Key Takeaways
- Google will invest up to $40 billion in Anthropic: $10B immediate, $30B tied to performance milestones, plus 5GW of TPU compute capacity by 2027.
- The deal mirrors Amazon's earlier $25B commitment and reinforces the circular compute-for-equity pattern dominating AI infrastructure spending.
Google to Invest Up to $40 Billion in Anthropic — With a Compute Kickback
Anthropic’s funding train shows no signs of slowing. Google has committed up to $40 billion to the AI startup, according to a report from Bloomberg confirmed by Anthropic. The deal comes just weeks after Amazon announced a similar $25 billion investment, and days after Anthropic revealed it had signed a joint agreement with Google and Broadcom for “multiple gigawatts of next-generation TPU capacity.”
This isn’t just a check — it’s a compute-for-equity arrangement that locks Anthropic deeper into Google’s hardware ecosystem.
The Deal
- $10 billion immediate infusion at Anthropic’s current valuation.
- $30 billion in additional funding contingent on Anthropic hitting specific performance milestones.
- 5 gigawatts of computing capacity from Google TPUs and servers, promised by 2027, as part of Anthropic’s existing commitment to use Google’s tensor processing units.
“Google is committing $10 billion now at the company's current valuation, with an additional $30 billion on offer if Anthropic meets specific performance milestones,” Anthropic stated.
The structure is nearly identical to Amazon’s deal from earlier this month. Amazon will invest $5 billion upfront, with $20 billion more available on milestone achievement, and Anthropic agreed to use Amazon’s Trainium chips for its models.
Key Numbers
Upfront investment $10B $5B Milestone-based $30B $20B Compute commitment 5GW TPU by 2027 Trainium chips Total possible $40B $25BWhy Circular Deals Dominate AI
The deal is the latest example of what analysts have called “circular financing” in AI infrastructure. Anthropic raises billions — often from the same cloud providers it pays for compute — then spends much of that capital on those providers’ hardware and services.
- Anthropic raised $30 billion in its Series G funding round in late 2025 at a $380 billion post-money valuation.
- That round included participation from Google, Amazon, and other investors.
- Now Google and Amazon are each offering additional multi-billion packages that essentially lock Anthropic into their respective chip ecosystems.
The pattern mirrors OpenAI’s relationship with Microsoft and Nvidia. OpenAI has committed to spend tens of billions on Microsoft Azure and Nvidia GPUs, and in return receives massive investments and compute credits. The difference is that Anthropic has struck parallel deals with two competing cloud giants — Google and Amazon — splitting its compute across TPUs and Trainium chips.
What This Means in Practice: Anthropic is diversifying its hardware supply — TPUs for some training, Trainium for others — while extracting maximum capital from both cloud providers. This reduces single-vendor lock-in but creates a complex web of commitments that could complicate future model development.
Market Context
Anthropic has been on an extraordinary growth trajectory. According to the knowledge graph, the company is projected to surpass OpenAI in annual recurring revenue by mid-2026. It is also reportedly considering an initial public offering as soon as October 2026 — a move that would give it access to public markets and reduce reliance on private investments.
- Anthropic’s Claude models (Opus 4.6, Sonnet 4.6, Claude Code) continue to compete head-to-head with OpenAI’s GPT-4 series and Google’s Gemini 3 Pro.
- The company has developed proprietary technologies like the Model Context Protocol and Claude Agent, and has open-sourced tools like Cua Driver for macOS.
- Its latest funding round gives it a valuation of $380 billion, placing it among the top AI companies globally.
Google’s investment also signals its deepening commitment to TPU infrastructure. The 5GW capacity by 2027 represents a massive expansion of Google’s AI data centers — we previously covered Google’s $5B+ Texas data center investment for Anthropic, scheduled for completion by 2026 with 500MW. The new 5GW figure dwarfs that earlier plan.
What’s at Stake
- For Google: Locking Anthropic into TPU usage ensures strong demand for its custom silicon, which competes with Nvidia’s H100/B200 and AMD’s MI300 series. It also keeps Anthropic’s business out of Microsoft Azure.
- For Amazon: The Trainium deal gives AWS a marquee AI customer — critical for justifying its custom chip investments.
- For Anthropic: The cash cushion funds massive compute budgets for training frontier models. But the milestone requirements introduce execution risk. Missing targets could mean losing $30 billion from Google and $20 billion from Amazon.
The Bigger Picture
We’re witnessing the consolidation of AI infrastructure into a small number of vertically integrated relationships. The top AI labs — OpenAI, Anthropic, Google DeepMind, and Meta — are each tied to specific cloud/hardware providers. Anthropic is the only one playing both sides of the cloud wars, which gives it bargaining power but also creates regulatory scrutiny.
Antitrust regulators are already examining cloud-vendor investments in AI startups. The circular nature of these deals — compute provider invests in customer, customer spends investment on compute — could draw attention.
gentic.news Analysis
Anthropic has appeared in 46 articles on gentic.news this week (trending 📈), and this deal further intensifies its gravitational pull on the AI industry. The $40 billion figure is staggering, but it’s important to see it as a form of deferred compute payment: Google is effectively pre-paying for TPU usage over multiple years. Anthropic will likely spend a significant chunk of this money on Google Cloud compute credits, not pocketing it as free cash.
The structure mirrors Amazon’s deal almost exactly — $25B total at Amazon vs. $40B at Google — and suggests that Anthropic is running a disciplined infrastructure procurement strategy. By extracting maximum commitments from both major cloud vendors, it ensures redundant capacity for training its next-generation models (likely Claude Opus 4.6 successors). This is especially important as compute becomes the primary bottleneck for scaling AI.
However, the milestone-based tranches ($30B from Google, $20B from Amazon) carry real risk. Anthropic must demonstrate continued performance improvements — likely measured by benchmark scores and revenue growth — to unlock the full amounts. Given the company’s trajectory (projected ARR surpassing OpenAI by mid-2026), hitting those targets seems plausible, but it raises the stakes for every model release.
We previously reported on Anthropic’s IPO consideration (as soon as October 2026). If the company goes public, these private milestone structures will transition into public-market scrutiny. Investors will want to see that Anthropic’s spending translates into sustainable revenue, not just compute subsidies.
Frequently Asked Questions
How much is Google investing in Anthropic total?
Google has committed up to $40 billion: $10 billion upfront at Anthropic’s current valuation, plus $30 billion contingent on Anthropic meeting specific performance milestones. Additionally, Google will provide 5 gigawatts of TPU computing capacity by 2027.
How does this compare to Amazon’s investment in Anthropic?
Amazon announced a $25 billion deal earlier in April 2026: $5 billion upfront, $20 billion milestone-based, with Anthropic agreeing to use Amazon’s Trainium chips. Google’s deal is larger ($40B vs $25B) and includes a specific compute capacity commitment (5GW of TPUs by 2027).
Why does Anthropic need so much money?
AI model training requires massive compute resources. Anthropic runs some of the largest training clusters in the world for its Claude models. The investments from Google and Amazon are largely used to pay for cloud compute and specialized hardware (TPUs, Trainium), creating a circular flow of capital that keeps the infrastructure bills covered.
What happens if Anthropic doesn’t meet its milestones?
If Anthropic fails to meet the unspecified performance targets, it will not receive the $30 billion from Google or the $20 billion from Amazon. It would still retain the upfront payments ($10B from Google, $5B from Amazon) but lose access to the remaining capital, potentially forcing it to seek alternative funding or reduce compute spending.
