Google will invest up to $40 billion in Anthropic, with $10 billion locked in immediately at a $350 billion valuation and the remaining $30 billion contingent on the AI lab hitting undisclosed performance targets, according to terms reported on April 24, 2026. The agreement layers another five gigawatts of Google Cloud capacity on top of an existing one-million-TPU commitment, deepening one of the most lopsided cloud relationships in the artificial intelligence industry.
The pricing is the part most coverage skipped. Google paid in at a roughly $30 billion valuation discount to Anthropic’s February 2026 Series G round, which closed at $380 billion post money, and at less than half the $800 billion mark venture firms reportedly floated in early April. In exchange, Google secures the buyer’s seat next to Amazon ahead of an Anthropic IPO that bankers expect as soon as October.
Inside the structure of Google’s $40 billion commitment
The deal arrives in two tranches. The first $10 billion lands now in cash, priced at a $350 billion pre-IPO valuation. The remaining $30 billion converts only if Anthropic clears specific commercial milestones that neither company has disclosed publicly.
Alphabet has already poured more than $3 billion into Anthropic since 2023 and most recently held roughly 14 percent of the company on a fully diluted basis. The new tranche tightens that grip without triggering a control filing, a structure that has become standard in cloud-AI tie-ups.
Google Cloud will also supply five additional gigawatts of compute capacity to Anthropic over the next five years. That sits on top of the October 2025 commitment to up to one million Tensor Processing Units, which Anthropic said would put more than a gigawatt online during 2026.
Why the $350 billion price tag is the real story
Anthropic closed its Series G on February 12, 2026, raising $30 billion at a $380 billion post-money valuation in a round led by Coatue and Singapore’s GIC, with checks from Microsoft and Nvidia. By mid-April, TechCrunch reported that venture firms were knocking with preemptive offers as high as $800 billion. Anthropic walked past them.
So why did Google get a markdown? Two reasons. The first $10 billion is structured as a strategic primary investment, not a secondary purchase from existing holders, which gives Google preferred terms. The second is that Google is the seller on the other side of the table. Most of the cash will return to Alphabet’s books as Google Cloud revenue, a structural reality that justifies discounted equity in private term sheets.
The discount is effectively a rebate on the next decade of TPU spending.
The circular flow: where the money actually ends up
Cash in, compute out. Anthropic on April 7, 2026, expanded its agreement with Google and Broadcom for multi-gigawatt deployments of seventh-generation Ironwood TPUs that begin coming online in 2027. Industry analysts at SemiAnalysis pegged the first tranche at 400,000 Ironwood units worth roughly $10 billion in finished racks sold directly to Anthropic by Broadcom, with another 600,000 rented through Google Cloud Platform.
Google’s Ironwood TPU technical brief claims a 10x peak performance jump over TPU v5p and more than 4x the per-chip throughput of last year’s Trillium silicon. Anthropic is the anchor customer that lets Google amortize a chip program competing with Nvidia at hyperscale.
The arithmetic looks circular because it is. Google writes a $10 billion check. Anthropic books the cash, then commits a multiple of it back across years of TPU rentals, custom-silicon racks, and Google Cloud services. Alphabet records the inflow as cloud revenue, which lifts segment growth and supports the share price.
Why Anthropic’s revenue made the bidding war worth it
Anthropic’s annualized run-rate revenue crossed $30 billion in March 2026, up from about $9 billion at the close of 2025. CFO Krishna Rao confirmed the figure in the company’s compute-deal release, calling it the justification for the largest infrastructure commitment in the firm’s history.
The growth concentrates in business buyers. Anthropic counts more than 300,000 enterprise customers, and the cohort spending over $1 million a year doubled in under two months heading into April. Claude Code, the company’s developer product, reached $2.5 billion in annualized revenue by February 2026.
