Anthropic is significantly expanding its artificial intelligence infrastructure plans, with sources telling The New York Times that the company expects to reach more than 3.5 gigawatts of compute capacity by 2026.reported This growth strategy involves a broader mix of hardware vendors and cloud partners than previously reported.reportedAnthropic is building a much larger system for running artificial intelligence, using chips from two different companies and services from big cloud firms.Anthropic is executing a multi-vendor compute strategy, integrating AMD silicon and hyperscaler capacity to achieve a 3.5GW target by 2026.
Evidence: reported
Sources detail Anthropic's compute deal spree: 3.5GW+ of expected capacity in 2026, diversifying beyond Nvidia's AI chips to Amazon, Google, and AMD, and more (New York Times)aifleet-techmeme
What was released or claimed
According to The New York Times, Anthropic is finalizing a series of compute deals that expand its direct hardware procurement.reported The report states the company is moving beyond relying solely on Nvidia’s AI chips. Instead, Anthropic is adding AMD to its list of direct hardware suppliers. The company is also securing capacity through major cloud providers, specifically naming Amazon and Google as partners in this expansion.
Evidence: reported
Sources detail Anthropic's compute deal spree: 3.5GW+ of expected capacity in 2026, diversifying beyond Nvidia's AI chips to Amazon, Google, and AMD, and more (New York Times)aifleet-techmeme
Who it affects
The primary entities involved in these transactions are Anthropic, the AI developer; Nvidia and AMD, the semiconductor manufacturers providing the hardware; and Amazon and Google, the cloud infrastructure providers. These deals reshape the supply chain for large-scale AI training and inference operations.
What is still unknown
The New York Times report does not specify the exact financial terms of these deals or the specific models of AMD and Nvidia chips being procured. It is also unclear how much of the 3.5 gigawatt target will come from direct hardware purchases versus cloud services. The timeline for when these capacities will come online remains unconfirmed.unverified
What to make of it
That suggests Anthropic is prioritizing supply chain diversification to secure the massive compute resources required for its future models. By engaging multiple chipmakers and cloud providers, the company may be mitigating the risk of supply constraints from any single vendor. This move indicates a shift toward a more complex, multi-vendor infrastructure strategy for large AI labs.
What buyers should weigh
Organizations evaluating AI infrastructure should consider how Anthropic’s diversification into AMD and cloud providers might affect market dynamics. The move signals that viable alternatives to exclusive Nvidia dependence are emerging for large-scale deployments. Buyers may need to assess the readiness of AMD-based stacks for their specific workloads. The inclusion of Amazon and Google suggests hybrid cloud strategies are becoming standard for top-tier AI development.
Why it matters
- The expansion to 3.5 gigawatts highlights the immense scale of capital and energy required for next-generation AI models.
- Diversifying to AMD reduces reliance on a single chip vendor, potentially increasing market competition and supply stability.
- Partnerships with Amazon and Google indicate that cloud providers remain critical partners even as companies buy direct hardware.
- The timeline of 2026 provides a benchmark for when significant new AI capacity may become available in the market.
Topics: Anthropic, AMD, Nvidia, AI infrastructure, compute capacity
Updated
Then / now / next
Anthropic primarily relied on Nvidia for its artificial intelligence compute infrastructure.
The company is diversifying its supply chain to include AMD and major cloud providers like Amazon and Google.
Industry observers will watch for the actual deployment of this capacity and the performance of AMD chips in large-scale AI workloads.
Story timeline
- Then Anthropic primarily relied on Nvidia for its artificial intelligence compute infrastructure.
- 02 Nov 2026 Sources detail Anthropic's compute deal spree: 3.5GW+ of expected capacity in 2026, diversifying beyond Nvidia's AI chips to Amazon, Google, and AMD, and more (New York Times)
- Now The company is diversifying its supply chain to include AMD and major cloud providers like Amazon and Google.
- Next Industry observers will watch for the actual deployment of this capacity and the performance of AMD chips in large-scale AI workloads.
The strongest doubt
Counterpoint from the sources
The New York Times report does not provide details on whether these diversified hardware choices will impact model performance or training efficiency compared to previous Nvidia-only setups.
For readers
Business
- Anthropic's diversification may signal a broader industry shift away from single-vendor dependency.
- The scale of 3.5 gigawatts indicates massive capital expenditure in the AI sector.
Policy
- The energy requirements for such capacity raise questions about data center power infrastructure.
- Supply chain diversification could impact semiconductor market dynamics and export controls.
Investors
- Increased demand for AMD chips may benefit the company's stock performance.
- Cloud providers like Amazon and Google are securing long-term contracts with major AI labs.
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Terms in this story
Compute capacity
The total processing power available for artificial intelligence tasks, often measured in gigawatts or teraflops.
Direct hardware procurement
Purchasing physical servers and chips directly from manufacturers rather than renting cloud services.
Gigawatt
A unit of power equal to one billion watts, used here to measure the scale of data center energy consumption.
Questions readers ask
How much compute capacity does Anthropic plan to have by 2026?
Sources tell The New York Times that Anthropic expects to reach more than 3.5 gigawatts of capacity by 2026.
Which hardware vendors is Anthropic working with?
The company is expanding its procurement to include both Nvidia and AMD for its hardware needs.
Are cloud providers involved in these deals?
Yes, the report states that Anthropic is securing capacity through partnerships with Amazon and Google.
Connected stories
Reading receipt
- Anthropic expects to reach over 3.5 gigawatts of compute capacity by 2026.
- The company is diversifying hardware suppliers to include AMD alongside Nvidia.
- New deals involve direct procurement and partnerships with Amazon and Google.
Still open: What specific AMD chip models are being integrated into Anthropic's infrastructure?
Still open: How will the performance of this mixed-hardware setup compare to previous Nvidia-only configurations?