Will Intel take significant Qualcomm business from TSM

[Arthur Hanson]

Looking at Qualcomm's chairman's interview on CNBC it looks like Intel is looking to take additional chip business from TSM. Any thoughts on this would be appreciated. Gelsinger is making claims to equal and even pass TSM in technology. Intel has increased cap-ex, but is still far below TSM. Is Gelsinger blowing a lot of smoke or does he have reality to back it up. Any thoughts or comments on this subject would be appreciated.

 

[count]

It's in Qualcomm's best interest to talk up Intel Foundry, they don't like being dependent on TSMC and want more foundry competition. I think Gelsinger is mostly blowing smoke.

 

[Daniel Nenni]

It's in Qualcomm's best interest to talk up Intel Foundry, they don't like being dependent on TSMC and want more foundry competition. I think Gelsinger is mostly blowing smoke.

Agreed. QCOM leverages foundries against each other, that is in their DNA.

From what I hear inside the ecosystem I don't think QCOM will use Intel Foundry until 2A so it will be a while before we see it. And I'm not sure if that will be a high volume SoC/Modem or other lower volume supporting chips. Remember, QCOM has stiff competition with the likes of Mediatek which is closely partnered with TSMC. If Intel Foundry processes put QCOM at a disadvantage in delivery or performance they will go with TSMC, absolutely.

 

[VCT]

Agreed. QCOM leverages foundries against each other, that is in their DNA.

From what I hear inside the ecosystem I don't think QCOM will use Intel Foundry until 2A so it will be a while before we see it. And I'm not sure if that will be a high volume SoC/Modem or other lower volume supporting chips. Remember, QCOM has stiff competition with the likes of Mediatek which is closely partnered with TSMC. If Intel Foundry processes put QCOM at a disadvantage in delivery or performance they will go with TSMC, absolutely.

In the past, it's very hard for Mediatek to compete with QCOM in mid end market. (not even high end).
Since QCOM failed it's Samsung made snapdragon 888 (China call it fire dragon because of overheat), it hurts QCOM brand image and help Mediatek improved it's brand in mid end market.
QCOM lost a lot of market share in mid end market because of that. It will cost QCOM a lot more than they know in the long term.

 

[Daniel Nenni]

In the past, it's very hard for Mediatek to compete with QCOM in mid end market. (not even high end).
Since QCOM failed it's Samsung made snapdragon 888 (China call it fire dragon because of overheat), it hurts QCOM brand image and help Mediatek improved it's brand in mid end market. QCOM lost a lot of market share in mid end market because of that. It will cost QCOM a lot more than they know in the long term.

This is true! Fire Dragon! QCOM also fumbled with Samsung 10nm with single digit yields for the launch.

Mediatek is an amazing company. They are very nimble and work very hard. I have nothing but respect for them. In the past Mediatek used mature nodes to get cost advantages but now they are on the leading TSMC processes. I also noticed that Mediatek has many former TSMC employees working there which is an advantage.

No matter what the politics are China prefers Taiwan business over US business but first and foremost they want the best chips at the best prices.

 

[hist78]

Can Qualcomm maintain three separate divisions, one each for TSMC, Samsung, and Intel, for leading edge semiconductor development and product marketing?

I assume each of them has different IP, manufacturing technologies, design rules and tools, and ecosystems. It's complicated. And don't forget about the NDA constraint between them and the foundries.

Qualcomm is probably among the few companies who have the capabilities and willingness to try such arrangements.

 

[Daniel Nenni]

Can Qualcomm maintain three separate divisions, one each for TSMC, Samsung, and Intel, for leading edge semiconductor development and product marketing?
I assume each of them has different IP, manufacturing technologies, design rules and tools, and ecosystems. It's complicated. And don't forget about the NDA constraint between them and the foundries. Qualcomm is probably among the few companies who have the capabilities and willingness to try such arrangements.

Qualcomm has done this since the beginning, 2nd, 3rd, even 4th sourcing of a single product. They had their own set of design rules that the smaller fabs accommodated. This was before FinFETs. With FinFETs products cannot be 2nd and 3rd sourced without modification. There are also NDA issues requiring separate teams for TSMC and Samsung designs. Much more complicated than in the CMOS days.

Nvidia also jumps between Samsung and TSMC as do others. Mostly for price and to keep TSMC in check. Nobody wants a foundry monopoly so Intel Foundry has a chance, absolutely.

 

[hist78]

Qualcomm has done this since the beginning, 2nd, 3rd, even 4th sourcing of a single product. They had their own set of design rules that the smaller fabs accommodated. This was before FinFETs. With FinFETs products cannot be 2nd and 3rd sourced without modification. There are also NDA issues requiring separate teams for TSMC and Samsung designs. Much more complicated than in the CMOS days.

Nvidia also jumps between Samsung and TSMC as do others. Mostly for price and to keep TSMC in check. Nobody wants a foundry monopoly so Intel Foundry has a chance, absolutely.

Yes, I totally agree that multiple foundries can bring more competition and technology advancement to the industry. What I'm not sure is how many fabless companies who can afford and is willing to do so on the leading edge products.

