# Samsung

> Source: https://aiwiki.ai/wiki/samsung
> Updated: 2026-06-23
> Categories: AI Companies, AI Hardware
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution to "AI Wiki (aiwiki.ai)".

Samsung Electronics is a South Korean electronics manufacturer headquartered in Suwon, the flagship company of the Samsung Group conglomerate and the world's largest maker of memory chips, the components that store data inside AI servers. Samsung sits at more points of the AI value chain than almost any other company: it supplies the [high-bandwidth memory](/wiki/high_bandwidth_memory) (HBM) that feeds [AI accelerators](/wiki/ai_chips), operates the world's second-largest contract chip foundry, ships AI-branded consumer devices under the Galaxy AI brand, runs its own Samsung Gauss large language models, and invests in humanoid and home robotics. Its AI story from 2023 to 2026 has been one of contrasts: a celebrated consumer AI launch with the Galaxy S24, a painful multi-year stumble in HBM that ceded leadership to [SK Hynix](/wiki/sk_hynix), and a recovery capped by becoming the first supplier to ship HBM4 and by a $16.5 billion [Tesla](/wiki/tesla) foundry contract. In the first quarter of 2026 Samsung held the largest share of the global DRAM market at 38.6 percent, ahead of SK Hynix and [Micron](/wiki/micron), even as it ranked a clear third in the narrower HBM segment that powers frontier AI training. [24]

## Overview

Samsung Electronics was founded in 1969 and is led by executive chairman [Lee Jae-yong](/wiki/lee_jae_yong). The company is organized into two main arms: the Device eXperience (DX) division, covering smartphones, TVs, and appliances, and the Device Solutions (DS) division, covering memory, system LSI (chip design), and foundry. Jun Young-hyun, appointed semiconductor chief in May 2024 amid the HBM crisis, leads DS. [5]

For fiscal 2025 Samsung reported record annual revenue of KRW 333.6 trillion and operating profit of KRW 43.6 trillion, with the fourth quarter setting all-time highs of KRW 93.8 trillion revenue and KRW 20.1 trillion operating profit on surging HBM and DRAM sales. [22] Within that quarter the Device Solutions division alone generated KRW 44.0 trillion in revenue and KRW 16.4 trillion in operating profit, with the memory business reaching record quarterly revenue and profit despite limited supply availability. [22] The AI memory supercycle continued into 2026: first-quarter operating profit rose more than eightfold year on year to a record as AI demand created a memory supply crunch, and industry-wide global DRAM revenue hit a record $97.1 billion in the quarter, up 85.3 percent sequentially. [23][24]

## What is Galaxy AI?

Galaxy AI is Samsung's brand for the on-device and cloud generative-AI features built into its Galaxy smartphones, tablets, and wearables. Samsung introduced it alongside the Galaxy S24 series, announced on January 17, 2024 at its Unpacked event in San Jose. [2] The launch was paired with a multi-year partnership with [Google](/wiki/google) Cloud that made Samsung the first Google Cloud partner to deploy [Gemini](/wiki/gemini) Pro and the [Imagen 2](/wiki/imagen_2) image model through [Vertex AI](/wiki/vertex_ai) on smartphones, while [Gemini Nano](/wiki/gemini_nano) ran on-device for tasks such as summarization in Notes, Voice Recorder, and the keyboard. [1][2] Headline S24 features included Live Translate for two-way call translation in up to 13 languages, Interpreter, Chat Assist, Note and Transcript summarization, Generative Edit photo fill, and the debut of Circle to Search, co-launched with Google. [1][2] Samsung also committed to seven generations of OS upgrades and seven years of security updates for the S24. [1]

Samsung pursues a hybrid strategy that mixes partner models with its own: cloud-based Gemini for heavier tasks, and on-device processing for latency- and privacy-sensitive ones. Its internal foundation model family, Samsung Gauss, was unveiled at the Samsung AI Forum on November 8, 2023 with language, code, and image variants; the code model, called code.i, is used internally for software development, and the language model supports Korean, English, French, Spanish, Chinese, and Japanese. [3] The second-generation Gauss2, announced at the Samsung Developer Conference Korea on November 25, 2024, is a multimodal model offered in Compact (on-device), Balanced, and Supreme (mixture-of-experts) configurations; it integrates language, code, and image processing, supports 9 to 14 languages depending on the variant, and runs 1.5 to 3 times faster than comparable open models on key metrics, according to Samsung. [3]

