Samsung SDS is seeking to become Asia’s first cloud provider to offer Nvidia’s Vera Rubin GPUs, in March 2027, The Korea Herald reported on 29 October 2026. Chief executive Lee Joon-hee set out the plan on the company’s third-quarter earnings call that day, starting with B300 GPU-as-a-service at its Dongtan data center by the end of this year.
Samsung SDS will operate 1,008 Vera Rubin units and 2,016 B300 units under South Korea’s 2.08 trillion won government GPU project, Seoul Economic Daily reported in June. That is 1,008 of the project’s 2,016 Vera Rubin units, or 50% by our arithmetic, and the same report said all 2,016 were allocated to government use.
Short explainer · Samsung SDS Vera Rubin
A vertical news Short narrated from the published headlines and opening lines of this piece and two related FourWeekMBA pieces.
Business Pill · Sovereign AI
A short explainer on sovereign AI: who controls the model and the data. It teaches the general idea only and says nothing about any company, person or government in this story.
The key insight: Samsung SDS says it will offer Nvidia Vera Rubin GPUs as a service from March 2027, which Seoul Economic Daily reports is expected to be a first in Asia; under Korea’s GPU project it operates 1,008 of the 2,016 Vera Rubin units, 50% by our arithmetic.
Lee said, as quoted by The Korea Herald: “Following our selection by the Ministry of Science and ICT to build and operate GPU infrastructure, we will begin offering B300 GPUaaS at the west wing of our Dongtan data center in Hwaseong, Gyeonggi Province, by the end of this year.”
Seoul Economic Daily reported that he added the company “plans to start a Vera Rubin-Based GPU service in March next year”, in what the paper said “is expected to be the first such rollout in Asia”.
The Korea Herald reported a separate stopgap ahead of the national AI computing center in Haenam, which it said opens in 2028. Lee said: “We are preparing to offer an interim service in April 2027 using part of the west wing of our Dongtan data center to meet surging demand for AI computing before the main center begins operations.”
South Korea’s 9,704-unit government GPU project: by operator, Samsung SDS 3,024 (1,008 Vera Rubin + 2,016 B300), Naver Cloud 4,120 and Elice Group 2,560; by use, 2,016 Vera Rubin and 4,360 B300 for government use and 3,328 B300 for the cloud firms’ own use. Our arithmetic: Samsung SDS runs 50% of the Vera Rubin units. Source: Ministry of Science and ICT allocation as reported by Seoul Economic Daily, 8 Jun 2026.
Where the Vera Rubin Units Sit
The Ministry of Science and ICT picked Naver Cloud, Samsung SDS and Elice Group for a project to secure 9,704 GPUs, 2,016 Vera Rubin and 7,688 B300, Seoul Economic Daily reported. By the ministry’s split as reported, Naver Cloud runs 1,008 Vera Rubin and 3,112 B300 units, Samsung SDS 1,008 Vera Rubin and 2,016 B300, and Elice Group 2,560 B300.
Seoul Economic Daily reported that the ministry decided to allocate “2,016 Vera Rubin and 4,360 B300 units of the 9,704 GPUs for government use”. The remaining 3,328 B300 units are for the cloud companies’ own services, the report said. By our arithmetic, every Vera Rubin unit in the project is in the government pool, and the cloud companies’ own-use share is 3,328 of 9,704 units, about 34%.
In June, Deputy Prime Minister Bae Kyung-hoon said, per Digital Today: “It looks like we will receive B300 on time, but Vera Rubin seems likely to be a bit delayed.” The ministry has said Vera Rubin services will start in the first half of 2027.
Our Arithmetic: The Scale and the Earnings
Samsung SDS’s 3,024 units are about 31% of the project’s 9,704, and Vera Rubin is about 21% of the project by unit count, by our arithmetic. For scale, Google Cloud plans to scale its Vera Rubin infrastructure to as many as 960,000 Rubin GPUs, as recorded in our Google Rubin piece of 21 October; that is roughly 950 times Samsung SDS’s 1,008 units, though the two counts may not use the same unit.
In the third quarter, Samsung SDS revenue rose 11.2% to 3.7725 trillion won and operating profit rose 3.7% to 241 billion won, Seoul Economic Daily reported from the company’s filing. Cloud revenue rose 18% to 796.1 billion won, which by our arithmetic is about 47% of IT services revenue of 1.681 trillion won and about 21% of group revenue.
