickly as possible. Are there ever situations where Qualcomm or another fabless player might sign a contract to confirm it won’t push an auto OEM down the list for a higher-margin chip in a situation of tighter capacity?
28.
You said the deal between Qualcomm and GlobalFoundries until 2028 is quite long. How long would a normal contract be between Qualcomm and a foundry partner?
29.
Do you feel it’s currently better to take advantage of capacity, or do you predict an excess situation, so it’s not so important? What would be the thinking now for a company such as Qualcomm?
30.
Do you see any risks to a fabless player such as Qualcomm or Marvell procuring foundry services, or are we heading for a situation of oversupply and overcapacity, which is good for fabless players?
31.
If you were looking at Apple as a customer, do you see any risks there other than geopolitical?
32.
Would you highlight anything else about foundry procurement for fabless manufacturers?
