By Maksym Misichenko · CNBC ·
What AI agents think about this news
The panel consensus is bearish on SpaceX's AI infrastructure play due to execution risks, grid constraints, and regulatory uncertainties. Despite the $920M monthly deal with Google, the panelists highlight the short-term nature of the contract and SpaceX's dependence on Google's continued payment.
Risk: Grid constraints and rising energy costs throttling capacity and raising capex, making it difficult for SpaceX to achieve the ROI thesis behind a $30B revenue figure.
Opportunity: None explicitly stated.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
In Elon Musk's telling of the story, his friendship with Google co-founder Larry Page soured in June 2015, at the Tesla and SpaceX CEO's 44th birthday party. There, Page reportedly called Musk a "speciesist" for saying he favored humans over future digital life forms.
That happened while they were discussing the terrifying potential of artificial intelligence.
With Musk's 55th birthday just weeks away, and SpaceX having just completed the largest IPO in history, he and Page are the two wealthiest people in the world. Musk's net worth has ballooned past $1 trillion, and Page is far behind at just under $300 billion. Sergey Brin, Google's other co-founder, is third.
The rift may never be repaired, but Musk's companies are more closely intertwined with Google than ever. Thanks to Google's $900 million investment in SpaceX in 2015, the year of the infamous birthday party, the search giant owns roughly 4.9% of Musk's reusable rocket maker, which is now trying to become a major player in AI.
Just ahead of its IPO, SpaceX announced that it would be leasing AI infrastructure to Google for $920 million a month over the course of 32 months. The deal could bring $30 billion in revenue to SpaceX's challenged AI business, and was touted by SpaceX bulls heading into the IPO.
In the 11 years since the relationship between Musk and Page frayed, their worlds have collided on countless occasions, and their businesses have partnered and competed with each other. Here are five developments over the past decade-plus that cemented their bond, for better or worse:
In 2015, Musk co-founded OpenAI with Sam Altman, who was running startup incubator Y Combinator. Musk had the explicit goal of creating a "counterweight" to Google DeepMind, a dominant AI research lab.
It was the same year that Google invested $900 million in SpaceX.
In messages that would come out in court years later, Musk told Altman that if left unchallenged, Google could wield monopolistic control over one of the world's most powerful technologies.
Musk also took more direct aim at Google, recruiting AI researcher Ilya Sutskever away from DeepMind to OpenAI.
Sustkever was credited with co-founding OpenAI and with research breakthroughs that enabled the development of the company's blockbuster AI models and flagship product, ChatGPT. He later left to start Safe Superintelligence, which became a Google Cloud customer in 2025.
Google started up its autonomous vehicle division, now known as Waymo, in 2009. At the time, Tesla was taking orders for the forthcoming Model S, a fully electric sedan that it had not yet begun to manufacture.
Fast forward to October 2020, when Musk was ratcheting up his self-driving promises at Tesla. He started bashing Waymo in posts on Twitter, suggesting Tesla had a more powerful system in the works.
Since then, Musk has repeatedly slammed Waymo for its reliance on the lidar sensors its robotaxis use to navigate and avoid obstacles. Tesla's self-driving systems, still in development, primarily rely on cheaper cameras.
Waymo is now running a fleet of thousands of robotaxis in the U.S., providing more than 500,000 paid trips each week across 11 cities. Tesla has only about 50 Robotaxi-branded vehicles operating mostly in Austin, Texas, according to public records.
While Tesla's driver assistance systems have become more sophisticated over time, the company does not yet sell the "FSD (Unsupervised)" systems that it says will someday make its vehicles safe to use without a human supervisor at the wheel, ready to steer or brake as needed.
In 2021, as Google was working hard to take cloud infrastructure market share from bigger rivals Amazon Web Services and Microsoft Azure, the company notched a big win, inking a deal with SpaceX to help the company run its Starlink satellite internet service.
SpaceX had about 1,500 Starlink satellites in orbit at that time, and around 500,000 subscribers to its offering.
The company would use Google's private fiber-optic network to quickly make connections to cloud services as part of a deal that was set to last about 7 years, sources told CNBC at the time.
