DEEP RESEARCH · SAMSUNG ELECTRONICS
Samsung-Tesla Alliance: What the AI6 Chip Order Means for Semiconductors
A 22.8 trillion won long-term foundry contract viewed through AI6, 2nm GAA, and packaging value chains.
Investment decisions are your responsibility. This material is research, not a buy or sell recommendation.
0. Bottom line first
This order means more than revenue for Samsung Foundry. If Tesla AI6 scales across vehicles, Optimus, and data centers, Samsung has to prove both 2nm GAA process credibility and advanced-packaging reliability.
- The source summarizes the contract at about KRW 22.8 trillion, or roughly $16.5 billion.
- The contract period is described as July 2025 through December 31, 2033.
- The post says the value equals 7.6% of Samsung Electronics' 2024 annual revenue and is the largest single-customer foundry order in Samsung Foundry history.
1. Tesla's silicon strategy
Official fact: The post says Tesla initially used external solutions such as Mobileye EyeQ3 and Nvidia-based HW2.5, then announced the Samsung 14nm FSD chip, or HW3, in 2019.
Official fact: HW3 is described as having dual neural-network accelerators and 2,300 frames-per-second image processing, 21 times better than the previous generation.
Tesla AI hardware roadmapParallel evolution of edge inference and data-center training
HW3Samsung 14nm FSD chip
AI4Samsung-built vehicle inference
AI5TSMC Taiwan and Arizona expected
AI6Samsung next-generation process
Tesla uses both TSMC and Samsung to diversify technology, pricing, and geopolitical risk.
2. AI6: scaled down and scaled up
Official fact: The source frames scalability as AI6's core design philosophy: low-power inference for FSD and Optimus, and hundreds or thousands of chips in data centers for training infrastructure.
Scaled Down
Vehicles and Optimus
Low-power, efficient inference matters. The chip acts as the real-time edge decision engine.
Scaled Up
AI data centers
The concept is to combine many chips into Dojo-like training infrastructure.
Supply Chain
Foundry diversification
The strategy reduces dependence on one foundry by using Taiwan, the U.S., and Korea production axes.
3. Technologies Samsung must deliver
Official fact: The post connects AI6 production to Samsung 2nm GAA/SF2, EUV tools, EUV photoresist, 2.5D interposers, FOPLP, HBM TC bonding, large FC-BGA substrates, DSP/VDP, and EDA tools.
| Value chain | Technology/contribution | Key company | Country |
|---|---|---|---|
| Design | Scalable AI architecture | Tesla | U.S. |
| Foundry | 2nm GAA process, SF2 | Samsung Electronics | Korea |
| Wafer | 2nm-class silicon wafer | SK Siltron | Korea |
| EUV tools | EUV lithography scanner | ASML | Netherlands |
| EUV materials | EUV photoresist | JSR / Dongjin Semichem | Japan / Korea |
| Advanced packaging | 2.5D interposer, I-CubeS | Samsung Electronics | Korea |
| Back-end, substrate, design service | FOPLP, HBM TC bonding, FC-BGA, DSP/VDP, EDA | Needs estimation | Needs estimation |
4. My interpretation
Interpretation: Tesla appears to be designing AI6 not as one chip for one product, but as a platform spanning robots, cars, and data centers. Samsung now has to use this contract to show credible 2nm GAA manufacturing and turnkey packaging competitiveness.
