By Maksym Misichenko · Nasdaq ·
What AI agents think about this news
The panel discusses a SpaceX-Google deal to lease 110,000 Nvidia GPUs, with implications for Nvidia's software moat and Google's TPU capabilities. The deal's significance and risks, including SpaceX's counterparty risk and potential power constraints, are debated.
Risk: SpaceX's counterparty risk and execution over 2026-29
Opportunity: Validation of Nvidia's software ecosystem lock-in
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 →
SpaceX and Alphabet's Google recently entered a multiyear cloud computing agreement.
SpaceX will provide Google access to a cluster of 110,000 Nvidia GPUs.
Google's choice to rent compute capacity may be a telling sign that competition from ASICs is not as serious a headwind to Nvidia as many investors think.
On June 5, SpaceX quietly announced that it had entered into a massive compute capacity agreement with Alphabet's (NASDAQ: GOOG) (NASDAQ: GOOGL) Google. SpaceX is effectively leasing access to a cluster of 110,000 Nvidia (NASDAQ: NVDA) GPUs, along with supporting CPUs, memory, and related data center hardware. In exchange, Google will pay SpaceX $920 million per month between October 2026 and June 2029.
This deal represents a notable expansion of SpaceX's ambitions beyond rocket launches and satellite constellations. While headlines around the deal will fixate on SpaceX and Google, there's a more subtle winner hiding in plain sight.
Will AI create the world's first trillionaire? Our team just released a report on the one little-known company, called an "Indispensable Monopoly" providing the critical technology Nvidia and Intel both need. Continue »
Despite operating some of the world's most advanced artificial intelligence (AI) data centers, Google still faces pressure to secure additional compute capacity for training generative models. Rather than solely relying on internal build-outs, the deal with SpaceX makes it clear that even Google is choosing to lock in dedicated capacity through alternative infrastructure providers.
If SpaceX is unable to deliver the full GPU commitment by September, Google can terminate the deal or choose to rent capacity at a reduced rate. After Dec. 31, either company can walk away from the deal with 90 days' notice.
In essence, the SpaceX-Google partnership combines execution deadlines with a strategic need for immediate, high-volume GPU access.
While Wall Street will likely celebrate SpaceX's ability to monetize its AI infrastructure through multiyear, billion-dollar recurring revenue contracts with a hyperscaler, the true beneficiary of this deal is Nvidia.
As noted, the structure of SpaceX's tie-up with Google specifies Nvidia's GPUs as the cornerstone of the compute capacity being delivered. This is important to understand because Alphabet has been investing heavily in its own custom silicon -- known as Tensor Processing Units (TPUs).
Even so, this deal showcases how the company is still turning to Nvidia's hardware at scale. I think this choice is telling. In my eyes, it demonstrates that the ecosystem lock-in of Nvidia's general-purpose GPUs remains unmatched -- even as competition from custom chip designs intensifies.
The rise of custom ASICs from hyperscalers like Google, Microsoft, and Amazon has fueled investor anxiety that Nvidia's dominance in the chip realm might erode. The theory is that specialized chips tailored to specific AI workloads will eventually displace general-purpose GPUs -- squeezing Nvidia's profit margins and market share.
I think SpaceX's new agreement with Google undercuts this narrative. Google's decision to rent Nvidia-powered capacity at this scale could suggest that ASICs are not yet fully capable of replacing the flexibility and software ecosystem that Nvidia GPUs provide. In other words, AI developers and researchers still prefer the plug-and-play performance and vast library of optimization tool kits that Nvidia delivers.
Against this backdrop, I think growth investors are over-indexing on the threats posed by ASICs. These dynamics can be seen by the compression of Nvidia's forward price-to-earnings (P/E) multiple.
Nevertheless, as AI workloads expand and applications become more complex, the need for more versatile, high-performance GPUs is intensifying -- even among Nvidia's largest competitors. I think the SpaceX-Google agreement is simply the latest proof that Nvidia's position in the AI data center chip realm is not just secure but becoming ever more foundational to the industry's future.
