Faraday Announces Plans to Develop Arm-Based 64-Core SoC on Intel 18A Technology

[Daniel Nenni]

Faraday Technology Corporation (TWSE: 3035), a leader in ASIC design services and IP solutions, announces its collaboration with Arm and Intel in spearheading the development of a 64-core System-on-Chip (SoC) utilizing Intel 18A technology. This innovative SoC seamlessly integrates Arm® Neoverse™ Compute Subsystems (CSS), delivering unparalleled performance and power efficiency tailored for the construction of scalable hyperscale datacenters, infrastructure edge, and advanced 5G networks.

The Arm-based SoC is designed to be a fundamental component of Faraday's upcoming SoC evaluation platform, which aims to empower customers in accelerating the development of data center servers, high-performance computing (HPC)-related ASICs and custom SoCs. The platform will further incorporate interface IPs from the Arm Total Design ecosystem, ensuring a comprehensive implementation and verification process for 18A technology. This integrated approach is poised to streamline SoC architecture exploration within the front-end design flow, ultimately reducing the overall chip development timeline. The solution is expected to be available in 1H2025.

"As a design service partner in Arm Total Design, Faraday strategically targets the most advanced technology nodes to fulfill the evolving needs of future applications," said Steve Wang, CEO of Faraday. "We are excited to announce the development of our new Arm Neoverse-based SoC platform, leveraging Intel 18A technology. This solution will benefit our ASIC and DIS (Design Implementation Service) customers, enabling them to expedite the time-to-market for cutting-edge data center and HPC applications."

"The advancements in AI and the proliferation of data throughout the entire infrastructure emphasizes the importance of Neoverse CSS and the need for an ecosystem like Arm Total Design to accelerate innovation," said Mohamed Awad, senior vice president and general manager, Infrastructure Line of Business, Arm. "We're thrilled to see industry leaders like Faraday and Intel on the cutting edge of Arm-based custom silicon development with their Neoverse CSS-based SoC and we look forward to how this innovation advances development in data center and HPC applications."

"We are pleased to work with Faraday in the development of the SoC based on Arm Neoverse CSS utilizing our most competitive Intel 18A process technology," said Stuart Pann, Intel senior vice president and general manager of Intel Foundry Services (IFS). "Our strategic collaboration with Faraday demonstrates our commitment to delivering technology and manufacturing innovation across the global semiconductor supply chain helping Faraday's customers to seamlessly meet the world-leading power and performance standards for SoC designs."

About Faraday Technology Corporation
Faraday Technology Corporation (TWSE: 3035) is dedicated to the mission of benefiting humanity and upholding sustainable values in every IC it handles. The company offers a comprehensive range of ASIC solutions, including total 3DIC packaging, Neoverse CSS design, FPGA-Go-ASIC, and design implementation services. Furthermore, its extensive silicon IP portfolio encompasses a wide array of offerings, such as I/O, Cell Library, Memory Compiler, ARM-compliant CPUs, LPDDR4/4X, DDR4/3, MIPI D-PHY, V-by-One, USB 3.1/2.0, 10/100 Ethernet, Giga Ethernet, SATA3/2, PCIe Gen4/3, and SerDes. For further details, visit www.faraday-tech.com or follow Faraday on LinkedIn.

[Faraday News] Faraday Announces Plans to Develop Arm-based 64-core SoC on Intel 18A Technology

#intel #umc #semiconductorindustry #semiconductor #semiconductors #ASIC #arm #semiconductormanufacturing

 

[Daniel Nenni]

Faraday is UMC's design arm, no pun intended. The Intel - UMC plot thickens!

 

[nghanayem]

Faraday is UMC's design arm, no pun intended. The Intel - UMC plot thickens!

What does that mean? I thought UMC spun out their design arm ala MediaTek?

 

[Daniel Nenni]

What does that mean? I thought UMC spun out their design arm ala MediaTek?

Faraday has always been UMC's design and IP partner. Like GUC and TSMC. Faraday is no longer exclusive to UMC (since the FinFET debacle) but they are still very close, literally next door. MediaTek has their own design services group like Marvell and Broadcom.

 

[hskuo]

Faraday is UMC's design arm, no pun intended. The Intel - UMC plot thickens!

Yes, like I mentioned before, Non-tsmc or intel alliance is forming. I will expect either IBM, Rapidus or GF to be in the list next.

 

[Daniel Nenni]

Yes, like I mentioned before, Non-tsmc or intel alliance is forming. I will expect either IBM, Rapidus or GF to be in the list next.

Agreed. Faraday is not a whale in my opinion but it is notable. Very interesting times. The Intel Foundry event is in two weeks in San Jose. We should know more then.

Intel Foundry Direct Connect – April 29, 2025 – Register now

Hear from Intel leaders, customers, industry technologists, and partners as they share details of our strategy, process technology, and advanced packaging and test.

www.intel.com

 

[MKWVentures]

this is one of their committed customers. They plan to make designs that some other company may decide to use someday. its like a mythical micro-whale.

It finally is leaking out.... Intel builds fabs for other foundries. will it be as successful as the IBM Samsung GF alliances?

 

[nghanayem]

this is one of their committed customers. They plan to make designs that some other company may decide to use someday. its like a mythical micro-whale.

That makes literally no sense. They are an IP provider, their website and Dan say as much. That would be like saying Synopsis and ARM are intel’s whales because IFS customers will license their i16/i3/18A IP to said customer's products.

 

[nghanayem]

It finally is leaking out.... Intel builds fabs for other foundries. will it be as successful as the IBM Samsung GF alliances?

How do you figure that? i3 and 18A are hardly licensed process technology.

