South Korea $950B AI Deal: Nvidia, SK Hynix & Europe

Samantha: Welcome to The State of Tech, The European Edition, Monday July twenty-seventh, 2026. I'm Samantha Lawrence.

Bob: And I'm Bob Russell. Today: South Korea unveils a nine hundred fifty billion dollar AI push anchored by a giant Nvidia and SK Hynix deal, Qualcomm hikes smartphone chip prices by double digits, China's Moonshot open-sources a model that beats the biggest names, Big Tech braces for a nervy earnings week focused on AI spending, and to close out, two things you can actually use today: the writing app Bibisco, and the free European map service Qwant Maps. Let's get into it.

South Korea locks in nine hundred fifty billion dollars of AI investment, with a five hundred billion dollar Nvidia and SK Hynix deal at its heart.

Samantha: We start in San Francisco, where South Korea has just made itself impossible to ignore in the global AI race. President Lee Jae Myung announced a package of AI initiatives worth nine hundred fifty billion dollars, and the centrepiece is a five hundred billion dollar partnership between Nvidia and SK Hynix. The idea is to lock in supply of next-generation high-bandwidth memory, the specialised chips that feed data into AI accelerators fast enough to keep them busy. Without that memory, even the most expensive AI hardware sits idle.

Bob: And the scale here is genuinely eye-watering. As part of the deal, SK Telecom plans to build a two-gigawatt data centre by 2027, powered by Nvidia's next-generation Vera Rubin chips and SK Hynix's newest memory. Two gigawatts is roughly the output of two large nuclear reactors, dedicated to a single AI campus. That tells you where the industry thinks the bottleneck is: not the software, not even the chips themselves, but the raw electricity and cooling needed to run them, plus enough memory to keep them fed.

Samantha: What makes this strategically interesting is that South Korea is positioning itself as the memory chokepoint of the AI economy. SK Hynix and Samsung already dominate high-bandwidth memory production, and Seoul is now formalising that leverage through direct government-backed partnerships with US chip designers. For Europe, this matters in two ways. First, any European company building AI infrastructure will now be competing against a Korean-American alliance that has priority access to the tightest part of the supply chain. And second, it further sidelines Europe's own ambitions, which so far amount to a handful of chip subsidies and no comparable partnership.

Bob: The other angle worth flagging is what this means for prices. High-bandwidth memory has been in acute shortage for over a year, and analysts have warned that supply is essentially spoken for through 2027. Locking in Nvidia and SK Hynix at this scale suggests the biggest buyers are willing to pay almost any price to secure allocation. European cloud providers, research institutes, and even car makers building in-house AI, they are all going to feel that squeeze. Expect longer lead times, higher prices, and more pressure on Brussels to accelerate its own chip strategy beyond the current, frankly modest, commitments.

Qualcomm tells customers to expect double-digit price hikes on smartphone chips from September first.

Samantha: Next up, a story that will quietly touch almost every phone sold in Europe. Qualcomm, the biggest supplier of smartphone processors outside of Apple's own silicon, told its customers on Friday that prices will rise by a double-digit percentage for products shipped after September first. The company says it has run out of room to absorb rising costs from its own suppliers, and attempts to find cheaper alternative components have not worked out.

Bob: And this is not a small technical footnote. Qualcomm's chips sit inside a huge share of Android phones sold in Europe, from mid-range Samsung devices to Xiaomi, Oppo, Motorola, and many others. When Qualcomm raises prices by ten percent or more, phone makers have three options: absorb the cost and take a hit on margins, pass it on to consumers, or downgrade the chip in the next model to keep the price stable. In practice, all three will happen, and consumers usually end up paying more for the same tier of phone.

Samantha: The timing is also notable. September is when the industry finalises component orders for the crucial autumn and Christmas launches. So the phones announced at big trade shows early next year, the ones European carriers will heavily promote, are being priced right now with these higher chip costs baked in. Some analysts expect flagship Android phones in Europe to climb by fifty to a hundred euros, and mid-range models by twenty to forty. That widens the gap with refurbished and second-hand phones, a market that is already growing fast in Germany, France, and the Netherlands.