Distribution is Samsung's core advantage. Galaxy AI reached more than 200 million devices in 2024 via flagship launches and One UI updates, and at its July 2025 Unpacked the company committed to 400 million Galaxy AI devices by the end of 2025; Samsung also said more than 70 percent of Galaxy S25 users engaged with Galaxy AI features, with Circle to Search used by 54 percent of S25 owners and Gemini usage tripling versus the prior generation. [4] The Galaxy S25 series (January 2025) deepened the Google integration, making Gemini the default assistant and adding agentic cross-app actions and the personalized Now Brief. The Galaxy S26 series, unveiled at an AI-focused Unpacked in San Francisco on February 25, 2026, introduced the Exynos 2600, the industry's first 2nm gate-all-around mobile processor, which powers the S26 and S26+ in most markets and brings an NPU with a 113 percent improvement in generative-AI performance over its predecessor. [17] Samsung has extended AI branding across the home as well, with Bespoke AI appliances and ambient intelligence features in its SmartThings platform.

## Why did Samsung fall behind SK Hynix in HBM?

Samsung is the world's largest memory chipmaker, but it fell conspicuously behind in the memory segment that matters most for AI. SK Hynix co-developed HBM with [AMD](/wiki/amd) in the 2010s, invested through the downturn, and became NVIDIA's primary supplier for the H100 and H200 era, while Micron also qualified ahead of Samsung on HBM3E. Samsung's fifth-generation HBM3E repeatedly failed [NVIDIA](/wiki/nvidia)'s qualification tests, with reports citing heat and power-consumption issues. The company replaced its semiconductor chief with Jun Young-hyun in May 2024, and on October 7, 2024 he issued a rare public apology after disappointing results, acknowledging delayed HBM3E sales to a major customer. "We have caused concerns about our technical competitiveness, with some talking about the crisis facing Samsung," Jun told investors. [5][25]

## Is Samsung supplying HBM to NVIDIA?

Yes, after a multi-year delay Samsung now supplies NVIDIA at the leading edge, and as of early 2026 it was the first company to ship HBM4. The turnaround came in stages. While NVIDIA qualification dragged on, Samsung shipped HBM3E to other buyers, including AMD's Instinct accelerator line, according to industry reports. In September 2025, roughly 18 months after the part was developed, Samsung's 12-layer HBM3E finally passed NVIDIA's tests, clearing it to supply the major HBM customers as the third qualified supplier behind SK Hynix and Micron; the chips target accelerators such as NVIDIA's B300 and AMD's MI350, and NVIDIA shipments began in the third quarter of 2025. [6][7] Samsung then moved aggressively on HBM4, pairing sixth-generation 1c DRAM with a 4nm logic base die from its own foundry, an integration competitors must source externally. Its HBM4 reportedly ranked first among suppliers in NVIDIA's qualification, achieving 11.7 Gbps pin speeds against a 10 Gbps requirement and passing without a redesign, and in February 2026 Samsung became the first company in the industry to begin mass shipments of HBM4 to NVIDIA and AMD, with the parts slated for NVIDIA's [Vera Rubin](/wiki/vera_rubin) platform. [8][9] Industry trackers still expect SK Hynix to take roughly two-thirds of NVIDIA's HBM4 allocation, with Samsung's initial share projected in the mid-20-percent range. [10]