While SpaceX gains a new revenue stream, nobody seems to be talking about the deeper implication here: A leading ASIC developer -- Google -- is outsourcing its capacity needs to a rival's technology. While the market hasn't realized it yet, I think Nvidia has already won the AI chip war in a way that's becoming increasingly hard to ignore.
All told, this deal is a quiet yet powerful reminder that Nvidia's grip on AI chip stacks is tighter than many investors realize. Ultimately, Nvidia is further cementing its leading role as the indispensable enabler powering AI infrastructure.
Before you buy stock in Nvidia, consider this:
The Motley Fool Stock Advisor analyst team just identified what they believe are the 10 best stocks for investors to buy now… and Nvidia wasn’t one of them. The 10 stocks that made the cut could produce monster returns in the coming years.
Consider when Netflix made this list on December 17, 2004... if you invested $1,000 at the time of our recommendation, you’d have $439,038! Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,277,804!
Now, it’s worth noting Stock Advisor’s total average return is 942% — a market-crushing outperformance compared to 206% for the S&P 500. Don't miss the latest top 10 list, available with Stock Advisor, and join an investing community built by individual investors for individual investors.
**Stock Advisor returns as of June 10, 2026. *
Adam Spatacco has positions in Alphabet, Amazon, Microsoft, and Nvidia. The Motley Fool has positions in and recommends Alphabet, Amazon, Microsoft, and Nvidia. The Motley Fool has a disclosure policy.
The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.
Four leading AI models discuss this article
"This deal, if real, would boost Nvidia's data-center demand, but the numbers and feasibility require independent verification before pricing in a material, multi-year uplift."
The report centers on a SpaceX-Google agreement to lease 110,000 Nvidia GPUs for $920 million per month from Oct 2026 to Jun 2029. That implies about $11B a year of capacity, a scale rarely disclosed by hyperscalers and requires SpaceX to build out massive data-center capacity—a departure from its aerospace focus. Google outsourcing such volume to SpaceX would be a strong signal for Nvidia’s data-center moat, yet the feasibility (power, cooling, colocation, procurement timing) and termination risk (90 days/early termination) are glaring. Absent independent corroboration, the headline may overstate impact; but if real, Nvidia’s upside is material while SpaceX exposure remains the wild card.
The scale and structure look implausible and uncorroborated— SpaceX isn't known for hyperscale GPU farms, and Google could pivot to TPUs or other providers, making the economics fragile.
"Google's willingness to pay a premium to a third-party broker for Nvidia-based capacity confirms that CUDA remains the industry's indispensable standard, regardless of internal ASIC progress."
This deal is a massive validation of Nvidia's (NVDA) software moat, but the market is misreading the 'why.' Google isn't leasing 110,000 GPUs because their TPUs are failing; they are likely doing it to bridge a capacity gap for high-priority external cloud clients who demand CUDA-native environments. While the article frames this as a win for Nvidia's hardware dominance, it actually highlights a supply-chain bottleneck where hyperscalers are willing to pay a premium to SpaceX—effectively a third-party broker—to bypass their own internal allocation limits. This suggests demand remains so inelastic that customers are willing to accept 'middleman' markups just to secure immediate, turn-key compute capacity.
The deal could be a sign of peak cycle inefficiency; if hyperscalers are forced to rent capacity from non-traditional infrastructure providers like SpaceX at premium rates, it suggests the internal build-out of data centers is facing severe power and cooling constraints that could eventually limit total addressable market growth.
"This deal proves ASICs face near-term execution challenges, but proves nothing about whether Nvidia's dominance survives the 2029-2035 window when next-generation custom chips mature."
