Hello, this is Daily Stock. Today, we will provide an in-depth analysis of the hot topic in the Nasdaq market: the exclusive partnership between NVIDIA (NVDA) and SpaceX, and the positive outlook for the Rubin architecture set to roll out in the second half of the year.
[Image: /stdaily/uploads/202608/gen_6a7671e41baf80.81582861.png]
Key Summary
SpaceX, which recently completed its first earnings release post-IPO, drew significant market attention by announcing an exclusive partnership with NVIDIA. SpaceX will fully adopt NVIDIA's next-generation "Vera Rubin" architecture to build its next-generation low Earth orbit (LEO) AI satellite network, "Starmind AI1."
This partnership signals the beginning of "Space Computing" to overcome power shortages in terrestrial data centers. Simultaneously, NVIDIA has secured a massive ally for its Rubin chips, scheduled for release in the second half of 2026, solidifying its monopoly in the AI semiconductor market.
Current Status Summary
According to SpaceX's Q2 2026 earnings, revenue surged 92% year-over-year to $7.81 billion, beating market expectations ($6.93 billion). Notably, the company executed a record-high Capital Expenditure (CapEx) of $18.37 billion, of which 86% ($15.83 billion) was funneled into AI infrastructure.
CEO Elon Musk praised NVIDIA's Vera Rubin NVL72 (codename Kyber) rack design during the conference call and stated that the company plans to unify all future AI data center infrastructure with NVIDIA chips. Accordingly, SpaceX plans to secure 2GW of computing capacity by the end of this year and expand it to 10GW to 20GW by the end of next year.
As of today, August 8, 2026, the Nasdaq index stands at 26,690.62 points, and the USD/KRW exchange rate closed at 1,410.50 won. According to the Daily Stock Fear & Greed Index, the Nasdaq is in the Greed (63.7) phase (up from Neutral 45.2 a week ago), while the KOSPI is at 6,258.77 points, indicating Extreme Fear (17.1).
[Image: /stdaily/uploads/202608/gen_6a7671eddf0d54.28485408.png]
Financial Analysis
NVIDIA's Vera Rubin platform is built on TSMC's cutting-edge 3nm (3N FinFET) process, featuring 336 billion transistors and next-generation HBM4 memory. It boasts extreme efficiency, reducing inference token costs by up to 10 times compared to the previous Blackwell architecture.
The "Space-1 Vera Rubin" module, designed for LEO satellites, is a specialized model that improves space-based AI computing power by up to 25 times compared to the existing H100. It offers financial advantages by drastically reducing cooling energy costs, leveraging unlimited solar energy in space and natural cooling in a vacuum environment.
| Category | Blackwell Ultra (B300) | Vera Rubin (GR100) |
|---|---|---|
| Manufacturing Process | TSMC 4NP Enhanced | TSMC 3nm (3N FinFET) |
| Transistor Count | Approx. 208 Billion | Approx. 336 Billion |
| Memory Standard | HBM3e (288GB) | HBM4 (288GB) |
| Memory Bandwidth | Approx. 8.0 TB/s | Approx. 22.1 TB/s |
| Peak Compute Performance | 15 PFLOPS (FP4 Inference) | 50 PFLOPS (FP4 Inference) |
Valuation
This partnership suggests that NVIDIA has dispelled concerns about a demand slowdown from hyperscalers and tapped into a new growth engine: the "space and defense AI" segment. SpaceX's massive pre-orders for Rubin chips are expected to act as a strong support floor for NVIDIA's 12-month Forward P/E multiple.
In particular, the premium valuation justification is strengthened as it dampens competitor AMD's chase. However, the key to maintaining this multiple will be whether SpaceX can successfully prove monetization while handling billions of dollars in annual CapEx.
Expert and Institutional Analysis
Adam Jonas, an analyst at Morgan Stanley, commented, "SpaceX's computing infrastructure investment goes beyond the realm of typical terrestrial cloud providers." He analyzed that trying to utilize infinite space energy resources for AI training and inference could be a breakthrough for the global tech ecosystem facing power supply bottlenecks.
However, Taiwanese market research firm TrendForce pointed out realistic supply chain constraints in its latest report. TrendForce warned, "The shortage of High Bandwidth Memory (HBM) will likely persist until 2027," raising the possibility that "NVIDIA might downgrade the HBM specifications of the Rubin Ultra products."
Risk Factors
Operational reliability in the extreme environment of space remains an unsolved challenge. The potential for semiconductor malfunctions due to cosmic radiation and the difficulty of maintaining orbiting satellites are persistent hardware risks.
Furthermore, supply bottlenecks due to delays in securing yield rates for HBM4 by memory manufacturers are also a concern. Even if hardware performance is proven, if supply does not back it up in time, NVIDIA's mass production schedule for the Rubin platform could face delays.
Investment Perspective Summary
The alliance between NVIDIA and SpaceX represents a historic turning point where AI infrastructure expands beyond Earth into orbit. Although the Rubin architecture possesses peerless specifications that transcend current technical limits, stable memory supply chain management is crucial for long-term implementation.
As Nasdaq's "Greed" sentiment persists, investors should carefully weigh their portfolio allocation based on the actual mass production yield of Rubin chips in the second half of the year and real launch data from the Starmind AI1 satellites.
Investor Checkpoint Q&A
Q1. Why did SpaceX decide to exclusively use NVIDIA chips?
A1. CEO Elon Musk clearly stated that it is because the design of NVIDIA's Vera Rubin platform and NVL72 Kyber rack delivers peerless performance in the industry.
Q2. Which chips will be used in the Starmind AI1 satellite project?
A2. A dedicated module combining NVIDIA's latest 'Vera CPU' and 'Rubin GPU' will be installed to perform real-time AI computing in Earth's orbit.
Q3. What is the potential specification downgrade for the Rubin chip raised by TrendForce?
A3. Since a global HBM shortage is expected until 2027, NVIDIA is reportedly considering adjusting specifications by lowering the HBM4 stacking layers of the Rubin Ultra.
Q4. What is the clear advantage of space computing over terrestrial data centers?
A4. It is free from power grid constraints and environmental regulations on Earth, and can lower costs by utilizing unlimited solar power and natural vacuum cooling in space.
Q5. What key metrics should investors watch for NVIDIA in the short term?
A5. The actual shipment volume of Rubin in H2 2026, the mass production and validation schedule of global HBM4 memories, and the actual ROI (Return on Investment) of major customers' CapEx.