Samsung also anchored itself in [OpenAI](/wiki/openai)'s buildout: in October 2025 Samsung Electronics signed a letter of intent to serve as a strategic memory partner for the [Stargate](/wiki/stargate) initiative, against OpenAI demand projected to reach up to 900,000 DRAM wafers per month, a volume that analysts estimated could equal as much as 40 percent of global DRAM output. [11][26] Affiliates Samsung SDS (data-center design and enterprise AI services) and Samsung C&T and Samsung Heavy Industries (floating data centers) joined the program, and OpenAI explored Stargate data centers in South Korea. [11][12] Together, Samsung and SK Hynix supply roughly 70 percent of the world's DRAM and about 80 percent of HBM, making the two Korean firms central to AI infrastructure planning. [26]

| Date | HBM milestone |
| --- | --- |
| May 2024 | Chip chief replaced by Jun Young-hyun amid HBM3E qualification failures [5] |
| October 2024 | Public apology over AI-memory delays and weak results [5][25] |
| September 2025 | 12-layer HBM3E passes NVIDIA qualification as third supplier [6][7] |
| October 2025 | Strategic memory partner for OpenAI's Stargate [11] |
| February 2026 | First in industry to ship HBM4, to NVIDIA and AMD at 11.7 Gbps [8][9] |

## What is the Samsung-Tesla foundry deal?

Samsung Foundry is the world's second-largest contract chipmaker but a distant runner-up to [TSMC](/wiki/tsmc), and it entered 2025 with low utilization and persistent losses after yield struggles on advanced nodes pushed marquee customers toward its rival. The inflection came on July 28, 2025, when Samsung disclosed a 22.8 trillion won (about $16.5 billion) foundry contract running through the end of 2033. [Elon Musk](/wiki/elon_musk) confirmed the customer was Tesla and that Samsung's new fab in Taylor, Texas would be dedicated to Tesla's next-generation AI6 chip, a part designed for inference workloads that will power Tesla's [Full Self-Driving](/wiki/full_self_driving) system and its Optimus humanoid robots. "Samsung's giant new Texas fab will be dedicated to making Tesla's next-generation AI6 chip. The strategic importance of this is hard to overstate," Musk wrote, adding that "the $16.5bn number is just the bare minimum" and that Tesla would help maximize manufacturing efficiency. [13][14][27] Samsung already manufactured Tesla's AI4 Full Self-Driving computer, with AI5 then planned for TSMC. [14] The relationship subsequently widened: Musk later confirmed that both AI5 and AI6 will be dual-sourced from Samsung and TSMC, with AI5 production starting in small volumes in late 2026 and high volume in 2027, and he remarked in October 2025 that the Taylor fab's equipment is slightly more advanced than TSMC's Arizona plant. [15]

The Taylor site anchors Samsung's more than $37 billion Texas investment, supported by a finalized $4.745 billion direct award under the [CHIPS and Science Act](/wiki/chips_act) covering two leading-edge logic fabs and an R&D fab in Taylor plus an Austin expansion, with all facilities expected to be operational by 2030. [16] Samsung began mass production of its first-generation 2nm process in Korea in late 2025, with the Exynos 2600 as lead product, and began shipping 4nm HBM4 base dies, while Taylor ramps 2nm production for customers including Tesla in the second half of 2026. [17][22]

## How is Samsung using AI in robotics and the home?

Samsung has pushed into robotics as a designated growth engine. In December 2024 it exercised call options to raise its stake in [Rainbow Robotics](/wiki/rainbow_robotics), the KAIST spinoff founded in 2011 by researchers who built Korea's first bipedal humanoid HUBO, from 14.7 percent to 35 percent for KRW 267 billion (about $181 million), becoming the largest shareholder and consolidating the firm as a subsidiary; regulators cleared the deal in early 2025. [18] Samsung simultaneously established a Future Robotics Office reporting directly to the CEO, with Rainbow Robotics founder Oh Jun-ho leading the effort, aiming to combine Samsung's AI and manufacturing scale with Rainbow's humanoid platforms for eventual use in manufacturing and beyond. [18][19]

Its consumer robot, Ballie, has had a rockier path. First shown as a rolling ball-shaped home companion at CES 2020 and revamped with a built-in projector at CES 2024, Ballie was slated, with Google Cloud's Gemini models on board, to launch in the United States and South Korea in summer 2025. The launch slipped repeatedly: by December 2025 Samsung confirmed the robot was still not ready, and Ballie was absent from CES 2026, where reporting indicated Samsung now describes it as an internal innovation platform rather than an imminent consumer product. [20][21]

