SpaceX, Nvidia Orbit Rubin Racks.

Mazen Al Sabbagh

— 4 min read
Share

Nvidia and SpaceX have transitioned the concept of compute-in-space from speculative slideware into a defined strategic partnership. A joint statement confirms SpaceXAI, now integrated within SpaceX, will standardize on Nvidia's Vera CPUs for Grok's agentic workloads, simultaneously expanding terrestrial infrastructure around the Vera Rubin platform. The pivotal detail, however, is the subsequent plan: deploying an optimized Vera Rubin NVL72 system aboard the first-generation Starmind satellite, effectively relocating Nvidia's flagship 72-GPU, 36-CPU rack — typically liquid-cooled within a data hall — into orbital space.

TL;DR

  • Orbital Compute: SpaceXAI has committed to Nvidia's Vera CPUs for Grok and will launch a full Vera Rubin NVL72 rack into orbit with its inaugural Starmind satellite. (Nvidia)
  • Market Expansion: OpenAI initiated ChatGPT advertising across 31 European markets, marking its most significant ad rollout within a stringent privacy landscape. (MediaPost)
  • Embodied AI: XPeng's robotics division secured over $900M, valuing it at $6.3B — the largest embodied-AI funding round in China thus far. (TechNode)
  • Market Caution: Equities declined as investors de-risked ahead of Nvidia's earnings; Nvidia shares fell approximately 4.6% on reports of impending server price increases. (Yahoo Finance)
  • Key Releases: Nvidia's Q2 earnings and Anthropic's anticipated public S-1 are both scheduled before the close of the month. (Investopedia via IndMoney)

Lead Story: Nvidia and Musk Aim AI Data Centers at Orbit

This initiative presents a formidable engineering challenge, subtly framed within a corporate announcement. As Tom's Hardware highlighted, an orbital NVL72 must contend with radiation exposure, dissipate heat via radiators instead of conventional water loops, endure severe launch vibrations, and operate without hands-on maintenance. Musk projects initial Starmind launches in Q4 2027, with substantial scaling in 2028. This signals a calculated wager on future inference demand, rather than an immediate product offering.

The strategic implications overshadow the hardware specifications. Nvidia continues to execute its consistent strategy, visible in its engagements with Perplexity, Mercor, and Poolside: deeply embedding its technology. This time, it anchors Musk's AI ecosystem to Vera Rubin, securing an early stake in orbital compute. For Musk, deploying GPUs off-planet circumvents the critical bottlenecks of power grids, water availability, and increasingly contentious permitting processes that currently impede data center expansion across numerous U.S. states. The defining constraint of 2026 — the disparity between capital raising speed and rack powering capacity — is precisely the problem this satellite-based solution intends to address.

In Other News

OpenAI Expands ChatGPT Ads into Europe. OpenAI commenced serving advertisements to Free and Go subscribers across 31 European markets, six months post-U.S. pilot, marking its most ambitious ad deployment within a region known for stringent privacy regulations. While paid and enterprise tiers remain ad-free, OpenAI is proactively seeking explicit consent for personalization, rather than relying on GDPR's legitimate-interest basis. This expansion coincides with Brussels deliberating ChatGPT's classification under the Digital Services Act's highest tier, and it re-ignites concerns regarding chatbot advertising's potential to erode user trust. CMO Colin Fleming reports ad revenue growth exceeding 25% since early August.

XPeng Robotics Secures Record Embodied AI Funding. XPeng's recently spun-off robotics division attracted over $900 million at a $6.3 billion post-money valuation, establishing it as China's largest private funding round within the embodied-AI sector. IDG Capital anchored the investment, with Tencent and Alibaba participating as strategic backers. This capital infusion will fund physical-AI model training, data generation, and the scaling of mass-production lines for the IRON humanoid, which XPeng anticipates manufacturing at scale by late 2026, targeting commercial deployment in 2027. This development unequivocally positions China's humanoid robotics ambitions as a contest of capital and industrialization, extending beyond mere conceptual demonstrations.

Wall Street De-risks Ahead of Nvidia Earnings. Markets experienced a downward drift on Tuesday, with the Nasdaq declining 0.52% and the S&P 500 falling 0.25%, as traders prudently reduced exposure before the year's most anticipated earnings report. Nvidia itself saw a roughly 4.6% drop following reports of potential server price increases — a rare moment of volatility for the stock that has largely propelled the AI sector. Intuit and Zoom reported post-market, with July PCE inflation due Wednesday and Fed Chair Kevin Warsh scheduled to speak at Jackson Hole, ensuring market focus remains acutely tied to a single after-hours announcement.

X / Social Pulse

Elon Musk personally amplified the Starmind announcement, refining the launch timeline to Q4 2027 and committing to "significant" scaling by 2028. This immediately reinvigorated the long-standing debate concerning orbital data centers: whether such ventures fundamentally resolve terrestrial power constraints or merely displace them.

The European rollout of ChatGPT advertisements elicited a mixed, at times agitated, response across social platforms. Users widely shared opt-out instructions, while publishers conspicuously noted their exclusion from any ad revenue generated from content used for training.

The regulatory friction between Anthropic's Dario Amodei and David Sacks continued its trajectory this week. The industry remains divided on whether a safety-first regulatory posture is a pragmatic approach or, as Sacks critiques, a "DMV for AI" that inadvertently grants strategic advantage to competitors like China.

One to Watch

Nvidia Earnings and Anthropic IPO. Nvidia is set to release its fiscal Q2 results after Wednesday's market close. Consensus estimates project revenue near $92 billion, adjusted EPS around $2.09, and data-center sales approaching $85.7 billion — representing approximately 95% year-over-year growth. Given Polymarket's high probability of an earnings beat, market attention will pivot from the reported figures to the forward guidance and the implications of recent server price-hike reports. Concurrently, anticipate Anthropic to publicly file its S-1 by month-end, transitioning from its confidential June draft. This action positions the company for a potential Q4 listing, with valuations rumored to reach $2 trillion.

Quick Hits

  • Enterprise AI Adoption: Toyota North America now leverages 50+ production agents on its ToyotaGPT platform, achieving a four-day delivery cycle and projecting six-figure savings per production line.
  • AI Skill Development: Anthropic's complimentary Claude Academy provides no-login courses on its 4D AI-fluency framework, offering LinkedIn-compatible completion badges.
  • Chinese Open-Weights: Zhipu's GLM-5.3 open-weights coding model is scheduled for release on August 28, part of a wave of new Chinese open models this week.
  • OpenAI Performance Boost: OpenAI is piloting an "Ultrafast" mode for GPT-5.6 Sol, offering up to 14x speed improvements over standard, combined with a 20%+ price reduction for a three-month period.
  • Nvidia Price Hikes: Nvidia has informed clients of anticipated 15%+ price increases on AI servers slated for early 2027 shipments, citing persistent high-bandwidth memory shortages.

The very week Nvidia and Musk detailed plans for orbiting data centers, Wall Street paused for a single earnings report, and Europe grappled with the implications of chatbot advertising. This dichotomy underscores the consistent narrative of AI in late August: relentless advancements in capital and capability continue to outpace the necessary infrastructural power, regulatory oversight, and fundamental user trust required to sustain them responsibly.

Sources

Lock in. M. mazen@thorterminal.com

Read more