No mater how big a company is, like Qualcomm, there is a finite number of talents the company can hire to develop the products. It can be very complicated and consume a lot resource when those talents are further divided into multiple groups according to foundries partners, manufacturing technology, and different editions of the same product line.

 

[Daniel Nenni]

Yes, I totally agree that multiple foundries can bring more competition and technology advancement to the industry. What I'm not sure is how many fabless companies who can afford and is willing to do so on the leading edge products. No mater how big a company is, like Qualcomm, there is a finite number of talents the company can hire to develop the products. It can be vary complicated and consume a lot resource when those talents are further divided into multiple groups according to foundries partners, manufacturing technology, and different editions of the same product line.

The available semiconductor talent pool is definitely getting shallower now that system companies are doing their own chips and they spend a lot more money on tools and people than the fabless chip companies and the IDMs. I have chip friends who now work at AirBnB and most of the automotive companies now. Microsoft has been bulking up as well. It really is a great time to be a semiconductor professional, absolutely. Also a great time to have a semiconductor design portal. Business is seriously booming.

 

[hist78]

The available semiconductor talent pool is definitely getting shallower now that system companies are doing their own chips and they spend a lot more money on tools and people than the fabless chip companies and the IDMs. I have chip friends who now work at AirBnB and most of the automotive companies now. Microsoft has been bulking up as well. It really is a great time to be a semiconductor professional, absolutely. Also a great time to have a semiconductor design portal. Business is seriously booming.

What's AirBnB plan to do? A super computer cluster powered by millions of AirBnB "hosts"?

 

[JeffreyHF]

In the past, it's very hard for Mediatek to compete with QCOM in mid end market. (not even high end).
Since QCOM failed it's Samsung made snapdragon 888 (China call it fire dragon because of overheat), it hurts QCOM brand image and help Mediatek improved it's brand in mid end market.
QCOM lost a lot of market share in mid end market because of that. It will cost QCOM a lot more than they know in the long term.

Really? Did you digest Qualcomm's 11/3 earnings and conference call? What about their bullish forecasts for the next quarter and next fiscal year? Perhaps it would be important to attend their 11/16 Investor's Day, and then their Snapdragon unveiling in Hawaii at around the beginning of December? Their QCT phone business is growing robustly, not shrinking, and their RFFE and WiFi 6 attach rates are hurting Quorvo, Skyworks, and Broadcom. The new Qualcomm is deriving 40% of its money from IoT, automotive, robotics, drones, and other adjacencies. There is always a danger of viewing Qualcomm through the lens of negative confirmation bias, and not seeing them for the company they've become.

 

[count]

Qualcomm stumbled a few years ago with overheating snap dragons, but they made a brilliant acquisition with Nuvia, gaining a lot of talent that had left Apple in the process, and they are now back on track. Qualcomm strength in 5G is not to be discounted, and I expect that they will be putting out some very very good cpus in the future.

 

[hist78]

Really? Did you digest Qualcomm's 11/3 earnings and conference call? What about their bullish forecasts for the next quarter and next fiscal year? Perhaps it would be important to attend their 11/16 Investor's Day, and then their Snapdragon unveiling in Hawaii at around the beginning of December? Their QCT phone business is growing robustly, not shrinking, and their RFFE and WiFi 6 attach rates are hurting Quorvo, Skyworks, and Broadcom. The new Qualcomm is deriving 40% of its money from IoT, automotive, robotics, drones, and other adjacencies. There is always a danger of viewing Qualcomm through the lens of negative confirmation bias, and not seeing them for the company they've become.

Qualcomm didn't explain why they can increase smartphone chip sales by 56% YoY. It's probably not caused by some new product released by Qualcomm. Or we can learn a little bit from this interview on CNBC:

Qualcomm shares pop after company reports earnings beat, 56% rise in smartphone chip sales

Qualcomm's revenue beat Wall Street expectations, driven by chip sales for smartphones, which were up 56% despite a global chip shortage.

www.cnbc.com

I remember some news reports in early 2021 stating that TSMC agreed to take some emergency orders from Qualcomm even TSMC itself was under a very tight production constraint due to the unusual market demand. Those reports also said Qualcomm's emergency orders were originally planned to be fullfiled by Samsung Foundry.

If those reports are true, it leads us to an interesting question: how can a fabless company do if their foundry partner can't deliver the products with right quality and quantity at the right time?

If TSMC really ran out the means and ways to accommodate Qualcomm's last minute order, what can Qualcomm do?

On the other hand I suspect the reason, partially, to allow Mediatek to lead Qualcomm in volume for several quarters is Samsung can't produce enough chips for Qualcomm in time.

Mediatek not only uses TSMC as the primary foundry partner but also counts Samsung Smartphone division as its customer.

But why some of the Samsung smartphones use Mediatek's chips instead of Qualcomm's? Is it because Samsung Foundry can't produce enough Qualcomm chips? Or Samsung smartphone division wants to do multi sourcing to avoid letting Qualcomm to become the single external supplier for their smartphone application processors?

To think further, it means TSMC is an important supplier of Samsung Smartphone division.

It's getting very interesting.