## Why does Samsung matter for AI?

Samsung is the only company that simultaneously builds flagship AI consumer devices, manufactures the memory underpinning frontier accelerators, and operates a leading-edge logic foundry, making it a bellwether for both AI demand and AI supply. Its HBM3E saga became the industry's cautionary tale about how quickly leadership can be lost in AI-era memory, reshaping a decades-old hierarchy in which Samsung had been the unquestioned number one; its first-to-ship HBM4 comeback and record 2025-2026 earnings show how fast fortunes can reverse in a supply-constrained market. [6][22][23] The Tesla contract and the Taylor fab give Samsung Foundry its most credible path yet to relevance at the leading edge in the United States, while Galaxy AI's 400-million-device footprint makes Samsung one of the largest distribution channels for on-device and [Gemini](/wiki/gemini)-powered AI in the world. [4][13][16] With the Rainbow Robotics consolidation and the [Sam Altman](/wiki/sam_altman)-led Stargate memory partnership, Samsung has positioned itself across nearly every layer of the AI buildout, from wafers to robots.

## References

1. Samsung Newsroom. "Samsung and Google Cloud Join Forces to Bring Generative AI to Samsung Galaxy S24 Series." January 2024. https://news.samsung.com/global/samsung-and-google-cloud-join-forces-to-bring-generative-ai-to-samsung-galaxy-s24-series
2. TechCrunch. "Samsung's Galaxy S24 will feature Google Gemini-powered AI features." January 17, 2024. https://techcrunch.com/2024/01/17/samsungs-galaxy-s24-will-feature-google-gemini-powered-ai-features/
3. Samsung Newsroom. "Samsung Electronics Hosts Samsung Developer Conference Korea 2024, Unveils Its Improved Gen AI Model." November 2024. https://news.samsung.com/global/samsung-electronics-hosts-samsung-developer-conference-korea-2024-unveils-its-improved-gen-ai-model
4. Samsung Newsroom. "[Galaxy Unpacked 2025] From AI to Actionable Care: Industry Leaders Chart the Future of Mobile Innovation at Galaxy Tech Forum." July 2025. https://news.samsung.com/global/galaxy-unpacked-2025-from-ai-to-actionable-care-industry-leaders-chart-the-future-of-mobile-innovation-at-galaxy-tech-forum
5. Bloomberg News (via BusinessMirror). "Samsung makes rare apology after results sag." October 8, 2024. https://businessmirror.com.ph/2024/10/08/samsung-makes-rare-apology-after-results-sag/
6. KED Global. "Samsung clears Nvidia hurdle for 12-layer HBM3E supply, setting stage for HBM4 battle." September 19, 2025. https://www.kedglobal.com/korean-chipmakers/newsView/ked202509190008
7. Tom's Hardware. "Samsung earns Nvidia certification for its HBM3E memory; stock jumps 5%." September 2025. https://www.tomshardware.com/tech-industry/samsung-earns-nvidias-certification-for-its-hbm3-memory-stock-jumps-5-percent-as-company-finally-catches-up-to-sk-hynix-and-micron-in-hbm3e-production
8. The Korea Times. "Samsung reports solid recovery in 2025, set to deliver HBM4 starting next month." January 29, 2026. https://www.koreatimes.co.kr/business/companies/20260129/samsung-reports-solid-recovery-in-2025-set-to-deliver-hbm4-starting-next-month
9. TrendForce. "Samsung Reportedly Set to Begin Official HBM4 Shipments to NVIDIA and AMD in February." January 26, 2026. https://www.trendforce.com/news/2026/01/26/news-samsung-reportedly-set-to-begin-official-hbm4-shipments-to-nvidia-and-amd-in-february/
10. TrendForce. "SK hynix Reportedly to Supply About Two-Thirds of NVIDIA HBM4; Samsung Targets Early Delivery." January 28, 2026. https://www.trendforce.com/news/2026/01/28/news-sk-hynix-reportedly-to-supply-about-two-thirds-of-nvidia-hbm4-samsung-targets-early-delivery/