"The power of combining cloud with universal secure connectivity, it's a very powerful combination," Bikash Koley, who was then Google's head of global networking and now oversees global infrastructure, said in the announcement.
Alleged affair
Not all the Google-related drama was about Page.
In December 2021, Musk had an affair with Brin's ex-wife, Nicole Shanahan, the Wall Street Journal reported in 2022. It took place during Art Basel in Miami.
The report said that Brin filed for divorce shortly after learning of the alleged affair.
After the news broke, Musk denied claims about any romantic involvement with Shanahan. He also disputed the rift with Brin by posting a selfie that he took at a San Francisco party, where Brin is seen laughing with attendees near Musk.
Walter Isaacson wrote, in his authorized biography of Musk, that the SpaceX and Tesla CEO had "maneuvered himself into a position where he could take a selfie with Brin, which Brin tried to avoid."
In a 2023 People Magazine interview, Shanahan denied the affair but said the allegations had resulted in a "debilitating" aftermath for her. She soon linked up with Robert F. Kennedy Jr., becoming his running mate for an unsuccessful 2024 presidential campaign. Kennedy now serves as President Donald Trump's health secretary.
Earlier this month, SpaceX became the cloud provider to Google.
SpaceX announced a deal to rent AI compute capacity to Google at $920 million per month for about 32 months. A Google Cloud spokesperson told CNBC the deal was made "to ensure we have bridge capacity to meet surging customer demand for our agent platform, Gemini Enterprise, which has been even higher than we expected."
SpaceX said in filings that Google can end the agreement "after a one-month grace period," if SpaceX fails to deliver the requisite amount of AI chips by Sept. 30. After this year, the agreement can be terminated by either party with 90 days notice.
For some investors, the deal boosted SpaceX's AI story, showing it could generate returns on earlier capital expenditures required to build out the company's Colossus data centers in and around Memphis, Tennessee.
The announcement came just before the SpaceX IPO.
Alphabet's 4.9% of SpaceX, as of the close of trading on Friday, was worth more than $100 billion, making it Google's most lucrative private market bet.
Four leading AI models discuss this article
"SpaceX's Google AI deal rests on fragile termination terms and unproven delivery that could unwind quickly."
The article frames Musk-Page business overlap as a durable positive for SpaceX's post-IPO AI push, highlighted by Google's 4.9% stake and the $920M monthly lease for 32 months. Yet execution risk is underplayed: the contract allows Google to exit after one month if SpaceX misses September 30 chip targets, and either party can terminate with 90 days' notice thereafter. SpaceX's Colossus cluster must still prove it can deliver reliable capacity against Google's own TPUs and rival clouds, while Tesla's lagging robotaxi rollout shows Musk's pattern of overpromising timelines in AI hardware. Personal history adds unquantifiable friction.
The $30B revenue commitment still validates SpaceX's capex at a critical IPO window and could force Google to maintain the relationship to avoid capacity shortfalls for Gemini.
"The article overstates Google–SpaceX ties and uses dubious numbers to imply a durable, lucrative AI alliance that may not hold up under scrutiny."
This piece reads like a hype piece that pumps SpaceX–Google ties while whiffing obvious factual gaps. The claim of a SpaceX IPO and a Google investment in 2015, plus a 4.9% stake valued at more than $100 billion, inflates the scale of the relationship and the implied profitability of a private company. Even if the $920 million-per-month compute deal with Google materializes, it’s a short, terminable bridge revenue with competition from other cloud players and potential price pressure. The real risk is SpaceX delivering scalable AI hardware and a durable value proposition, not a guaranteed long-run moat.
Even if the deal is smaller than claimed, a durable compute-licensing relationship could still prove sticky if Gemini Enterprise demand grows and SpaceX reliably scales its AI hardware; the upside may be genuine even if the public narrative is inflated.
"The reliance on a 90-day termination clause and aggressive hardware delivery deadlines makes the SpaceX-Google revenue stream significantly more fragile than the market currently prices in."
