Samsung, SK and Nvidia put $700bn on the US-Korea table
A wave of US-Korea AI deals announced between 24 and 25 July 2026 puts roughly $700 billion of capital behind data centres, memory and next-generation chips. The hard part is who pays, who builds, and on whose timeline.
Samsung Electronics and SK Group announced on Friday 25 July 2026 that they will join Nvidia and other US companies in deals worth a combined $700 billion tied to artificial-intelligence infrastructure, according to Nikkei Asia. The commitments, disclosed across a single 24-hour window, fold Korean memory and packaging capacity into Nvidia's data-centre build-out at a scale that pushes the country's industrial policy directly into the centre of the US-China technology contest.
The headline figure is large, the counterparties are familiar, and the timing is not accidental. What is genuinely new is the structure: Korean capital, Korean fabrication, and Korean land and power, being deployed under an arrangement whose political gravity sits in Washington. That is the line worth following.
What was actually announced
The $700 billion figure, reported by Nikkei Asia, aggregates Samsung Electronics, SK Group, Nvidia and other unnamed US companies. Within that envelope, SK Group separately unveiled a more than $500 billion AI initiative with Nvidia covering large-scale AI data centres and next-generation memory, according to LiveMint's coverage of an NVIDIA statement issued on the same day. The two announcements are clearly nested rather than additive: the SK-Nvidia deal sits inside the broader $700 billion tally rather than alongside it.
SK Group's memory affiliate, SK hynix, is the most consequential piece of the stack. It is one of two credible global suppliers of high-bandwidth memory (HBM), the chip packaging that sits next to Nvidia's accelerators and effectively determines how much AI compute can be shipped per quarter. Tying SK's memory roadmap into Nvidia's data-centre plan means committing a constrained bottleneck to a single customer's cadence.
Samsung's role is broader and less defined in the public reporting. The company is the world's largest memory maker by revenue and a major foundry, but the Nikkei summary does not specify which Samsung entities or business lines carry which portion of the $700 billion. The available source items do not specify the breakdown.
Why this is not just another capex story
The dollar number invites comparison with prior semiconductor investment waves. But the structure of these deals is different. Korean conglomerates are not simply announcing capacity expansions; they are committing capacity to a US-aligned AI supply chain whose principal customer is one company, Nvidia, and whose principal end-buyer is a small group of US hyperscalers training frontier models.
Two facts sharpen the read. First, HBM is sold into a market where Nvidia controls architectural pull: each new accelerator generation sets the spec, and the memory vendors build to it. SK hynix has been the lead supplier for Nvidia's H100 and H200 generations, and Samsung has been trying to qualify for that slot. The current round of announcements reads, in part, as the formalising of that supply relationship.
Second, Korea is doing this under conditions set by Washington. The US export-control regime on advanced semiconductors to China is the background condition that makes Korean memory strategically load-bearing for the American side. The deals are framed, in the public reporting, as commercial, but their strategic alignment is the selling point for both governments. Monexus assessment: the headline figure should be read as a political-industrial package, not a market signal in the conventional sense.
The Chey factor
The same SK Group that is committing more than $500 billion to AI infrastructure is controlled by a chairman whose personal balance sheet was, as of late July 2026, the subject of a South Korean court order. On 24 July 2026, a South Korean court ordered SK Group Chairman Chey Tae-won to pay his ex-wife Roh Soh-yeong around $644 million after a legal battle that ran nearly ten years, according to a post on X by the account @Pirat_Nation documenting the ruling. The case has been called South Korea's largest divorce settlement.
That detail is not gossip. Chey controls SK Group through a web of shareholdings, and the divorce ruling reshuffles that web. Corporate-governance questions about succession, voting blocs, and the stability of SK's strategic commitments follow directly. The available source items do not specify whether the ruling has triggered any immediate change in SK Group's capital allocation or its AI commitments, and this article has not independently established that link. The proximity in time, however, is the kind of detail that Korean and US regulators will be asked about.
The stakes, in plain terms
If the $700 billion aggregate holds in anything close to the announced form, it does three things at once. It locks Korean memory and foundry capacity into the US AI supply chain for the duration of the build-out. It gives Washington a lever over Korean industrial policy that runs through commercial contracts rather than treaties. And it pushes Chinese AI hardware further into a constrained position, because the highest-spec memory and the densest packaging are now spoken for on a multi-year horizon.
The countervailing risk runs the other way. The figure is large enough to invite scepticism: $700 billion across Samsung, SK and Nvidia is more than the combined annual capital spending of those three companies' relevant divisions in any recent year, and the public reporting does not specify the time horizon, the financing structure, or the contingent elements. A reasonable alternative read is that some portion of this is option value, not committed capex, and that the actual disbursement curve will be set by Nvidia's accelerator roadmap and by US hyperscaler orders rather than by the Korean side's stated ambition. Monexus assessment: treat the headline as a commitment ceiling, not a spend schedule, until SK and Samsung publish time-phased breakdowns.
What remains genuinely uncertain, and where the sources thin out, is the role of the Korean state. The country has historically used a mix of tax incentives, concessional financing through the Korea Development Bank, and regulatory speed to support its chaebol. None of the available source items specify which Korean public institutions, if any, are underwriting the AI commitments. That is the next filing worth watching.
Desk note: the wire version of this story (Nikkei, LiveMint, X) leans on the headline figure. Monexus pushed on structure: who is tied to whom, on whose clock, and what is left unsaid about Chey's court order and the Korean state's role.
Wire provenance
This editorial synthesis draws on the following public wire/social posts:
- https://t.me/NikkeiAsia/21067
- https://t.me/nikkeiasia/21067
- https://t.me/LiveMint/21956
- https://www.livemint.com/companies/news/nvidia-sk-group-announce-500-billion-ai-initiative-huang-says-powering-next-wave-of-growth-11784970047880.html
- https://x.com/Pirat_Nation/status/2080790379448709224