Elon Musk Updates Colossus 2, Boosting Nvidia GPU Demand

Elon Musk Updates Colossus 2, Boosting Nvidia GPU Demand

THE MONEYWISE JOURNAL  ·  finance  ·  Monday, September 28, 2026
Elon Musk Updates Colossus 2, Boosting Nvidia GPU Demand

Key Takeaways Colossus 2 currently hosts 110,000 GB200 and 440,000 GB300 Nvidia GPUs.

By MoneyWise Guide

Elon Musk posted on X that SpaceX’s Colossus 2 AI cluster will double its Nvidia GPU count by year‑end, adding 440,000 GB300 chips and 110,000 GB200 chips, with further additions scheduled for next week and November.

Key Takeaways

  • Colossus 2 currently hosts 110,000 GB200 and 440,000 GB300 Nvidia GPUs.
  • SpaceX plans to add 220,000 GB300 GPUs next week and another 220,000 in November.
  • SpaceX’s long‑term goal is 1 million GPUs by 2026, potentially exceeding the target if orders are fulfilled.
  • xAI spent roughly $18 billion on its initial 550,000 GPUs, a mix of older and newer models.
  • Nvidia’s Vera Rubin platform and new Vera CPUs are being integrated into SpaceX’s AI and orbital data center plans.
Metric Value Why it matters
Current Colossus 2 GPU count 550,000 (110,000 GB200 + 440,000 GB300) Baseline for future expansion
Upcoming GPU additions 440,000 (220,000 next week + 220,000 November) Immediate increase in Nvidia demand
SpaceX GPU target 2026 1 million GPUs Potential to surpass current orders
xAI initial GPU spend $18 billion for 550,000 GPUs Significant capital commitment to Nvidia chips

Colossus 2 Expansion Details

SpaceX’s Colossus 2 supercomputer currently houses 110,000 Nvidia GB200 chips and 440,000 GB300 chips. Musk announced that 220,000 additional GB300 GPUs will be online next week, with another 220,000 slated for November. He added a hopeful note that a further 220,000 GB300s might arrive in late December, “if we get lucky.”

SpaceX’s GPU Ambitions

SpaceX’s original plan was to scale to 1 million GPUs by the end of 2026. The recent additions suggest the company may surpass this target if the pending orders are completed. The expansion aligns with SpaceX’s broader compute goals: ending the year with roughly 2 gigawatts of computing power and targeting about 10 gigawatts by 2027, up from 5 gigawatts. One gigawatt of compute power costs between $50 billion and $60 billion, with Nvidia chips and related components accounting for about $35 billion of that total.

Nvidia–SpaceX Partnership

Musk previously confirmed that SpaceX would exclusively use Nvidia chips. He praised Nvidia’s Vera Rubin platform as the best architecture available and highlighted the partnership’s value. The collaboration extends beyond training clusters to the development of GPUs for SpaceX’s planned orbital data centers. Cosmic radiation can flip bits in standard chips, creating hard‑to‑detect errors. To mitigate this, Musk said Nvidia and SpaceX have designed a Vera Rubin NVL72 system optimized for space, with a launch planned for the end of 2027 and scaling to begin in 2028.

Nvidia’s Market Position

Nvidia remains the dominant player in AI infrastructure, with its GPUs the primary chips for AI model training. The company’s CUDA software platform provides a wide moat in this space. Nvidia’s recent acquisition of Groq and its language processing unit (LPU) technology has been integrated into the CUDA ecosystem, giving the firm a differentiated edge in the fast‑growing inference market. The company’s ARM‑based Vera CPUs are also gaining traction, especially for agentic AI workloads, with SpaceX slated to use them for such tasks. Nvidia’s expansion into software layers, through acquisitions of SchedMD and the pending Hugging Face deal, further solidifies its position as a comprehensive AI infrastructure provider.

Investment Outlook

With Nvidia’s forward price‑to‑earnings ratio at just 14 times fiscal 2028 (ending January 2028), analysts view the stock as a buy. The company’s growing relationships with high‑profile customers like SpaceX, combined with its broad product portfolio and software ecosystem, support a positive outlook.

Why it matters

SpaceX’s commitment to Nvidia chips signals a significant boost in demand for the company’s GPUs, especially as the firm pushes toward its 1 million‑GPU goal. The partnership also underscores Nvidia’s expanding role beyond GPUs into complete AI infrastructure, including CPUs and software solutions. For investors, the alignment with a high‑growth customer and the attractive valuation suggest a compelling opportunity.

Frequently Asked Questions

What is Colossus 2?

Colossus 2 is SpaceX’s artificial intelligence computing cluster that uses Nvidia GPUs for training large AI models.

How many GPUs will Colossus 2 have by year‑end?

Elon Musk indicated that the cluster could double its current count, adding 440,000 GB300 chips and 110,000 GB200 chips, with further additions of 220,000 GB300 chips next week and another 220,000 in November.

Why is SpaceX’s GPU target significant for Nvidia?

SpaceX’s goal of 1 million GPUs by 2026, combined with the recent orders, could push Nvidia’s sales beyond current projections, reinforcing the company’s leadership in AI hardware.

What is the Vera Rubin platform?

The Vera Rubin platform is Nvidia’s architecture for high‑performance AI computing, praised by Musk for its excellence and used in SpaceX’s training and orbital data center plans.

Is Nvidia stock a good investment?

Given Nvidia’s dominant position, expanding customer base, and a forward P/E of 14x for fiscal 2028, many analysts consider the stock a buy.

Bottom Line

Elon Musk’s update on Colossus 2 confirms a rapid expansion of Nvidia GPU usage within SpaceX, potentially exceeding the company’s 1 million‑GPU target by 2026. Coupled with Nvidia’s broad AI ecosystem and attractive valuation, the partnership strengthens the case for Nvidia as a solid investment in the AI hardware space.

This is educational information, not personalized financial advice.

Reporting based on: original source.