That revenue ramp briefly put Anthropic ahead of OpenAI in annualized sales, with OpenAI reported near $25 billion ARR in the same window. For Google, those numbers are evidence that Claude is winning the enterprise contracts Gemini has struggled to close.
https://x.com/AnthropicAI/status/2041275563466502560
Amazon’s parallel $25 billion bet four days earlier
Google’s announcement landed four days after Amazon agreed to invest up to $25 billion more in Anthropic, on top of the $8 billion AWS had already committed. The Amazon deal mirrors Google’s structure: $5 billion now at the same $380 billion February valuation, $20 billion contingent on commercial milestones.
The compute side of Amazon’s agreement is even larger. Anthropic pledged to direct more than $100 billion toward AWS technologies over the next decade, including nearly one gigawatt of Trainium2 and Trainium3 silicon online by year-end. AWS remains Anthropic’s primary cloud provider despite the Google capacity surge.
| Backer | Headline commitment | Cash now | Contingent | Compute attached |
|---|---|---|---|---|
| Google (Apr 24, 2026) | Up to $40B | $10B at $350B valuation | $30B on milestones | 5 GW over 5 years |
| Amazon (Apr 20, 2026) | Up to $25B | $5B at $380B valuation | $20B on milestones | ~1 GW Trainium by end of 2026 |
Antitrust scrutiny is no longer hypothetical
Federal regulators flagged exactly this pattern more than a year ago. The Federal Trade Commission’s January 2025 staff report on cloud-AI partnerships reviewed the Microsoft-OpenAI, Amazon-Anthropic, and Google-Anthropic relationships, finding that the deals already involved more than $20 billion in cumulative funding and significant non-monetary control rights. Staff warned the structures could increase switching costs and limit access to compute for outside developers.
Senators Elizabeth Warren and Ron Wyden have since sent inquiry letters probing the cloud-AI tie-ups, and DOJ antitrust officials have publicly questioned whether the arrangements function as quasi-acquisitions that escape merger review. The April 24 Google commitment expands a relationship the FTC already had under its lens.
Madhavi Singh, deputy director of the Thurman Arnold Project at Yale School of Management, argued in an April 22 analysis that adjacent industry consortiums Anthropic helps lead may carry independent Sherman Act exposure on top of the bilateral deal terms.
The clock is the IPO
Anthropic executives have told banking partners they are targeting an October 2026 listing on Nasdaq, with Goldman Sachs and JPMorgan Chase advising on what would be a $60 billion-plus raise. Wilson Sonsini is finalizing the S-1 draft, which is expected to land in late summer.
That timeline explains the speed of the April deal sequence. Google needed its position priced and disclosed before public-market investors set the float. Amazon needed the same. Both want their cloud commitments locked into a registration statement that retail buyers will read as evidence of demand durability.
Bankers are privately modeling a debut valuation between $400 billion and $500 billion, well above Google’s entry point and below the $800 billion VC indication. If that range holds, Alphabet’s $10 billion April tranche is already worth a paper premium before Anthropic prints its first ticker.
What it means for the AI market in mid-2026
The competitive picture is inverting. A year ago Google sold Gemini as the answer to OpenAI. In April 2026, Alphabet is paying its largest-ever strategic check to make sure a competing model family runs on its silicon and inside its data centers.
Dario Amodei, Anthropic’s chief executive, has framed the spending as a response to demand he says shows no sign of slowing. He has also warned, in interviews this year, that the AI industry is engaged in a period of “YOLO spending” that will not pay off for everyone writing checks.
For background on Anthropic’s product trajectory feeding this growth, see our coverage of the April 16 launch of Claude Opus 4.7 and its coding benchmark scores, and our reporting on the source-code leak that exposed unreleased Anthropic systems earlier this month. The competitive backdrop also includes Meta’s recent Muse Spark debut from its Superintelligence Labs, which has reset enterprise expectations for proprietary models.
Anthropic’s next disclosed milestone is the S-1 filing. Until then, the only real number that matters is the one Google paid: $350 billion, on a company that may print a public market cap with a five in front of it within six months.
Leave a Comment