11. OpenAI. "Samsung and SK join OpenAI's Stargate initiative to advance global AI infrastructure." October 1, 2025. https://openai.com/index/samsung-and-sk-join-stargate/
12. Samsung Newsroom. "Samsung and OpenAI Announce Strategic Partnership To Accelerate Advancements in Global AI Infrastructure." October 2025. https://news.samsung.com/global/samsung-and-openai-announce-strategic-partnership-to-accelerate-advancements-in-global-ai-infrastructure
13. KED Global. "Samsung clinches $16.5 billion chipmaking deal from Tesla." July 28, 2025. https://www.kedglobal.com/korean-chipmakers/newsView/ked202507280001
14. CNN Business. "Elon Musk says Tesla and Samsung have signed a $16.5 billion chip deal." July 28, 2025. https://www.cnn.com/2025/07/28/business/tesla-samsung-chip-deal
15. Tom's Hardware. "Musk confirms Tesla AI5 and AI6 will be made at both Samsung and TSMC, reinforcing dual-foundry strategy." 2026. https://www.tomshardware.com/tech-industry/semiconductors/musk-confirms-tesla-ai5-and-ai6-will-be-made-at-samsung-and-tsmc
16. NIST CHIPS Program Office. "Samsung Electronics (Texas)." https://www.nist.gov/chips/samsung-electronics-texas-taylor
17. TrendForce. "Samsung Unveils Exynos 2600: Industry-First 2nm GAA AP." December 19, 2025. https://www.trendforce.com/news/2025/12/19/news-samsung-officially-unveils-exynos-2600-industry-first-2nm-gaa-ap-with-113-ai-performance-uplift
18. TechCrunch. "Samsung pays $181M to become largest shareholder of Rainbow Robotics." December 30, 2024. https://techcrunch.com/2024/12/30/samsung-pays-181m-to-become-largest-shareholder-of-rainbow-robotics/
19. The Robot Report. "Samsung increases stake in Rainbow Robotics, establishes Future Robotics Office." January 2025. https://www.therobotreport.com/samsung-increases-stake-in-rainbow-robotics-establishes-future-robotics-office/
20. 9to5Google. "Samsung Ballie launch window flies by, and it's still delayed." December 8, 2025. https://9to5google.com/2025/12/08/samsung-ballie-still-delayed-again/
21. Gizmochina. "Why Samsung's Ballie robot has disappeared from CES 2026." January 7, 2026. https://www.gizmochina.com/2026/01/07/samsung-ballie-robot-shelved-after-delays/
22. Samsung Newsroom. "Samsung Electronics Announces Fourth Quarter and FY 2025 Results." January 29, 2026. https://news.samsung.com/global/samsung-electronics-announces-fourth-quarter-and-fy-2025-results
23. CNBC. "Samsung profit surges over eightfold to beat estimates as AI boom fuels memory chip crunch." April 30, 2026. https://www.cnbc.com/2026/04/30/samsung-q1-earnings-ai-memory-chip-demand-profit-record.html
24. TechTimes. "Samsung Leads DRAM Market Share at 38.6%: SK hynix Trails on Revenue but Tops Profit Margins." June 9, 2026. https://www.techtimes.com/articles/318052/20260609/samsung-leads-dram-market-share-386-sk-hynix-trails-revenue-tops-profit-margins.htm
25. The Guardian / Reuters (via BusinessMirror). "Jun Young-hyun statement to investors on technical competitiveness." October 8, 2024. https://businessmirror.com.ph/2024/10/08/samsung-makes-rare-apology-after-results-sag/
26. Tom's Hardware. "OpenAI's Stargate project to consume up to 40% of global DRAM output, inks deal with Samsung and SK hynix to the tune of up to 900,000 wafers per month." October 2025. https://www.tomshardware.com/pc-components/dram/openais-stargate-project-to-consume-up-to-40-percent-of-global-dram-output-inks-deal-with-samsung-and-sk-hynix-to-the-tune-of-up-to-900-000-wafers-per-month
27. Tom's Hardware. "Samsung inks $16.5 billion Tesla AI chip deal: Elon Musk says 'The strategic importance of this is hard to overstate'." July 2025. https://www.tomshardware.com/tech-industry/semiconductors/samsung-inks-usd16-5-billion-tesla-ai-chip-deal-elon-musk-says-samsung-will-produce-new-a16-chips-the-strategic-importance-of-this-is-hard-to-overstate