Bob: There is also a wider signal here about the state of the chip supply chain. Qualcomm citing exhausted capacity and failed attempts to find alternatives suggests the whole ecosystem around advanced chips, packaging, testing, specialised materials, is still under enormous strain. That strain is being amplified by the AI build-out we just discussed. When TSMC and its suppliers prioritise AI accelerators, everything else, including phone chips, gets squeezed. European regulators watching smartphone affordability, and consumer groups pushing right-to-repair, will want to keep a close eye on this.

Samantha: Quick interruption. If you listen to The State of Tech regularly, hit that like button and subscribe, that way you'll never miss an episode. Okay, moving on.

Chinese startup Moonshot open-sources Kimi K3, outperforming top proprietary models from Anthropic and OpenAI.

Bob: Chinese startup Moonshot AI has released Kimi K3 as fully open-source today, meaning anyone can download it, inspect it, and modify it. According to benchmark tests published alongside the release, Kimi K3 outperforms leading proprietary models from Anthropic and OpenAI, especially on programming tasks. It also ranks first on the LLM Arena leaderboard for building web interfaces, a category where developers vote on which model produces the best working code. With two point eight trillion parameters, it is now the largest open-source model in the world.

Samantha: The strategic context matters here. The United States has spent the last two years tightening chip export rules to slow Chinese AI development. The theory was that without access to Nvidia's top accelerators, Chinese labs would fall behind. Instead, groups like Moonshot, DeepSeek, and Alibaba have doubled down on efficiency, and on giving their models away for free. The message is clear: if we cannot out-spend the Americans on hardware, we will out-share them on software. And it is working, developers everywhere are downloading these models by the millions.

Bob: For European companies and researchers, this is genuinely useful. An open model of this quality means you can run serious AI work on your own servers, inside European data centres, without paying per-token fees to an American provider or sending sensitive data across the Atlantic. Universities, hospitals, government agencies, and mid-sized companies that could never afford to train something like this from scratch now get a top-tier starting point for free. That fits well with the EU's stated preference for open, auditable AI systems.

Samantha: There is, of course, a debate about whether releasing such powerful models openly is safe. Critics argue that open weights make it easier for bad actors to remove safety guardrails. Supporters counter that transparency lets independent researchers actually verify what a model does, rather than trusting a company's press release. That debate is very much live in Brussels, and it connects directly to a letter we will touch on in a moment: two dozen major tech firms, including Nvidia, Microsoft and Meta, have urged US lawmakers not to restrict open-source AI, arguing it drives both innovation and security.

Big Tech earnings week puts AI spending under a microscope as investors question when it will pay off.

Bob: This week Meta, Microsoft, Apple, and Amazon all report quarterly earnings, and the question dominating every analyst call will be the same: when does the enormous AI infrastructure spending start to pay off? Alphabet's recent results already showed capital expenditure dragging on free cash flow, and investors are getting twitchy. Combined AI-related spending across the top American tech firms is now running at hundreds of billions of dollars a year, and the revenue attached to it, while growing, is still a fraction of that.

Samantha: The tension is straightforward. On one side, executives argue this is a once-in-a-generation land grab: whoever builds the biggest AI infrastructure wins the next decade. On the other side, shareholders are starting to ask uncomfortable questions about payback periods, depreciation schedules on hardware that becomes obsolete in three years, and whether the current gross margins on AI services can ever justify the spend. When Alphabet posted its results, the stock reaction was volatile precisely because those numbers cut both ways.

Bob: For Europe, this earnings week is a health check on the whole AI economy. European pension funds, insurers, and sovereign wealth funds hold significant stakes in these American giants, so the results directly affect European savers. And European companies that have committed to major cloud contracts with Microsoft Azure, AWS or Google Cloud will be watching to see whether prices go up, or whether hyperscalers absorb the cost to keep customers locked in.

