The Compute Scarcity

Google is paying SpaceX $920 million per month — $30 billion total — for GPU access at xAI's Colossus data centers. Google has its own vast infrastructure. It is still renting compute from a competitor because there is not enough AI compute in the world to satisfy demand through internal build-out alone. Meta is spending $125-145 billion on infrastructure this year while cutting 8,000 employees. OpenAI filed its confidential S-1 targeting a $1 trillion-plus valuation. The compute scarcity that was described as theoretical 18 months ago is now showing up as nine-figure monthly payments to rivals.

In this edition: Google pays SpaceX $920M/month for AI compute  ·  OpenAI files confidential S-1: Goldman Sachs + Morgan Stanley, $1T+ target  ·  Meta: 8,000 layoffs, 7,000 redirected to AI, $125-145B infrastructure spend  ·  Harness $240M at $5.5B: AI-native software delivery

The Price of Compute Scarcity

Google signed a deal to pay SpaceX $920 million per month from October 2026 through June 2029 for access to approximately 110,000 NVIDIA GPUs at xAI's Colossus data centers. The total contract value is approximately $30 billion. Google is one of the most infrastructure-rich companies in the world. It is paying a competitor $11 billion per year for GPU capacity because the alternative is not having enough compute. That is the clearest available signal of where AI infrastructure stands in mid-2026.

Google has been building AI infrastructure since before most of its competitors understood the term. Its custom Tensor Processing Units are among the most powerful AI accelerators available. Its data center footprint is second only to Amazon's and Microsoft's in global scale. It still cannot build fast enough. The capital expenditure pipeline for AI data centers — driven by hyperscalers who committed to $725 billion combined in 2026 — has outrun the supply chain for the GPU clusters, power connections, cooling systems, and physical real estate required to house them. SpaceX built Colossus faster than anyone anticipated. Google needed compute faster than its own construction pipeline could deliver. The deal is the market-clearing price for that gap.

For enterprise leaders, the $920 million monthly figure is not primarily a story about Google's spending power. It is a data point about the AI infrastructure environment every organization is navigating. The same GPU scarcity that causes Google to rent from a competitor determines the availability and pricing of cloud AI inference for every enterprise customer on AWS, Azure, and Google Cloud. The governance dimension deserves equal attention: Google is running AI workloads on infrastructure owned by xAI, a direct competitor. The agreement includes data protection terms — Google retains ownership of its models and data — but the arrangement raises questions enterprise legal teams should apply to their own vendor agreements: when AI vendors outsource compute to third parties, what does the counterparty relationship look like?

Moving Pieces

Three developments that matter to enterprise leaders this week

OpenAI Files Confidential S-1 — Goldman Sachs and Morgan Stanley, $1 Trillion Target

OpenAI confirmed on June 8 that it had submitted a confidential S-1 to the SEC, joining Anthropic in the public market pipeline. Goldman Sachs and Morgan Stanley are underwriters. Analysts expect the IPO valuation to exceed $1 trillion, with a listing window targeted between September and November 2026. OpenAI was valued at $852 billion in its most recent private round. The simultaneous IPO track for both companies — the two largest independent AI labs — will produce the most significant data set on AI company financial structures and revenue quality that enterprise procurement teams have ever had access to. Enterprise revenue now accounts for more than 40% of OpenAI's total, on track for parity with consumer by year-end. When the S-1 becomes public, the unit economics of enterprise AI vendor relationships will be readable in a way they never have been before.

Meta: 8,000 Layoffs, 7,000 Redirected to AI Teams, $125-145 Billion in 2026 CapEx

Meta began notifying approximately 8,000 employees of layoffs — the largest companywide reduction since the 2022-23 "Year of Efficiency" that eliminated 21,000 positions. An additional 7,000 employees are being redirected into newly created AI-focused teams. Context: Meta has committed to $125-145 billion in capital expenditures in 2026, more than twice its 2025 outlay, almost entirely directed at AI infrastructure. Chief People Officer Janelle Gale described the cuts as freeing up $8-10 billion for AI data center construction. The arithmetic is explicit — human labor is being converted into infrastructure budget. For enterprise leaders making similar rebalancing decisions: the Meta restructuring provides a documented framework at scale. Workforce rationalization funds the infrastructure investment that funds the AI capability that the organization's competitive position requires. The sequencing is deliberate and the trade-off is public.

Harness Raises $240 Million at $5.5 Billion — AI-Native Software Delivery for Enterprise

Harness closed a $240 million Series E at a $5.5 billion valuation, focused on automating the "after-code" phase of software delivery — the complex steps of testing, securing, and deploying software that follow code generation. The company uses AI agents and a proprietary software delivery knowledge graph to orchestrate testing, security scanning, and deployment pipelines without manual configuration, serving more than 1,000 enterprise customers. The funding round is significant in the context of GitHub Copilot's move to usage-based billing and Microsoft's MAI-Code-1-Flash: as AI coding assistants generate more code faster, the bottleneck in software delivery has shifted downstream. Harness is the enterprise bet that the largest productivity gains in software delivery are not in code generation — that problem is increasingly solved — but in the testing, security, and deployment steps that determine whether generated code ships safely and at scale.

the total value of Google's compute agreement with SpaceX — $920 million per month from October 2026 through June 2029, for 110,000 NVIDIA GPUs at xAI's Colossus data centers.

The $30 billion figure is not a forecast or a capital plan — it is a signed contract between two of the world's most sophisticated infrastructure operators, disclosed in an SEC filing. Google, which operates more than two dozen global data center regions and builds its own AI chips, is paying a competitor $30 billion because its own infrastructure pipeline cannot keep pace with demand. For enterprise AI teams assessing multi-year infrastructure commitments, vendor availability assumptions, and cloud inference pricing: the conditions that produced this deal are not temporary. They are the operating environment for AI infrastructure in 2026.

Sources: TechCrunch  ·  SEC Filing / SpaceX  ·  June 5, 2026

When Google Rents from xAI

The Google-SpaceX deal is worth thinking about slowly. Google is not a company that lacks resources, infrastructure expertise, or capital. It is paying $920 million per month to an Elon Musk entity — one whose AI product directly competes with Google's Gemini.

Google looked at its internal build timeline, looked at its demand curve, and concluded that paying a competitor $11 billion per year was the rational decision. That is the signal. Not the dollar amount. Not the unusual counterparty. The signal is that the most infrastructure-capable company in the world found that gap and could not close it internally. Enterprise leaders drawing AI infrastructure investment plans are often working with assumptions built in 2024: that cloud compute availability would normalize, that GPU prices would fall as supply caught up, that AI inference would become a commodity utility within 18-24 months. The Google-SpaceX deal is a direct contradiction of each of those assumptions.

The practical question for enterprise AI teams is not whether their organization can sign a $30 billion compute deal. It is whether their infrastructure planning reflects the environment that deal describes: one where compute availability should be locked in, not assumed, and where waiting for normalization is a strategy that costs more than acting now. Review your cloud AI commitments against a three-year time horizon. Factor infrastructure availability risk into your AI roadmap. When the Anthropic and OpenAI S-1s become public, read the infrastructure cost section — it will tell you more about sustainable unit economics of AI vendor relationships than any sales conversation can.

Supply is not catching up. Prices for premium GPU capacity are not falling. The companies with the most aggressive long-term compute commitments will have the most predictable cost structures for the AI workloads that determine their competitive position. The Google-SpaceX deal is the benchmark. Plan against it.

The compute environment is not temporary. Plan accordingly.

AK  /  Spearhead  /  Building AI systems, not tools