Nvidia's Billion-Dollar Balancing Act: Mega Projects, New Chip Wars, and Insider Caution

Nvidia partners with SK Hynix for HBM4, launches Vera CPU, plans Asian AI factories, and a massive Ohio data center — while insiders sell shares.

Jensen Huang is playing a game of extremes. While Nvidia forges alliances across Asia and plots a $500 billion data center in Ohio, the stock has slipped more than 7% in a week, insiders have unloaded shares, and major hedge funds are reducing exposure. The contrast between corporate ambition and market skepticism has rarely been starker.

Locking Down the Memory Supply Chain

The GPU giant is moving aggressively to secure its most critical bottleneck: high-bandwidth memory. CEO Jensen Huang describes the shortage of HBM chips not as a cyclical squeeze but as a structural constraint driven by insatiable AI demand. To solve it, Nvidia is deepening its ties with South Korea's SK Hynix.

Under the expanded partnership, SK Hynix will supply up to 70% of Nvidia's HBM4 requirements. The two companies are co-developing the next-generation memory standard, and South Korean factories are being reconfigured into exclusive Nvidia production hubs. This vertical lock-up is designed to prevent the supply chain fractures that plagued earlier AI build-outs.

A New Front: The Vera CPU

Nvidia isn't just defending its GPU fortress — it's launching an assault on Intel and AMD's core turf. The new "Vera" data-center processor is a fully independent CPU with 88 cores. According to reports, it runs 1.8 times faster than existing x86 systems, making it a direct challenger in the server room.

Mass production of Vera is slated for the second half of 2026. SK Hynix will supply the specialized memory for this platform, further integrating the two companies' roadmaps. The message is clear: Nvidia wants to own the entire data center stack, not just the GPU accelerators.

Asian Alliance: Seoul as a Hub

Alongside the memory deal, Nvidia has struck a series of partnerships in Seoul. Working with SK Telecom, Naver, and Doosan, the company is building a chain of AI factories. The first facility, a 55-megawatt plant, is scheduled to come online in 2027, and capacity will expand to 200 megawatts a year later.

Naver expects this infrastructure to generate roughly 20 trillion won in revenue over the next five years. Nvidia is also collaborating with LG on robotics, while Doosan provides the construction technology for the facilities. The Korean peninsula is emerging as a key production base for Nvidia's AI hardware.

Ohio's $500 Billion Bet

Half a world away, Nvidia is backing an even larger project. Together with OpenAI and SoftBank, the company is planning a massive data-center campus in Ohio. The estimated cost: at least $500 billion, with a power draw of ten gigawatts.

Nvidia is reportedly guaranteeing the lease agreements for OpenAI, which will rent the servers for twenty years. The first construction phase is expected to begin in 2028. If realized, it would be one of the largest infrastructure projects in the tech industry, dwarfing current AI data centers by orders of magnitude.

Insiders and Institutions Take Profits

While these grand plans unfold, some of the smart money is heading for the exits. Nvidia director Mark Stevens sold one million shares in early June, netting approximately $221 million. The sale came just before the stock's recent pullback from its May all-time high.

Hedge funds are also paring back. David Tepper reduced his Nvidia position by 13% in the first quarter. Dan Loeb's Third Point sold nearly its entire stake. The combined insider and institutional selling suggests that some investors see the current valuation as rich, even with the growth story intact.

Market Reaction and Technical Levels

The selloff has pushed Nvidia's stock to €174.20, well below the record high set in May. On a weekly basis, the shares are down around 7%. The decline has also broken the stock below its 50-day moving average. However, the longer-term trend remains intact: the price still holds a comfortable cushion above the 200-day line at €161.85. The Relative Strength Index sits at 42, inching toward oversold territory.

Analysts remain broadly bullish. Loop Capital has an outperform rating and a price target of $350, arguing that GPU shipments will double and selling prices will rise in coming months. Huang himself has dismissed the dip as a buying opportunity, insisting that global AI infrastructure build-out is still in its infancy.

Nvidia is betting big on three fronts — memory security, CPU expansion, and data-center scale — all at once. Whether the stock can keep pace with the vision is another question entirely.