 

[MKWVentures]

it will become clear in time.

How do you figure that? i3 and 18A are hardly licensed process technology.

1) does intel consider Faraday a customer or not?
2) If I build a fab for other foundries to use to sell to their customers, I am building Fabs for foundries. when you build wafers for other companies you dont give them a license.

 

[Daniel Nenni]

this is one of their committed customers. They plan to make designs that some other company may decide to use someday. its like a mythical micro-whale.

It finally is leaking out.... Intel builds fabs for other foundries. will it be as successful as the IBM Samsung GF alliances?

I wouldn't go that far. Does TSMC build fabs for Intel? Yes and no. They build fabs based on wafer agreements but the fabs are not exclusive to Intel. TSMC builds processes for Apple that are excusive to Apple but the fabs are used by others. Splitting hairs I guess. Faraday is primarily in the ASIC business but they do license/sell IP to enable that business.

The IBM alliance was doomed by too many cooks in the kitchen in my opinion. If Intel is the only cook it could work.

 

[hskuo]

it will become clear in time.

1) does intel consider Faraday a customer or not?
2) If I build a fab for other foundries to use to sell to their customers, I am building Fabs for foundries. when you build wafers for other companies you dont give them a license.

My inputs:
1) If you check tsmc's Design Alliance eco-system, it is clear that company like Faraday is customer and also enabler of foundry business for system house without design backend capability. From intel POV, Faraday is definitely a customer and also partner.
2) "Build" a fab you mentioned is quite ambiguous. It should be "maintain a fab with decent wafer delivery." Advanced Fab is capital intensive investment which might cost more than $8B USD for 20kwpm capacity. To get "decent wafer" (economically reasonable yield), you need high skilled workers and management with reputation, not mentioned about the reliable utility support from local government and more. If my understanding is correct, we might check the SMIC model in China. That is China government pays the fab-build-cost and SMIC manage it. Where they get the manufacturing technology license?

 

[nghanayem]

it will become clear in time.

1) does intel consider Faraday a customer or not?

“Our strategic collaboration with Faraday demonstrates our commitment to delivering technology and manufacturing innovation across the global semiconductor supply chain helping Faraday’s customers to seamlessly meet the world-leading power and performance standards for SoC designs.” --- Stu Pann GM of IFS

Seems clear cut that they are a design enablement partner

2) If I build a fab for other foundries to use to sell to their customers, I am building Fabs for foundries. when you build wafers for other companies you dont give them a license.

Which fab is intel building for foundries; none. The new fabs intel has announced have only ever been talked about as using intel 4 and beyond. Process technologies that IFS is peddling to fabless design houses. Those fabs are not being built on the behalf of TSMC/Samsung/UMC/GF to run their process technologies nor technologies that intel co-developed with either firm. The fabs in question (Fab 11/11X and Fabs 12-32) are ancient fabs that intel evidently didn't want to bring forward beyond intel 7. In both instances intel has said this is about extracting more value from their existing fab shells and tooling. Something TSMC and to a lesser extent Samsung already do with their existing 3rd party foundry ecosystems. In Samsung's case they also make display, CMOS image sensors, power ICs, etc for Samsung's internal needs. If intel is not going to convert the floor space to store the non EUV tools for their future nodes, and intel's BUs continue to have no need for anything that isn't N and N+1 process technology; then intel has only 3 things they can do with their legacy fabs:
1) Close said fabs once the last internal product production runs are done
2) Do what they are doing and be a foundry for other mature foundries and re-open Fab 9 as an advanced packaging facility
3) Spend the time and money to re-engineer 14nm and intel 7 as foundry technologies ala intel 16/22FFL. Once that is done, they get the privilege of spending the time and money to build a foundry ecosystem for said nodes. Then, and only then, they can attempt to get customers on these nodes.

If forced to pick from those three options I know which sounds far and away like the better idea.

 

[MKWVentures]

“Our strategic collaboration with Faraday demonstrates our commitment to delivering technology and manufacturing innovation across the global semiconductor supply chain helping Faraday’s customers to seamlessly meet the world-leading power and performance standards for SoC designs.” --- Stu Pann GM of IFS

Seems clear cut that they are a design enablement partner


Which fab is intel building for foundries; none. The new fabs intel has announced have only ever been talked about as using intel 4 and beyond. Process technologies that IFS is peddling to fabless design houses. Those fabs are not being built on the behalf of TSMC/Samsung/UMC/GF to run their process technologies nor technologies that intel co-developed with either firm. The fabs in question (Fab 11/11X and Fabs 12-32) are ancient fabs that intel evidently didn't want to bring forward beyond intel 7. In both instances intel has said this is about extracting more value from their existing fab shells and tooling. Something TSMC and to a lesser extent Samsung already do with their existing 3rd party foundry ecosystems. In Samsung's case they also make display, CMOS image sensors, power ICs, etc for Samsung's internal needs. If intel is not going to convert the floor space to store the non EUV tools for their future nodes, and intel's BUs continue to have no need for anything that isn't N and N+1 process technology; then intel has only 3 things they can do with their legacy fabs:
1) Close said fabs once the last internal product production runs are done
2) Do what they are doing and be a foundry for other mature foundries and re-open Fab 9 as an advanced packaging facility
3) Spend the time and money to re-engineer 14nm and intel 7 as foundry technologies ala intel 16/22FFL. Once that is done, they get the privilege of spending the time and money to build a foundry ecosystem for said nodes. Then, and only then, they can attempt to get customers on these nodes.

If forced to pick from those three options I know which sounds far and away like the better idea.

OK. Lets see what happens in the next 1-2 years. maybe it will be more clear on Intel's plans for the foundry ecosystem then.