The AI6 chip will be built on Samsung's advanced 2nm Gate-All-Around (GAA) process, expected to deliver roughly double the performance of Tesla's AI5 chip at the same die size . Samsung's Taylor plant, a $37 billion investment, held its equipment move-in ceremony on April 24, 2026, and is targeting full-scale mass production in the second half of 2026 .

Musk later clarified that both Samsung and TSMC will produce versions of Tesla's AI5 and AI6 chips, giving the automaker a diversified supply base. "Tesla is using both," he confirmed in November 2025 . Samsung has also reportedly agreed to allow Tesla to assist in maximizing manufacturing efficiency at the Taylor site .

BYD: Autonomous Driving Chips in Active Discussion

Chinese EV giant BYD is in talks with Samsung Foundry to manufacture autonomous-driving system-on-chips (SoCs) on both 4nm and 2nm processes. The discussions, first reported by Seoul Economic Daily in June 2026, cover chips that would power advanced driver-assistance systems for BYD and potentially other Chinese automakers .

BYD recently unveiled its Xuanji A3, which it calls China's first 4nm intelligent driving chip, supporting Level 3 and Level 4 autonomy . However, BYD lacks in-house 4nm manufacturing capability, and China's largest domestic foundry, SMIC, is constrained at advanced nodes. Industry sources note that SMIC's 7nm capacity is largely allocated to key clients like Huawei, leaving its remaining output concentrated on 14nm and older processes . This makes Samsung "the most realistically feasible alternative to TSMC" for Chinese automakers, according to one industry insider .

No contract has been finalized yet. If signed, BYD would become Samsung's second major automotive customer after Tesla.

Google: A Split-Manufacturing Plan for Next-Gen TPU

Google is considering Samsung to manufacture a key component of its upcoming 10th-generation Tensor Processing Unit (TPU), codenamed "Icefish." According to a June 2026 report from The Information, the proposed plan would split manufacturing across two foundries :

• TSMC would produce the main compute die on its 1.4nm process.
• Samsung would fabricate the memory I/O die on its 2nm process, connecting the processor to High Bandwidth Memory (HBM).

The deal represents a strategic shift for Google. It would mark the first time the company outsourced part of its TPU production from TSMC, driven by capacity constraints at TSMC's most advanced nodes . The memory I/O die is a critical but less performance-intensive component, making Samsung's 2nm process a suitable fit .

The talks remain preliminary, with mass production targeted as early as 2028. Google is also reportedly working with MediaTek and discussing large-scale TPU production with Intel . Samsung currently supplies over 60% of the HBM for Google's seventh-generation Ironwood TPU, giving it an existing relationship advantage .

AMD's EPYC Venice: No 2nm Deal with Samsung

Despite persistent rumors, there is no evidence that AMD has shifted its flagship EPYC Venice server CPU to Samsung Foundry. The 6th-generation EPYC "Venice" processor has entered volume production on TSMC's N2 2nm-class node, not Samsung's .

In May 2026, AMD confirmed that Venice, a 256-core Zen 6 processor, was the first high-performance computing product in the industry to reach production on N2 . In a March 2026 interview, an AMD Senior Vice President stated that "from AMD's perspective, TSMC is currently the undisputed leader in 2nm process technology" .

Earlier speculation stemmed from reports in December 2025 that Samsung Foundry was in discussions with AMD about a potential 2nm deal using its SF2P process, with a decision expected around January 2026 . Those talks did not translate into a confirmed contract for Venice. AMD had previously terminated its 4nm cooperation with Samsung Foundry in favor of TSMC for EPYC server CPUs, a decision reported in May 2025 .

A single February 2026 source claiming AMD "officially confirmed" a shift to Samsung appears to contradict the weight of more recent and higher-trust evidence . The consensus among reliable industry sources is clear: EPYC Venice is a TSMC product.

Samsung Foundry's Profitability Outlook

No source in the provided evidence explicitly confirms a Q3 2026 return to profitability for Samsung Foundry. However, the business landscape has improved markedly.

The Tesla AI6 contract provides Samsung with the marquee customer it needed to justify its Taylor, Texas investment. Before the Tesla deal, Samsung had delayed equipment deliveries and pushed back the facility's operational launch due to a lack of major customers . Now, the Taylor fab is preparing for high-volume 2nm production, with Tesla reportedly seeking to scale output to 40,000 wafers per month .

Additional customer discussions—with BYD for automotive SoCs and Google for a TPU I/O die—signal growing demand for Samsung's 4nm and 2nm offerings. Analysts have noted that if the Google order is confirmed, "the turnaround to profit at Samsung Electronics' foundry and System LSI divisions" could accelerate .

Still, the path to profitability depends on signing the deals currently under negotiation and executing smoothly on the Tesla ramp. The outlook is more constructive than a year ago, but a specific Q3 2026 profit inflection remains a positive scenario rather than a confirmed target.

Other Companies: Nvidia and Groq

Reports also indicate that Samsung Foundry is producing the Groq LP30 AI accelerator chip for Nvidia using 4nm and 8nm processes, and is involved in manufacturing for Nvidia's Drive AGX Thor autonomous driving platform . Discussions about next-generation projects between Nvidia and Samsung are reportedly ongoing, though the provided sources offer limited detail on the scope or status of any future deals.