Sources
- Source 1: https://m.sbs.co.kr/search/allSearchList.do?collection=nnews&sort=DATE&query=%EB%A8%B8%EC%8A%A4%ED%81%AC&m_searchclick=all_news_more_00_00
- Source 2: https://v.daum.net/v/20250728131340219
- Source 3: https://www.munhwa.com/article/11522046
- Source 4: https://www.g-enews.com/article/Global-Biz/2025/07/202507281323418666fbbec65dfb_1
- Source 5: https://www.newsspirit.kr/news/articleView.html?idxno=23314
- Source 6: https://www.inews24.com/view/1870151
- Source 7: https://www.yna.co.kr/view/AKR20250728028753003
- Source 8: https://tripuratimes.com/ttimes/musk-says-samsung-will-produce-teslas-ai6-chip-at-us-plant-29971.html
- Source 9: https://en.wikipedia.org/wiki/Tesla_Autopilot_hardware
- Source 10: https://semiwiki.com/forum/threads/tesla-ai5-hardware-n3-silicon-slips-one-year-ai6-to-scale-from-robots-to-ai-data-center.23229/
- Source 11: https://www.tesla.com/AI
- Source 12: https://en.wikipedia.org/wiki/Tesla_Dojo
- Source 13: https://datasciencedojo.com/blog/what-is-tesla-dojo/
- Source 14: https://www.infoq.com/news/2021/09/tesla-dojo-ai-models/
- Source 15: https://www.businesspost.co.kr/BP?command=sub&sub=8
- Source 16: https://www.notateslaapp.com/news/2428/un
- Source 17: https://applyingai.com/2025/07/decoding-teslas-core-ai-and-hardware-architecture-a-ceos-perspective/
- Source 18: https://library.techinsights.com/hg-asset/e2ce0c3b-93ff-4137-8c47-db2dabc5e570
- Source 19: https://awavesemi.com/resource/advancing-ai-the-role-of-hbm-in-the-chiplet-ecosystem-era/
- Source 20: https://www.counterpointresearch.com/insight/memory-solutions-for-gen-ai-part-6-edge-computing
- Source 21: https://news.samsung.com/global/samsung-electronics-to-provide-turnkey-semiconductor-solutions-with-2nm-gaa-process-and-2-5d-package-to-preferred-networks
- Source 22: https://semiconductor.samsung.com/foundry/advanced-package/
- Source 23: https://www.hpcwire.com/off-the-wire/synopsys-expands-samsung-collaboration-to-advance-ai-and-hpc-chip-designs/
- Source 24: https://indianexpress.com/article/technology/tech-news-technology/musk-says-tesla-samsung-electronics-sign-16-5-billion-chip-supply-deal-10154024/
- Source 25: https://www.financialexpress.com/life/technology-tesla-partners-with-samsung-for-long-term-chip-supply-says-elon-musk-3928601/
- Source 26: https://www.chinadailyasia.com/article/616820
- Source 27: https://gulfnews.com/technology/tesla-taps-samsung-for-ai6-chips-in-165-billion-deal-elon-musk-says-1.500213371
- Source 28: https://www.ainvest.com/news/tesla-samsung-strategic-alliance-revolutionizing-ai-semiconductors-accelerating-autonomous-driving-2507/
- Source 29: https://resources.system-analysis.cadence.com/blog/msa2022-comparing-finfets-vs-gaafets
- Source 30: https://techlevated.com/gaafet-vs-finfet-transistor/
- Source 31: https://electronics.stackexchange.com/questions/714379/can-someone-explain-the-working-operation-of-finfet
- Source 32: https://anysilicon.com/the-ultimate-guide-to-gate-all-around-gaa/
- Source 33: https://semiengineering.com/knowledge_centers/integrated-circuit/transistors/3d/gate-all-around-fet/
- Source 34: https://www.edn.com/the-2-nm-process-node-and-samsungs-foundry-crossroads/
- Source 35: https://www.fierceelectronics.com/ai/samsung-speeds-production-ai-chips-making-use-new-gaa-tech
- Source 36: https://semianalysis.com/2024/10/01/clash-of-the-foundries/
- Source 37: https://www.reddit.com/r/hardware/comments/v6p01c/eetimes_tsmc_commits_to_nanosheet_technology_at_2/
