The $600 Trillion Stargate Project: What's Behind the Hype?
You've probably heard about the Stargate Project that's been shaking up the AI industry lately. This joint venture between SoftBank and the U.S. government, featuring global AI giants like OpenAI, Oracle, and NVIDIA, is literally the world's largest AI infrastructure development project.
The scale of the data center being built in Texas is mind-boggling. We're talking about a total investment of $500 billion – that's roughly $600-700 trillion in Korean won. This massive data center will house NVIDIA's Blackwell systems with 50,000 GPUs per cluster across 8 clusters, totaling 400,000 GPUs. The power consumption alone will reach 1.2 gigawatts – equivalent to what a mid-sized city uses.
Land acquisition is complete, basic construction is underway, and they're targeting completion by mid-2026. This project isn't just about building a big data center; it's presenting a new infrastructure paradigm for the AI era.
Shifting Dynamics in the Traditional Cloud Market
Until now, the cloud market has been dominated by the "hyperscaler" trio – Amazon Web Services (AWS), Microsoft Azure, and Google Cloud – controlling over 80% of the market. But with the explosive growth of the AI market, new players are emerging.
Enter companies like CoreWeave and Oracle. CoreWeave's stock price has soared from $40 at IPO to $165 currently, while Oracle has nearly doubled recently. What's interesting is their market share. CoreWeave holds just 2% and Oracle 5%, but when you look at 2025 NVIDIA GPU allocations, CoreWeave gets 11% and Oracle gets 19%.
What does this mean? It signals the formation of a new cloud ecosystem specialized for AI workloads.
NVIDIA's Strategic Partnerships
NVIDIA's move to break free from the influence of traditional hyperscalers is also noteworthy. As Big Tech companies develop their own AI chips to reduce NVIDIA dependency, NVIDIA has started seeking new partners.
The relationship with CoreWeave goes beyond a simple customer-supplier dynamic. NVIDIA has invested trillions of won in CoreWeave and supplies them with next-generation GPUs like the Blackwell GB300 first. This is essentially a strategic partnership.
While Microsoft has somewhat slowed down its custom chip development, Google's TPU and Amazon's Trainium are actually entering mass production and showing results. In this context, NVIDIA building new alliances is a natural progression.
CoreWeave's Differentiated Competitive Edge
The reason CoreWeave is getting attention is clear: their overwhelming advantage in GPU utilization rates. While typical cloud providers achieve 35-50% GPU utilization, CoreWeave pushes it to 75% – nearly double the efficiency.
The secret behind this high efficiency lies in their "bare metal" architecture. While traditional hyperscalers build virtualization layers through hypervisors to offer various services, CoreWeave focuses solely on AI. By providing infrastructure specialized for AI workloads instead of general-purpose cloud services, they can maximize hardware resource utilization.
This efficiency translates directly into cost competitiveness. Using H100 GPUs on Amazon costs $180 per hour, while CoreWeave and Oracle are 30-40% cheaper. Same performance at a lower price – naturally, customers flock to them.
Many companies are actually adopting a hybrid strategy, keeping existing workloads on AWS while moving only AI-related tasks to CoreWeave. SemiAnalysis, the most authoritative research firm in semiconductors, recently gave CoreWeave the only platinum rating in their GPU cloud evaluation.
Oracle's Multi-Cloud Strategy
Oracle's approach differs from CoreWeave's. If we categorize AI infrastructure into five types, fully cloud-based and API provision are CoreWeave's territory, while hybrid, dedicated cloud, and fully on-premises are Oracle's strengths.
Oracle's core strategy is "multi-cloud." It's risky for companies to depend on just one cloud provider. What happens when Google Cloud goes down? That's where backup systems come in, and Oracle performs this role perfectly.
Oracle's biggest advantage is the ability to backup data from Google, Amazon, or Azure directly. This guarantees data sovereignty for enterprises. Especially in on-premises environments, security concerns virtually disappear.
This characteristic has brought attention to the concept of "Sovereign AI." For national-level AI, government systems, and defense sectors, uploading data to Google or Azure feels uncomfortable. Oracle's cloud infrastructure precisely targets these needs.
Oracle's participation in the Stargate Project also fits this context. Security-critical companies like OpenAI and Palantir have no choice but to select Oracle.
Investment Perspective Analysis
CoreWeave's growth trajectory is remarkable. Their 2024 revenue growth rate hit 737%, with over 150% growth expected in 2025. However, from an investment standpoint, there are areas requiring caution.
The current stock price significantly exceeds the analyst average target of $76, creating valuation pressure. More importantly, CoreWeave is a company that raised funds using GPUs as collateral. They have high debt ratios and face risks from significant depreciation of GPU assets due to technological advancement.
Nearly 100% dependency on NVIDIA is also a concern. However, the business model itself is excellent. Their strategy of accurately identifying and entering market gaps deserves high praise.
The Future of AI Infrastructure Market
With the explosive growth of the AI market, the infrastructure market is also diversifying. Cracks are appearing in the monopolistic system of traditional hyperscalers, while new players offering specialized services are emerging.
CoreWeave has established positioning as a high-efficiency cloud specialized for AI workloads, while Oracle has built differentiation as infrastructure guaranteeing multi-cloud and data sovereignty. Their success isn't just from technological superiority, but from accurately reading market changes and understanding customer needs.
As AI technology advances further and usage expands, market segmentation will likely accelerate. We'll probably see a structure where specialized services compete in respective areas: general-purpose cloud, AI-specialized cloud, security-focused cloud, and more.
In Conclusion
The AI infrastructure market has moved beyond simply providing computing power into an era of specialized solution competition. Large-scale investments like the Stargate Project, along with the rise of new players like CoreWeave and Oracle, demonstrate how dynamically this market is changing.
Existing Big Tech companies won't sit idle either. However, considering the rapid growth of the AI market and diverse customer needs, companies providing specialized services will have ample opportunities. The key is differentiated value propositions based on technological advantages.
We need to continue watching these changes and new opportunities in the AI infrastructure market. The true beneficiaries of the AI revolution might not simply be companies developing AI models, but those providing the infrastructure to operate them efficiently.