6 Questions & Answers About Samsung-Tesla Foundry Deal

Tesla recently announced a $16.5 billion foundry deal with Samsung. What does this deal mean for both parties?

By Mark LaPedus

Late last month, Tesla announced a new and major $16.5 billion foundry deal with Samsung. Under the terms, Samsung will make select chips for Tesla’s electric vehicles (EVs).

The deal generated a buzz in the market. There were (and still are) multiple facets to the announcement. So, it’s time to examine the deal in depth, and look at Samsung and its struggles in the foundry business, as well as Tesla’s chip roadmap and related topics.

Both Samsung and Tesla are iconic brands. South Korea’s Samsung Electronics, a large multinational conglomerate, generated $216.5 billion in sales in 2024. In the semiconductor market, Samsung is best known as major supplier of memory chips. Samsung also has a foundry unit, which manufactures chips for other companies in large facilities called fabs. Samsung Foundry competes in the foundry business against TSMC, Intel, SMIC, UMC, GlobalFoundries and others.

The memory business is a cash cow for Samsung, but it’s a different story for the foundry unit. For some time, Samsung Foundry has been beset by losses, lackluster chip yields and dwindling market share.

Suddenly, Tesla last month announced the multi-billion-dollar foundry deal with Samsung, a move that could help revive its underperforming, loss-ridden foundry unit. The deal is also critical for Tesla’s AI chip efforts.

Tesla and Samsung are no strangers to each other. Since 2019, Samsung has been manufacturing Tesla’s so-called Full Self Driving (FSD) processor chips, which are used to power the self-driving functions in Tesla’s EVs. The latest FSD chip, called A14, is designed by Tesla and incorporated inside its latest EV models.

Electric vehicles (EVs) are vehicles that use an electric motor to move the vehicle. An on-board battery pack is used to power the electric motor. Source: U.S. Energy Information Administration (EIA)

Meanwhile, in 2026 or so, Tesla is expected to release a new and faster FSD chip for its upcoming EVs. That Tesla-designed chip, dubbed A15, is being manufactured by TSMC.

Then, for the next FSD chip, Tesla has once again selected Samsung as its foundry partner. Under the terms of the $16.5 billion foundry deal, Samsung will manufacture Tesla’s next-generation FSD chip, dubbed A16. Based on Samsung’s new 2nm process, the Tesla-designed A16 system-on-a-chip (SoC) design is expected to power the car maker’s future EVs, robotaxis and robots.

Samsung’s foundry unit will manufacture the A16 chip for Tesla in its new but delayed fab in Taylor, Texas. Samsung’s $17 billion fab in Taylor was supposed to be in production in 2024. But the company has been unable to secure any foundry customers for the fab, causing it to delay the production ramp until 2026.

Tesla will help solve that problem. The A16 chip, however, is not expected to appear until 2028 or so. Until then, Samsung will need to find other foundry customers to help fill the fab. Samsung is also ramping up its 2nm process within its fabs in Korea. It needs to secure customers for those fabs as well. That won’t be an easy task for Samsung.

Nonetheless, there are several pressing questions about the Tesla-Samsung foundry deal, including:

*Why is the A16 chip so important for Tesla and Samsung?

*Will the Tesla deal help revive Samsung’s loss-ridden foundry unit? (Or will Apple save the day?)

*What are the future business opportunities for Samsung at Tesla?

*Will the Tesla deal help improve Samsung’s lackluster chip yields?

*Will this deal help generate new business for Samsung Foundry?

*Will it help Samsung’s foundry unit close the gap with TSMC?

In this analysis piece, I have answered these and other questions, based on my observations and opinions (See below).

Samsung’s rise and troubles

Before answering these questions, let’s take step back and look at Samsung. Founded in 1938, Samsung started off as a trading company. Over time, the company diversified into various sectors. And eventually, it became the largest chaebol, or business conglomerate, in South Korea.

In 1969, the company entered the electronics business, making black-and-white TVs and cheap appliances. More recently, Samsung has become a major supplier of appliances, smartphones and TVs.

Meanwhile, in 1974, Samsung entered the semiconductor business by acquiring Korea Semiconductor. Initially, Samsung’s fledgling semiconductor unit sold simple chips for watches. In the 1980s, it entered the memory business. And by 1992, Samsung became the world’s largest supplier of memory chips.

In another major move, the company entered the foundry business in 2005. The foundry unit was riding high in 2007, when it garnered a major customer—Apple. Samsung supplied the processor for Apple’s first iPhone in 2007.

Then, starting in 2010, Apple began to design its own SoCs for the iPhone. Apple’s SoCs incorporate processor cores and other functions on the same device. Samsung manufactured Apple’s SoCs for the early iPhones.