The company is targeting more than 1 trillion won in operating profit next year, The Korea Herald reported. Four quarters at the third quarter’s 241 billion won would be about 964 billion won, by our arithmetic, so the target sits a little above the current quarterly pace.
How the Lever Works
As we read the reports, the Vera Rubin units in Korea’s GPU project are placed through a state allocation: the ministry chose the operators, set the unit counts and reserved the Vera Rubin units for government use. The reports we read do not say whether Samsung SDS’s March service runs on those units or on separately bought chips.
Naver, the other Vera Rubin operator in the project, also plans a 200-megawatt AI factory “featuring NVIDIA Vera Rubin and Blackwell platforms”, with a tentative 55MW in the first half of 2027, as set out in our Naver AI factory piece. Samsung SDS aims to expand its AI infrastructure capacity to more than 800 megawatts by 2031, The Korea Herald reported.
As we read the June allocation, Korea’s first tranche of Vera Rubin is a state allocation: the ministry chose the three operators, set each one’s unit count and put all 2,016 Vera Rubin units in the government-use pool.
As we read the call reports, Samsung SDS is sequencing B300 at Dongtan by the end of 2026, an interim service in April 2027 and Vera Rubin from March 2027, while the national AI computing center in Haenam is due in 2028.
As we read the quarter, cloudrevenue, which Seoul Economic Daily said was driven by Samsung Cloud Platform demand and GPUaaS supply, is about 47% of Samsung SDS’s IT services revenue and about 21% of group revenue by our arithmetic; the reports give no separate GPUaaS revenue figure.
Lee Joon-hee, Samsung SDS CEO, Q3 earnings call, as quoted by The Korea Herald, 29 October 2026
“We are preparing to offer an interim service in April 2027 using part of the west wing of our Dongtan data center to meet surging demand for AI computing before the main center begins operations.”
Three Implications
FOR KOREAN AI DEVELOPERS The ministry has said the government-use units will go to independent foundation models, national AI projects and industry-academia-research model development, per Seoul Economic Daily in June.
FOR SAMSUNG SDS’S TARGETS The Korea Herald reported a target of more than 1 trillion won of operating profit next year and more than 800 megawatts of AI infrastructure by 2031; four quarters at the third quarter’s 241 billion won would be about 964 billion won, by our arithmetic.
FOR NVIDIA’S ASIAN SUPPLY Deputy Prime Minister Bae Kyung-hoon said in June that Vera Rubin “seems likely to be a bit delayed” and that Korea would ask for supply as a top priority, Digital Today reported.
The framework argues that AI’s physical constraints are creating “geopolitical leverage that operates independently of who has the best models.”
As we read it, a government deciding which operators get the first Vera Rubin units, and reserving them for national use, is one way that leverage shows up. The framework was published in December 2025, is about physical supply chains, and does not discuss Samsung SDS or Korea’s GPU project.
What Is Not Established
The March 2027 date and the expectation that it will be the first such rollout in Asia come from the earnings call as reported by The Korea Herald and Seoul Economic Daily; Seoul Economic Daily’s text is AI-translated from Korean. We did not read a Samsung SDS release or call transcript, and Nvidia has not confirmed the timing in anything we read.
Pricing for the Vera Rubin service was not reported. The 50%, 34%, 31%, 21%, 47% and 964 billion won figures and the comparison with Google Cloud are our arithmetic; the four-quarter figure is not a forecast.
This Week In The AI Supercycle
Execution Fault Lines, Concentrated Demand, and Geopolitical Firewalls
This week’s issue: the AI build is not short of demand for its parts. It is short of cheap money, and its end revenue rests on a narrow base of large, often venture-funded buyers.
Samsung SDS says it will offer Vera Rubin GPUs as a service from March 2027, which Seoul Economic Daily reports is expected to be a first in Asia. Under Korea’s GPU project it operates 1,008 of 2,016 Vera Rubin units, all allocated to government use: 50%, by our arithmetic.
A note on sourcing. The March 2027 plan, the Asia-first expectation and the quotes come from The Korea Herald’s and Seoul Economic Daily’s reports of the 29 October earnings call; Seoul Economic Daily’s text is AI-translated from Korean. The GPU allocation comes from the Ministry of Science and ICT as reported in June. Shares and multiples are our arithmetic; linked earlier pieces carry their own sources.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