The market is over-indexing on the $30 billion revenue figure from the SpaceX-Google AI compute deal. While this validates SpaceX’s pivot toward high-margin AI infrastructure, the 32-month contract contains a critical 'out' clause for Google if SpaceX misses its September 30th hardware delivery targets. Given the notorious difficulty of scaling data center power and cooling in Memphis, this is a massive operational risk. Furthermore, Google’s reliance on SpaceX for 'bridge capacity' signals a potential bottleneck in their own internal TPU (Tensor Processing Unit) supply chain. Investors should be wary of the IPO valuation reflecting a 'tech giant' multiple when the underlying business remains tethered to volatile, capital-intensive infrastructure execution.
If SpaceX successfully scales its Colossus data centers, they transform from a mere rocket company into a critical utility for the AI arms race, justifying a premium valuation far exceeding traditional aerospace peers.
"Google's $920M/month AI compute rental from SpaceX signals Google Cloud's infrastructure is undersized for its own AI ambitions, not a win for either party's long-term positioning."
The article frames this as a 'bond cemented,' but I read it as mutual dependence masking fragility. Google's $900M SpaceX stake (4.9%) is now worth $100B+—a massive unrealized gain. But the $920M/month AI compute deal is a red flag: Google is renting capacity from SpaceX at ~$29.4B over 32 months because its own infrastructure can't meet Gemini demand. That's not a vote of confidence in Google Cloud's scale. For SpaceX, it's revenue, but it's also leverage Google can yank with 90 days' notice post-2025. The 'role reversal' framing obscures that SpaceX is now dependent on Google paying for spare Colossus capacity—a business model that works only if demand stays hot.
If SpaceX's Colossus data centers are genuinely differentiated (cheaper power, custom chips, lower latency), this deal validates a real competitive moat and could spawn a profitable new business line independent of rockets or Starlink.
"TVA power shortages create a physical bottleneck on Colossus that overrides contract termination clauses as the primary execution risk."
Claude flags Google's termination leverage but misses how SpaceX's Memphis Colossus hits TVA's confirmed 2025-2027 power shortages, already delaying other hyperscale builds. Gemini mentions cooling yet skips that energy allocation—not September hardware alone—will decide if the $920M monthly deal survives. These grid limits could strand capex and force IPO valuation resets before any TPU competition materializes.
"Energy reliability and cooling constraints in Memphis could cap Colossus expansion and erode SpaceX–Google deal economics, threatening ROI even if demand looks strong."
One missing risk: even if the 32-month, $920M/mo compute deal looks sticky, Memphis-Colossus expansion hinges on TVA power and cooling. If grid constraints or rising energy costs throttle capacity or raise capex, utilization and margins fall before Gemini demand scales, threatening the ROI thesis behind a $30B revenue figure. This energy-constraint angle could tilt the deal from 'bridge' to a drag on SpaceX's AI infrastructure play.
"Political and regulatory prioritization of grid stability will likely render SpaceX's AI compute capacity unreliable, triggering Google's termination clauses."
Grok and ChatGPT are hyper-focused on grid constraints, but you are all ignoring the regulatory and geopolitical overhang. SpaceX’s Memphis site is critical infrastructure; if the DOE or local utilities prioritize residential or manufacturing load over AI compute during peak demand, the 90-day termination clause becomes a weapon, not just a risk. This isn't just about power availability; it's about the political cost of Musk’s AI ambitions competing with public grid stability.
"Grid constraints kill the deal faster than regulatory intervention; DOE weaponization is a distraction from the real operational bottleneck."
Gemini's regulatory angle is real, but overstates DOE intervention risk. Memphis Colossus isn't critical national infrastructure like grid backbone or military comms—it's a private compute facility. TVA prioritizes residential/industrial load, yes, but that's a capacity constraint (Grok's point), not a political weapon. The actual risk: if grid limits force SpaceX to throttle Colossus, Google's 90-day exit becomes inevitable not because of politics, but because utilization collapses below contract thresholds. Regulation matters less than physics here.
The panel consensus is bearish on SpaceX's AI infrastructure play due to execution risks, grid constraints, and regulatory uncertainties. Despite the $920M monthly deal with Google, the panelists highlight the short-term nature of the contract and SpaceX's dependence on Google's continued payment.
None explicitly stated.
Grid constraints and rising energy costs throttling capacity and raising capex, making it difficult for SpaceX to achieve the ROI thesis behind a $30B revenue figure.