Samsung then suffered a major setback in the mid-2010s, when Apple switched foundry camps. Starting with the iPhone7 in 2016, TSMC became the exclusive foundry vendor for Apple’s iPhone SoCs. As a result, TSMC’s foundry business took off.

Apple switched from Samsung to TSMC—and for good reason. At the time, Samsung was (and still is) competing against Apple in the cellphone business. Nonetheless, Samsung, which lost Apple’s foundry business, never fully recovered in the foundry market.

Fast forward. In 2024, Samsung reclaimed the No.1 spot as the world’s largest semiconductor company with $66.5 billion in sales, according to Gartner. But lately, it’s been a tough period for the Korean company. Samsung has struggled in the lucrative high bandwidth memory (HBM) market. As a result, South Korea’s SK Hynix is now the world’s largest memory supplier, surpassing Samsung, according to Counterpoint Research.

Meanwhile, Samsung Foundry currently operates several 200mm and 300mm fabs in Korea and the U.S. The company offers a broad range of trailing- and leading-edge processes, including everything from 180nm to 3nm technologies. It also provides an advanced packaging service.

Despite its impressive portfolio, Samsung’s foundry unit continues to struggle amid numerous reorganizations, losses and lackluster chip yields. It continues to lose market share.

In the first quarter of 2025, TSMC had a dominant 67.6% market share in the foundry business, according to TrendForce. Samsung was a distant second with 7.7% share, according to TrendForce. Samsung’s foundry sales reached $2.89 billion in the first quarter, down 11.3% from the previous period.

In 2024, Samsung’s foundry unit posted an operating loss of $4.1 billion, according to a report from CNN, which cited CLSA as its source. The foundry unit is expected to lose $4.8 billion in 2025, according to the report.

Tesla’s EV chip roadmap

Clearly, Samsung’s foundry unit is headed in the wrong direction. But suddenly, Samsung last month got a huge shot in the arm, thanks to the $16.5 billion foundry deal with Tesla.

In a recent posting on X, Elon Musk, chief executive of Tesla, described the terms of the deal. “Samsung’s giant new Texas fab will be dedicated to making Tesla’s next-generation A16 chip. The strategic importance of this is hard to overstate,” Musk said. “Samsung currently makes A14. TSMC will make A15, which just finished design, initially in Taiwan and then Arizona.”

Before deciphering Musk’s comments, let’s look at Tesla and its EV hardware/chip roadmap. In 2003, Tesla was founded by Martin Eberhard and Marc Tarpenning. Musk was an early investor and a board member at Tesla. Musk took over as chief executive in 2008.

The automotive industry began to take Tesla seriously in 2017, when the company rolled out the Model 3 EV. The mid-size sedan was a hit with select consumers. And the car demonstrated that EVs were finally ready for prime time.

In 2019, Tesla incorporated its new Hardware 3 (HW3) technology in its newer Model 3 EVs and other vehicles. Basically, HW3 was Tesla’s third-generation Full Self-Driving (FSD) computer. HW3 was a board-level onboard computer, which was incorporated inside a vehicle.

HW3 consisted of an FSD chip, memory and other devices. The Tesla-designed FSD chip is a complex SoC. The HW3 SoC is a 2.2GHz CPU with 12 cores and two neural network processors on the same device, according to Munro Live, an automotive website. The SoC is manufactured using Samsung’s 14nm finFET process.

In Tesla’s EVs, HW3 and subsequent FSD computers are designed to power two of the company’s key technologies--Autopilot and the Full Self-Driving (FSD) function. Autopilot and FSD are different. Autopilot is an advanced-driver assistance system, which provides automatic emergency braking, forward collision warning and related functions in Tesla’s EVs. The FSD function enables the self-driving capabilities in Tesla’s vehicles.

Generally, autonomous driving is a complex technology. “Autonomous cars rely on sensors, actuators, complex algorithms, machine learning systems, and powerful processors to execute software,” according to Synopsys in a blog. “Sophisticated software then processes all this sensory input, plots a path, and sends instructions to the car’s actuators, which control acceleration, braking, and steering. Hard-coded rules, obstacle avoidance algorithms, predictive modeling, and object recognition help the software follow traffic rules and navigate obstacles.”

In the world of autonomous driving, the Society of Automotive Engineers (SAE) defines six levels of driving automation. This ranges from 0 (fully manual) to 5 (fully autonomous), according to Synopsys.

For years, Tesla claimed it would achieve Level 5. But at best, HW3 achieved Level 2 driving, which is partial automation. Nonetheless, Tesla’s self-driving technology is controversial with a checkered safety record.