- Source 38: https://www.anandtech.com/show/21444/samsung-foundry-unveils-updated-roadmap-2nm-evolution-through-2027
- Source 39: https://zdnet.co.kr/view/?no=20240820092613
- Source 40: https://www.tsmc.com/english/dedicatedFoundry/technology/logic/l_2nm
- Source 41: https://www.astutegroup.com/news/industrial/tsmcs-2nm-venture-a-testament-to-innovation-amid-complexity/
- Source 42: https://en.wikipedia.org/wiki/2_nm_process
- Source 43: https://www.kontronn.com/hardware/tsmc-2nm-process-pilot-production-has-high-yield-rates-of-90-percent/
- Source 44: https://www.reddit.com/r/electronics/comments/1jvye6v/taiwans_2nm_chip_can_be_a_game_changer_in_tech/
- Source 45: https://www.tomshardware.com/tech-industry/semiconductors/intel-might-axe-the-18a-process-node-for-foundry-customers-essentially-leaving-tsmc-with-no-rival-intel-reportedly-to-focus-on-14a
- Source 46: https://www.hwcooling.net/en/intel-cancels-20a-node-meant-for-arrow-lake-cpu-production/
- Source 47: https://www.techpowerup.com/326289/intel-20a-node-cancelled-for-foundry-customers-arrow-lake-mainly-manufactured-externally
- Source 48: https://www.catch.co.kr/News/RecruitNews/296064
- Source 49: https://sitemap.zdnet.co.kr/view/?no=20240820092613
- Source 50: https://english.hani.co.kr/arti/english_edition/e_business/923990.html
- Source 51: https://www.digitimes.com/news/a20241119PD214/jsr-photoresist-euv-plant-japan.html
- Source 52: https://koreajoongangdaily.joins.com/2020/07/02/business/industry/photoresist-TOK-EUV/20200702200000423.html
- Source 53: https://dongjin.com/promotion/news_view.php?seq=129&s_keyname&s_keyword&page=3
- Source 54: https://www.hankyung.com/article/2022040426471
- Source 55: http://www.wikileaks-kr.org/news/articleView.html?idxno=129334
- Source 56: https://www.etoday.co.kr/news/view/2355227
- Source 57: http://www.dailyinvest.kr/news/articleView.html?idxno=50603
- Source 58: https://www.cadgraphics.co.kr/newsview.php?pages&sub=lecture03&catecode=9&num=70215
- Source 59: https://mbiz.heraldcorp.com/article/10496296
- Source 60: https://www.sedaily.com/NewsView/2DACIG0GQ0
- Source 61: https://aei.dempa.net/archives/29283
- Source 62: https://semiconductor.samsung.com/technologies/package/
- Source 63: https://semiconductor.samsung.com/foundry/advanced-package/advanced-heterogeneous-integration/
- Source 64: https://www.donga.com/news/Economy/article/all/20210429/106671833/1
- Source 65: https://zdnet.co.kr/view/?no=20240307144711
- Source 66: https://www.nepes.co.kr/kr/rnd/electronic_devices.php
- Source 67: https://mse-semi.tistory.com/entry/%EB%B0%98%EB%8F%84%EC%B2%B4%EC%A7%80%EC%8B%9D-HBM-2
- Source 68: https://tristanchoi.tistory.com/458
- Source 69: https://www.businesspost.co.kr/BP?command=article_view&num=330599
- Source 70: https://dmi.kims.re.kr/bbs/board.php?bo_table=05_02_02&wr_id=3045&sca=%EC%82%B0%EC%97%85&page=12
- Source 71: https://www.chosun.com/english/nsearch/?query=FC-BGA&siteid&sort=1&date_period=all&writer&field&emd_word&expt_word&opt_chk=false&app_check=0&website=www,chosun
- Source 72: https://www.etnews.com/20240329000014
- Source 73: https://heraldk.com/2025/05/27/%EC%82%BC%EC%84%B1-%ED%8C%8C%EC%9A%B4%EB%93%9C%EB%A6%AC-%EB%95%80%EB%82%98%EA%B2%8C-%EB%9B%B0%EC%97%88%EB%8B%A4%E2%80%A6%ED%8C%8C%ED%8A%B8%EB%84%88%EC%82%AC-1%EB%85%84%EC%83%88-%EB%8C%80%ED%8F%AD/
- Source 74: https://theguru.co.kr/news/article.html?no=83943