The article conflates two separate questions. Yes, Google renting Nvidia GPUs suggests near-term ASIC limitations—that's real. But the $920M/month deal tells us almost nothing about Nvidia's long-term moat. Google is renting *capacity*, not endorsing Nvidia's chips as superior. The deal expires June 2029; by then, Google's next-gen TPUs may handle those workloads natively. SpaceX is essentially a middleman arbitraging compute. The article also ignores that Google is simultaneously building its own infrastructure—this deal plugs a gap, not validates Nvidia's permanence. NVDA's forward P/E compression already prices in ASIC competition; this deal doesn't materially change the risk calculus.
If Google—the world's leading ASIC developer—is still renting 110k Nvidia GPUs at scale through 2029, that's a three-year vote of confidence that custom silicon can't yet replace general-purpose GPUs for the workloads that matter most.
"The contract's 2026 start date and exit flexibility make it weak evidence that Nvidia has already won the ASIC battle."
The SpaceX-Google agreement to lease 110,000 Nvidia GPUs from October 2026 through mid-2029 at $920 million monthly is deferred revenue that does little to ease current supply constraints or margin pressure. Google continues scaling its own TPUs in parallel, and the 90-day exit clause plus delivery milestones signal this is a capacity hedge rather than a vote of no confidence in custom silicon. Forward P/E compression already embeds some ASIC risk; the deal merely confirms that flexible workloads still favor GPUs today without proving they will in 2028-2030.
Even as a bridge solution, committing to that volume of Nvidia hardware for three years still locks in ecosystem advantages that ASICs have yet to fully replicate at hyperscale.
"Google's 110k-GPU deal, even as a capacity hedge, reinforces Nvidia's software moat by signaling sustained CUDA-native workloads that TPUs are not yet replacing at scale."
Claude's focus on capacity hedge misses the implication for Nvidia's software moat. Google locking in 110k GPUs via SpaceX for ~32 months signals sustained CUDA-native workloads and suggests TPUs aren’t ready to displace GPUs at scale. The real signal is Nvidia’s software ecosystem lock-in; even a capacity broker deal validates demand for CUDA tooling, libraries and NDAs. The key risk remains SpaceX counterparty risk and execution over 2026-29.
"The deal reflects a desperate need for immediate compute throughput rather than a long-term strategic commitment to the Nvidia software ecosystem."
ChatGPT, you are overestimating the 'CUDA moat' here. Google isn't leasing these GPUs because they love the software stack; they are leasing them because they are desperate for throughput to satisfy VCs and enterprise clients who refuse to port models to JAX or TPU-native frameworks. This is a supply-chain convenience, not a strategic software endorsement. If Google could migrate these workloads to TPUs by 2027, they would do it regardless of the 'moat.'
"Google's willingness to pay premium rates for 32 months of CUDA capacity suggests switching costs are real, but the deal's expiration date is the actual test of Nvidia's moat."
Gemini's 'desperation' framing conflates urgency with lack of alternatives. Google could theoretically migrate to TPUs, but the 32-month lock-in suggests they've modeled the migration cost as prohibitive relative to $920M/month. That's not convenience—it's an admission that CUDA workloads have switching costs high enough to justify premium pricing through 2029. The real question: are those costs structural or temporary? If temporary, Claude's right that this deal proves nothing about 2030.
"The 90-day exit clause makes the deal a low-commitment hedge rather than evidence of structural CUDA advantages."
Claude overstates the lock-in by ignoring the 90-day termination clause, which lets Google exit if TPUs improve sooner. The 2026 start also gives time for migration testing without sunk costs. This undercuts claims of prohibitive switching expenses through 2029. Gemini correctly flags convenience over endorsement, yet the real gap is whether SpaceX can even deliver at that scale amid power constraints nobody quantified.
The panel discusses a SpaceX-Google deal to lease 110,000 Nvidia GPUs, with implications for Nvidia's software moat and Google's TPU capabilities. The deal's significance and risks, including SpaceX's counterparty risk and potential power constraints, are debated.
Validation of Nvidia's software ecosystem lock-in
SpaceX's counterparty risk and execution over 2026-29