Meanwhile, in 2023, Tesla introduced its fourth-generation FSD Computer, dubbed Hardware 4.0 or A14. The A14 hardware not only includes a faster computer, but it also incorporates new sensors for the car. This hardware is incorporated in Tesla’s newer EVs. Like HW3, the A14 hardware is capable of Level 2 partial automation.

The A14 onboard computer consists of a faster SoC. This Tesla-designed SoC consists of a 2.35GHz CPU with 20 cores and three neural network processors, according to Munro Live. The device is reportedly manufactured using Samsung’s 7nm process.

What’s next? Tesla will soon release its fifth-generation FSD Computer, dubbed Hardware 5.0 or A15, which is due out in 2026 or so. Tesla is designing a new SoC for the A15 hardware. TSMC will manufacture the A15-based SoC, reportedly based on its 3nm process. Tesla hasn’t provided any details about the new SoC.

Then, in R&D, Tesla is working on a next-generation FSD Computer, dubbed A16. Samsung will make Tesla’s A16 SoC, based on its 2nm gate-all-around (GAA) transistor technology.

A transistor is a key building block in chips. Transistors are tiny structures, which are used to switch the electrical signals in chips. Today’s advanced chips have billions of transistors all on the same device.

Today, the mainstream high-performance transistor type is called the finFET. GAA, a next-generation transistor type, provides more performance than finFETs but it is harder to produce in the fab.

Samsung and others face several challenges in making automotive chips using GAA transistors. The automotive industry demands high chip yields with zero defects.

FinFET transistors vs. gate-all-around (GAA) transistors Source: Lam Research

Pressing questions and answers

After providing some background on Samsung and Tesla, let’s get back to the questions listed above. Here are six pressing questions and answers regarding the Tesla-Samsung foundry deal:

1-Why is the A16 SoC important for Tesla and Samsung?

For Samsung, the A16 SoC is based on its new 2nm process. Samsung has invested billions of dollars to develop this technology. The company will likely develop its own 2nm SoC for its future smartphones. That way, it can remain competitive and keep up with Apple and others in the smartphone market.

The 2nm process is also critical for Samsung Foundry. It hopes to garner foundry customers (i.e. Tesla) to recoup its R&D investments. 2nm technology will also allow Samsung to compete in the high-margin leading-edge foundry business.

For its part, Tesla wants to provide more self-driving capabilities for its future EVs and robotaxis. This will require several new breakthroughs, including new and faster FSD chips (i.e. A15 and A16 SoCs).

As stated, TSMC will make the A15 SoC for Tesla. Then, Tesla will move back to Samsung for the A16 chip. Tesla may have selected Samsung for the A16 SoC, simply because Samsung offered a better price and more capacity. Samsung will also allow Musk to take an active role in the fab operations.

That would never happen at TSMC. Plus, TSMC’s prices may have been too high for Tesla’s tastes. TSMC’s pricing for its 2nm process is a whopping $30,000 per wafer or higher, according to Counterpoint Research. Incidentally, Tesla never seriously considered Intel for the A16 foundry business. Intel has no real track record in the foundry market.

There is another issue at play here. Tesla developed an in-house supercomputer, which was designed to train the AI models for its self-driving technologies. The supercomputer, called Dojo, is based on a Tesla-designed AI processor.

Tesla is also separately designing the A16 SoC, which is expected to be used for AI inference capabilities in the car.

For Tesla, it doesn’t make sense to have two AI chip architectures--Dojo and A16--in development. So, Tesla has recently disbanded the Dojo supercomputer unit.

Tesla is incorporating its two AI chip architectures into a single design based on the A16 SoC. “Built on Samsung’s 2nm‑class SF2 process, A16 unifies Tesla’s Dojo AI training architecture and vehicle/robot inference hardware into one design. This eliminates the need to adapt models between training and deployment, thereby streamlining the development process,” said Jon Peddie of Jon Peddie Research.

2-Will the deal with Tesla help revive Samsung’s loss-ridden foundry unit? (Or will Apple save the day?)

The Tesla-Samsung foundry deal is important for both companies. The deal ensures that Tesla has a steady supply of advanced chips for its EVs, robotaxis and robots. It will also help Samsung fill its fab in Texas. And it will help boost its 2nm volumes. And needless to say, the $16.5 billion foundry deal should help the bottom line.

But Tesla alone won’t save Samsung Foundry. Samsung’s foundry unit will need other large customers to help fill its fabs and drive its 2nm volumes. It needs a high-volume foundry customer like Qualcomm.

Perhaps Apple can help matters. Samsung, according to reports, will supply its CMOS image sensors for Apple’s future iPhone18. In addition, Samsung will convert part of its existing fab in Austin, Texas to make image sensors for Apple, according to reports. That will help Samsung’s bottom line.

3-What are the future business opportunities for Samsung at Tesla?

