Azure Exclusivity Is Over, Google Backs Anthropic with $40B, DeepSeek V4 Opens Its Weights

April 30, 2026

Executive Summary

Four developments defined this week. Google committed $40 billion to Anthropic. OpenAI launched GPT-5.5 and, one day after ending Azure exclusivity, made it available on Amazon Bedrock. And EU AI Act trilogue talks collapsed on April 29, leaving the August 2, 2026 high-risk compliance deadline in place.

Top Stories

1. Google Commits $40 Billion to Anthropic

Google commits $40B to Anthropic

Google is investing up to $40 billion in Anthropic: $10 billion at a $350 billion valuation, plus $30 billion contingent on performance targets. The deal includes 5 gigawatts of TPU compute capacity over five years. Anthropic retains its existing arrangements with Amazon ($100B cloud spending commitment) and CoreWeave, meaning it now has serious GPU commitments from two major cloud providers. Estimated current valuation is $800 billion or higher, with an IPO reported for late 2026.

The shift: AI investment is no longer about picking early-stage winners. It is about securing compute access. Google is protecting infrastructure alignment with Anthropic before AWS dominates the arrangement.

For enterprise teams, the practical implication is a clearer signal that Claude is a credible long-term platform with redundant infrastructure backing.

2. GPT-5.5 Launches; Azure Exclusivity Ends; Bedrock Access Follows in 24 Hours

GPT-5.5

OpenAI launched GPT-5.5 on April 23, with API access on April 24. It outperforms GPT-5.4, Gemini 3.1 Pro, and Claude Opus 4.5 on benchmarks, with stronger performance in agentic coding and multi-step reasoning. It is more token-efficient than the prior model. Pricing: $5/$30 per million input/output tokens, double GPT-5.4.

On April 27, Microsoft and OpenAI ended Azure exclusivity. Key changes: OpenAI can distribute on any cloud, the IP license becomes non-exclusive through 2032, the AGI clause is removed, and Microsoft no longer pays revenue sharing to OpenAI (OpenAI continues paying Microsoft through 2030, with a new cap).

On April 28, GPT-5.5 launched on Amazon Bedrock in preview, accessible via existing Bedrock APIs and IAM policies. General availability is expected within weeks.

For teams already on AWS: you can now run GPT-5.5 without Azure. For teams that delayed OpenAI adoption because of Azure lock-in: that barrier is gone. Benchmark token efficiency on real workloads before upgrading, given the 2x price increase.

3. DeepSeek V4 Preview: Open-Weight Models at 100x Lower Cost Than GPT-5.5

DeepSeek V4

DeepSeek released V4 Flash and V4 Pro in preview. Both are mixture-of-experts models with 1 million token context windows and open weights. V4 Pro has 1.6 trillion total parameters (49B active). V4 Flash has 284B total (13B active). Both are text-only at launch.

Performance: V4 Pro-Max beats GPT-5.2 and Gemini 3.0 Pro on certain reasoning benchmarks but trails current leading closed-source models by roughly one generation on knowledge tasks. Pricing: V4 Flash at $0.14/$0.28 per million input/output tokens. V4 Pro at $0.145/$3.48. Legacy aliases deprecate July 24, 2026.

V4 Flash is worth evaluating for cost-sensitive inference. Self-hosting is possible via open weights. Wait for GA before production deployments.

Quick Bytes

  • EU AI Act talks collapsed April 29 after 12 hours. The proposed Omnibus amendment would have delayed Annex III high-risk compliance to December 2027. The original August 2, 2026 deadline stands. Next session: approximately May 13. Do not plan around a delay that has not been confirmed. (The Next Web, IAPP)
  • Meta and Microsoft cut 20,000-plus jobs on April 24, citing AI automation in content moderation and software testing. Both stocks rose on the news. Total 2026 tech layoffs: 92,000-plus globally. (CNBC)
  • ICLR 2026 ran April 23-27 in Rio de Janeiro with a record 19,797 submissions. Agentic AI has overtaken LLMs as the top research category. (ICLR 2026)

Industry Impact

The bilateral cloud-model structure that defined enterprise AI since 2023 is breaking down. OpenAI is on AWS. Anthropic has compute on both Google and Amazon. Enterprise teams now have genuine cross-cloud optionality, but that means more deliberate decisions about which model, on which cloud, for which workload.

The DeepSeek V4 Flash pricing at $0.28/M output tokens versus GPT-5.5 at $30/M is a 100x gap. High-volume inference teams should evaluate whether every workload requires frontier pricing.

On compliance: the EU AI Act August 2 deadline is 95 days away with no confirmed delay. Conformity assessments and technical documentation take longer than 95 days from scratch. Start now.

About Azumo

Azumo builds AI and software systems for product teams moving from prototype to production. LATAM-based senior engineers, your time zone, 300-plus projects, SOC 2, 95% NPS.

Our team has production experience deploying OpenAI, Anthropic, and open-weight models across AWS, Azure, and Google Cloud, and we help teams build EU AI Act compliance infrastructure. If you are re-evaluating your cloud AI strategy or building agentic workflows, we are ready to help.

https://azumo.com/artificial-intelligence

Sources

Are You New to Outsourcing?
We Wrote the Handbook.

We believe an educated partner is the best partner. That's why we created a comprehensive, free Project Outsourcing Handbook that walks you through everything from basic definitions to advanced strategies for success. Before you even think about hiring, we invite you to explore our guide to make the most informed decision possible.