AI Daily Brief: Seoul's $880 Billion Chip Bet, TSMC Opens the Books on a Record Quarter
Seoul puts $880 billion behind Samsung and SK Hynix, TSMC opens its books on the AI boom's record quarter, Nvidia and ServiceNow ship a desktop agent that never forgets, and Gemini 3.5 Pro arrives tomorrow. Plus IBM's warning shot at Indian IT, and the day in AI money.
The theme of the day is commitment, the kind measured in decades and hundreds of billions. A national government just bet the equivalent of half its GDP on silicon, the world's most important foundry opens its books this afternoon, and the model everyone has waited six weeks for lands tomorrow.
Seoul goes all in with Samsung and SK Hynix
South Korean President Lee Jae-myung unveiled a decade-long AI infrastructure plan worth 1,350 trillion won, roughly $880 billion, one of the largest national AI commitments ever made. The breakdown reads like a country rebuilding itself around the technology: about $518 billion for memory chip manufacturing through Samsung and SK Hynix, roughly $550 billion for AI data centers led by SK Group, GS Group and Naver, 8.4 gigawatts of data center capacity targeted by 2029, and a plan to grow Korea's humanoid robotics market share from 1 percent to 20 percent by 2028.
Korea already sits at the center of the AI hardware economy, with SK Hynix holding around 60 percent of high-bandwidth memory and Samsung's foundry hungry for a marquee AI customer. This plan is designed to defend that position against Taiwan, the US and China, and it lands the same week frontier labs came shopping: Anthropic's chip talks with Samsung, covered in yesterday's brief, suddenly look less like a one-off and more like the first customer through a door Seoul intends to hold wide open. Industrial policy is back, and AI is its arena.
TSMC opens the books on the boom's record quarter
TSMC holds its second-quarter earnings conference at 2pm Taiwan time today, and the preliminaries already set a high bar: record revenue of NT$1.27 trillion, about $39.6 billion, up 36 percent year over year, with June revenue alone up 68 percent and AI chips now contributing 61 percent of total sales. The company guided gross margin to 65.5 to 67.5 percent and pushed its capital budget toward the high end of $52 to 56 billion.
The number to listen for isn't revenue, it's CoWoS. Advanced packaging has replaced wafer starts as the binding constraint on AI chip production, so TSMC's update on packaging capacity will say more about how many accelerators actually ship in 2027 than any hyperscaler press release. If management raises full-year guidance above the current "more than 30 percent" growth outlook, the market will read it as confirmation that the buildout has not found its ceiling yet.
Nvidia and ServiceNow ship an agent that never forgets
Nvidia and ServiceNow expanded their partnership with Project Arc, a long-running desktop agent for knowledge workers, running on Nvidia's OpenShell secure runtime and its open Nemotron models. The pitch is a different shape of AI product: instead of a stateless chatbot you prompt and forget, Arc persists on the desktop, learns a specific worker's routines over days and weeks, and improves as it goes.
This is also Nvidia moving up the stack. The chip giant supplying the silicon now wants to own the runtime and the models that enterprises run on it, with ServiceNow contributing the thing Nvidia lacks: a grip on the workflow backbone of thousands of large companies. The same week Meta rolled out its Business Agent Platform globally and started selling excess compute like a cloud provider, the enterprise agent land grab gained two heavyweight entrants in as many days. The reliability bar for an always-on agent is far higher than for a demo, and that is exactly the test that will separate this generation of products.
Tomorrow: Gemini 3.5 Pro walks onto the biggest stage of the year
Google's rebuilt Gemini 3.5 Pro is expected to reach general availability tomorrow, though Google has confirmed neither the date nor the leaked specs, which include a 2-million-token context window and a Deep Think reasoning tier. The model arrives six weeks late, a week behind GPT-5.6, and after a bruising stretch of talent-drain headlines, so it needs to win at least one headline benchmark to change the narrative.
It shares the date with Shanghai's World AI Conference, which opens tomorrow with Xi Jinping attending in person for the first time since the event began in 2018. The West's most anticipated model of the summer and the East's most powerful leader on the same calendar square is 2026 in miniature: the AI race is now a contest of national systems, not just labs, and Seoul's $880 billion announcement this week makes it a three-way conversation.
What it means for India
The uncomfortable message for Indian IT this week came from an unexpected direction. IBM forecast quarterly revenue below Wall Street estimates, and the Nifty IT index fell 0.7 percent on the read-through: enterprise customers are diverting technology budgets toward AI infrastructure, the servers and networking gear TSMC and Korea are racing to supply, at the expense of software and IT services. That is the global capex story showing up as a direct tax on the sector that employs five million Indians.
The counter-argument is playing out in the same earnings season. TCS put a $2.6 billion number on its AI revenue last week, HCLTech opened over 2 percent higher this morning as IT stocks rebounded, and the LTIMindtree tie-up with Anthropic shows the services majors converting the AI budget shift into deal flow rather than just losing to it. India is not in the $880 billion sovereign-silicon game Seoul just joined, but the deployment layer, wiring AI into enterprises that IBM says are hungry to spend on it, is precisely where Indian IT lives. The question this results season answers is who is converting, and how fast.
Markets and AI money
Wednesday's US session rotated rather than rallied. Cooling inflation data kept the bid alive, but money moved out of chip stocks and into the megacaps: Apple jumped 4 percent to a record high, Amazon, Alphabet and Microsoft all gained around 3 percent, and fresh US strikes on Iran were largely shrugged off. Mumbai told a similar story of resilience, with the Sensex surrendering a 590-point morning surge to close modestly higher as financials offset the IT drag.
| Index / stock | Jul 15 close | Move |
|---|---|---|
| S&P 500 | 7,572.40 | +0.38% |
| Nasdaq Composite | 26,269.23 | +0.62% |
| Dow Jones | 52,658.64 | +0.29% |
| Sensex | 77,185.43 | +0.17% |
| Nifty 50 | 24,078.50 | +0.11% |
| Nifty IT | — | −0.7% |
TSMC's numbers land this afternoon, Gemini and Shanghai take the stage tomorrow. The capex is committed; now the models have to earn it.
Keep reading
AI Daily Brief: Google's New Flash Models Land as AMD and Anthropic Circle a Chip Deal
Google quietly shipped three new Gemini models overnight, Anthropic locked up 3.5 gigawatts of TPU capacity, and AMD kicked off its biggest AI event of the year with Anthropic rumors swirling. Below, what the compute land grab means for India, and how the chip trade is pricing it all in.
AI Daily Brief: Nvidia Ships Its Groq Chip and a New Model of Its Own, Qualcomm Bets $14 Billion Against It
Nvidia's first Groq-derived inference chip heads into production on Samsung's line just as Samsung's chairman prepares to meet Jensen Huang, Qualcomm commits $14 billion to not need Nvidia at all, and Nvidia's own research lab quietly shipped a new open model that decodes six times faster than the competition. Plus what the chip realignment means for India, and the day's market moves.