Samantha: There is also a knock-on effect for European AI startups. If Wall Street decides the AI capex cycle has peaked, funding for smaller AI companies, including those in Paris, Amsterdam, Berlin and Stockholm, will tighten quickly. On the other hand, if these earnings reassure the market, expect another wave of AI investment to flow into European ventures. Either way, by Friday evening, the tone for the rest of the year in tech will largely be set. It is not overstating things to call this the most consequential earnings week of 2026 so far.

Bibisco is the free, open-source writing app that helps novelists plan characters, plots and settings without any subscription.

Bob: To close out, two things you can actually use today. First up, for anyone who has ever thought about writing a book, or is stuck halfway through one: Bibisco. It is a free, open-source writing app built specifically for novelists and long-form storytellers. Unlike a generic word processor, Bibisco is organised around the things fiction writers actually struggle with: characters, chapters, timelines, locations, and how they connect.

Samantha: What makes it stand out is the character workshop. You can build detailed profiles for every character, their background, personality, appearance, relationships, and Bibisco then tracks where each one appears across your chapters. If your protagonist has not shown up in fifty pages, the app will tell you. It also has a mind-map view that lets you see how storylines interconnect, which is genuinely useful when you are juggling multiple plot threads and losing track of who knows what.

Bob: There is a free community edition that runs on Windows, macOS and Linux, and it stores everything locally on your own machine, so your unfinished novel does not sit on someone else's server. Users often praise how calming the interface feels compared to word processors bloated with formatting tools. There is a paid version with cloud sync if you want to work across devices, but the free version is more than enough to actually write a book. And it is translated into a dozen languages, including German, French, Spanish, Italian and Dutch.

Samantha: The barrier to entry is basically zero. You download it, open it, and it walks you through setting up a project with prompts for premise, setting, and main characters. Reviewers describe it as the tool they wish they had known about years ago. For anyone in Europe eyeing NaNoWriMo in November, or just wanting to finally get that story out of their head this autumn, Bibisco is worth an afternoon.

Qwant Maps is the free European map service that offers turn-by-turn directions without tracking your journeys.

Bob: And second, a genuinely European alternative to Google Maps that has been quietly maturing: Qwant Maps. It is a free web-based map service from the French search company Qwant, and it offers the basics most people actually need: address search, points of interest, walking, cycling, driving and public transport directions across European cities.

Samantha: The pitch is straightforward. Qwant Maps does not build a profile of where you go, what you search for, or which restaurants you looked at last Tuesday. It runs on European servers, under European data protection rules. The map data itself comes from OpenStreetMap, the open community project, which means coverage in Europe is often surprisingly detailed, especially for hiking paths, cycle routes, and public transport stops that Google sometimes misses.

Bob: It works directly in any web browser, on desktop or phone, no app install required, no account needed. You just open the site and search. For occasional trips, holidays, or planning a route across a European city you do not know well, it is a perfectly capable tool. Reviewers point out that it is not as polished as Google Maps for things like live traffic or business hours, but for the majority of everyday map lookups, it does the job while keeping your location history out of the advertising economy.

Samantha: A nice small habit to try this week: next time you look up an address or plan a route, open Qwant Maps instead of your default. If it works for what you need, keep using it. If not, you have lost nothing. It is one of those quiet European alternatives that only survives if people actually try it.

Samantha: Today we covered: South Korea's nine hundred fifty billion dollar AI push with Nvidia and SK Hynix, Qualcomm's double-digit smartphone chip price hikes, Moonshot's open-source Kimi K3 beating the top proprietary models, the make-or-break Big Tech earnings week focused on AI spending, and two things to try yourself: the novel-writing app Bibisco, and the European map service Qwant Maps.

Bob: Want to know more or react? Visit stateoftech.eu or email us at info@doorzetters.net.

Bob: State of Tech, the tech world in 15 minutes.