As stated, Tesla’s A16 SoC isn’t expected to appear until 2028 or later. So, it will take some time before Samsung can generate any meaningful foundry sales here.

Even then, the volumes are expected to be modest. Generally, the volumes in the automotive chip market are relatively small, compared to those in the smartphone business.

In addition, Tesla’s EV sales have declined in recent times. It’s a challenging period for the company. First, China’s BYD recently surpassed Tesla as the world’s largest EV maker. Second, with the exception of China, the worldwide EV market is sluggish. And third, Musk’s political whims alienated some potential EV customers.

Going forward, Tesla is banking on several product areas to jumpstart its future growth. (Samsung is crossing its fingers, hoping these technologies will translate into foundry sales.) The product areas are:

*Tesla will soon ship a new and more affordable EV. It’s unclear if that will boost Tesla’s EV sales. Tesla faces increased competition in the EV market. Tariffs won’t help matters. And a federal EV tax credit is set to expire on Sept. 30, which represents a major setback for EV vendors.

*Tesla recently launched its first robotaxi service. Tesla has high hopes for the robotaxi market, but there are some issues here. Tesla’s checkered safety record is an issue. Plus, Tesla has never endorsed Lidar, a remote sensing technology used by other robotaxi providers. Lidar helps prevent collisions. It’s unclear if that will impact consumer confidence for Tesla’s robotaxi service.

*Tesla is working on humanoid robots. This market will take a long time to develop.

4-Will the Tesla deal help improve Samsung’s lackluster chip yields?

For some time, Samsung’s logic chip yields have been sub-par at best. Samsung’s yield issues have spread to other parts of the company, namely its HBM products. Yield is the percentage of usable chips produced from a wafer.

Samsung’s foundry fortunes hinge upon its ability to ship its new 2nm process on time with good yields. That’s still a work in progress. Samsung is expected to ship its new 2nm process in the second half of 2025. Presently, the yields for the 2nm process are 40%, according to KeyBanc Capital Markets. This compares to 65% at TSMC and 55% at Intel, according to KeyBanc.

Tesla’s Musk is expected to take an active role in Samsung’s fab in Texas. It might take a tough task master like Musk to bring Samsung’s chip yields up to speed.

That may not do the trick. The ability to deliver chips with good yields is a corporate mindset. TSMC seems to have found the right formula here. That’s why TSMC is the foundry of choice.

5-Will the Tesla deal help generate new business for Samsung Foundry?

Getting Tesla’s foundry business is a major coup for Samsung. But that doesn’t mean Samsung will generate any new or future foundry business. Each foundry customer is different. Each one has different and stringent requirements.

That being said, there are several companies evaluating Samsung’s new 2nm process, including Broadcom, Qualcomm and others.

6-Will the Tesla deal help Samsung’s foundry unit close the gap with TSMC?

This question involves two separate components: 1) technology gap; 2) market share gap.

Today, Intel, Samsung and TSMC are racing each other to ship their respective 2nm technologies in the market. 2nm represents the state-of-the-art in process technology. Right now, no one is in mass production—yet. So, it’s unclear who is leading the technology race right now.

TSMC, however, says that it has a long list of customers lining up for its 2nm technology. If that is the case, TSMC will likely dominate the 2nm foundry business. Samsung could gain some 2nm share, if it can execute. The same is true with Intel.

There is another issue at play here. For years, TSMC has been the foundry of choice at the leading edge, but customers would like to have other options here, especially at the 2nm node.

There is an opportunity here for both Samsung and Intel—if they can execute. There are other options as well. Rapidus, a foundry startup in Japan, hopes to ship its 2nm process by 2027. That’s easier said than done.

SMIC, China’s largest foundry vendor, is developing a 5nm process with 3nm and 2nm in R&D. SMIC, however, is unable to obtain select leading-edge equipment, which will stymie its efforts at the leading edge.

All told, it’s unclear which non-TSMC foundry company will step up to the leading-edge plate. But one thing is clear: It will remain difficult to compete, and gain any share, against mighty TSMC.

Conclusion

To be sure, Samsung has not been a smashing success in the foundry business. But it could reverse its misfortunes with a successful launch in the 2nm arena.

My prediction: Samsung will initially struggle at 2nm and will fall to third place in the foundry rankings, behind TSMC and SMIC. Later, Samsung will figure 2nm out.

Another prediction: Intel will likely get a government bailout, but it will still struggle and lose money in the foundry business.

Tesla also faces some challenges. And it is also losing share. China’s BYD maintained its lead over Tesla in EV sales during 1Q25, claiming the top spot with a 15.4% market share, while Tesla followed with 12.6%, according to TrendForce (See chart below).

For Tesla, it’s hard to regain the hearts and minds of consumers. But a new and affordable EV might do the trick.