The AI6 chip will be built on Samsung's advanced 2nm Gate-All-Around (GAA) process, expected to deliver roughly double the performance of Tesla's AI5 chip at the same die size . Samsung's Taylor plant, a $37 billion investment, held its equipment move-in ceremony on April 24, 2026, and is targeting full-scale mass production in the second half of 2026 .

Musk later clarified that both Samsung and TSMC will produce versions of Tesla's AI5 and AI6 chips, giving the automaker a diversified supply base. "Tesla is using both," he confirmed in November 2025 . Samsung has also reportedly agreed to allow Tesla to assist in maximizing manufacturing efficiency at the Taylor site .

BYD: Autonomous Driving Chips in Active Discussion

Chinese EV giant BYD is in talks with Samsung Foundry to manufacture autonomous-driving system-on-chips (SoCs) on both 4nm and 2nm processes. The discussions, first reported by Seoul Economic Daily in June 2026, cover chips that would power advanced driver-assistance systems for BYD and potentially other Chinese automakers .

BYD recently unveiled its Xuanji A3, which it calls China's first 4nm intelligent driving chip, supporting Level 3 and Level 4 autonomy . However, BYD lacks in-house 4nm manufacturing capability, and China's largest domestic foundry, SMIC, is constrained at advanced nodes. Industry sources note that SMIC's 7nm capacity is largely allocated to key clients like Huawei, leaving its remaining output concentrated on 14nm and older processes . This makes Samsung "the most realistically feasible alternative to TSMC" for Chinese automakers, according to one industry insider .

No contract has been finalized yet. If signed, BYD would become Samsung's second major automotive customer after Tesla.

Google: A Split-Manufacturing Plan for Next-Gen TPU

Google is considering Samsung to manufacture a key component of its upcoming 10th-generation Tensor Processing Unit (TPU), codenamed "Icefish." According to a June 2026 report from The Information, the proposed plan would split manufacturing across two foundries :

• TSMC would produce the main compute die on its 1.4nm process.
• Samsung would fabricate the memory I/O die on its 2nm process, connecting the processor to High Bandwidth Memory (HBM).

The deal represents a strategic shift for Google. It would mark the first time the company outsourced part of its TPU production from TSMC, driven by capacity constraints at TSMC's most advanced nodes . The memory I/O die is a critical but less performance-intensive component, making Samsung's 2nm process a suitable fit .

The talks remain preliminary, with mass production targeted as early as 2028. Google is also reportedly working with MediaTek and discussing large-scale TPU production with Intel . Samsung currently supplies over 60% of the HBM for Google's seventh-generation Ironwood TPU, giving it an existing relationship advantage .

AMD's EPYC Venice: No 2nm Deal with Samsung

Despite persistent rumors, there is no evidence that AMD has shifted its flagship EPYC Venice server CPU to Samsung Foundry. The 6th-generation EPYC "Venice" processor has entered volume production on TSMC's N2 2nm-class node, not Samsung's .

In May 2026, AMD confirmed that Venice, a 256-core Zen 6 processor, was the first high-performance computing product in the industry to reach production on N2 . In a March 2026 interview, an AMD Senior Vice President stated that "from AMD's perspective, TSMC is currently the undisputed leader in 2nm process technology" .

Earlier speculation stemmed from reports in December 2025 that Samsung Foundry was in discussions with AMD about a potential 2nm deal using its SF2P process, with a decision expected around January 2026 . Those talks did not translate into a confirmed contract for Venice. AMD had previously terminated its 4nm cooperation with Samsung Foundry in favor of TSMC for EPYC server CPUs, a decision reported in May 2025 .

A single February 2026 source claiming AMD "officially confirmed" a shift to Samsung appears to contradict the weight of more recent and higher-trust evidence . The consensus among reliable industry sources is clear: EPYC Venice is a TSMC product.

Samsung Foundry's Profitability Outlook

No source in the provided evidence explicitly confirms a Q3 2026 return to profitability for Samsung Foundry. However, the business landscape has improved markedly.

The Tesla AI6 contract provides Samsung with the marquee customer it needed to justify its Taylor, Texas investment. Before the Tesla deal, Samsung had delayed equipment deliveries and pushed back the facility's operational launch due to a lack of major customers . Now, the Taylor fab is preparing for high-volume 2nm production, with Tesla reportedly seeking to scale output to 40,000 wafers per month .

Additional customer discussions—with BYD for automotive SoCs and Google for a TPU I/O die—signal growing demand for Samsung's 4nm and 2nm offerings. Analysts have noted that if the Google order is confirmed, "the turnaround to profit at Samsung Electronics' foundry and System LSI divisions" could accelerate .

Still, the path to profitability depends on signing the deals currently under negotiation and executing smoothly on the Tesla ramp. The outlook is more constructive than a year ago, but a specific Q3 2026 profit inflection remains a positive scenario rather than a confirmed target.

Other Companies: Nvidia and Groq

Reports also indicate that Samsung Foundry is producing the Groq LP30 AI accelerator chip for Nvidia using 4nm and 8nm processes, and is involved in manufacturing for Nvidia's Drive AGX Thor autonomous driving platform . Discussions about next-generation projects between Nvidia and Samsung are reportedly ongoing, though the provided sources offer limited detail on the scope or status of any future deals.