Google TPUs, DALL-E Deprecation, Clinical AI, and the RAISE Act: A Dense Week for Enterprise AI

April 23, 2026

Executive Summary

Google Cloud Next 2026 dominated this week. Over three days starting April 22, Google announced eighth-generation TPUs, a full enterprise agent platform built on Gemini, and a $750 million partner fund to accelerate agentic deployments. The conference set a clear direction: Google is betting its cloud business on agents, and it wants its 120,000-partner ecosystem to help build them. Two separate OpenAI launches also landed: ChatGPT Images 2.0 (with native reasoning) on April 21, and ChatGPT for Clinicians on April 23. On the regulatory side, New York's amended RAISE Act is now finalized, creating new transparency requirements for frontier AI developers ahead of its January 2027 effective date.

Top Stories

1. Google Cloud Next 2026: New TPUs, Gemini Enterprise Agent Platform, $750M Partner Fund

Google Cloud Next 26

Google Cloud Next 2026 (April 22-24, Las Vegas) produced the most substantial single-week enterprise AI announcement of the year so far. Three interconnected pillars define the strategy.

8th-Generation TPUs. Google announced two new chips designed for distinct workloads in the agentic era. TPU 8t is built for training: 9,600 chips in a single superpod, delivering 121 exaflops of compute and two petabytes of shared high-bandwidth memory. TPU 8i is built for inference and reinforcement learning: it triples on-chip SRAM to 384MB, increases high-bandwidth memory to 288GB, and delivers 80% better performance per dollar than the previous generation. The focus on low-latency inference is a direct response to the demands of real-time agentic workflows.

Gemini Enterprise Agent Platform. Google introduced a full-stack agent management layer inside Gemini Enterprise, including an Agent Designer for building trigger- and schedule-based agents, an Inbox for managing agent activity, long-running agents for complex multi-step processes, and a Skills library for recurring tasks. Partners including Adobe, Atlassian, Salesforce, ServiceNow, Workday, and Deloitte are building on the platform. Google also announced that its first-party models now process 16 billion tokens per minute via direct API, up from 10 billion last quarter.

$750 Million Partner Ecosystem Fund. Google committed $750 million to accelerate agentic AI development across its 120,000-member partner ecosystem. Deloitte simultaneously launched a dedicated Google Cloud Agentic Transformation Practice. The fund is paired with a cross-cloud Lakehouse and Knowledge Catalog under what Google is calling the Agentic Data Cloud.

Business Impact: For engineering and product teams evaluating cloud infrastructure for agent deployments, Google just made its platform significantly more competitive. The TPU 8i inference numbers are particularly relevant: agents that need to run millions of concurrent sessions need low-latency, cost-effective inference at scale. The Gemini Enterprise Agent Platform also provides the governance layer, including agent activity monitoring, that was identified last week as the primary gap in enterprise agentic rollouts.

2. OpenAI Launches ChatGPT Images 2.0 With Native Reasoning

ChatGPT Images 2.0

OpenAI released ChatGPT Images 2.0 on April 21, built on the new gpt-image-2 model. It is OpenAI's first image model with native reasoning, meaning it processes instructions before generating output rather than mapping prompts directly to pixels. The practical result is meaningfully better performance on tasks where earlier image models struggled: accurate small text in any language or script, dense compositions with 100-plus objects, UI element rendering, and stylistic consistency across a batch of images.

Key specifications: 2K resolution, aspect ratios from 3:1 to 1:3, up to eight coherent images per prompt with consistent characters and objects across the batch. The editing endpoint supports inpainting and outpainting via mask images. API pricing: $5 per million text input tokens, $8 per million image input tokens, $30 per million image output tokens. DALL-E 2 and DALL-E 3 are being deprecated and retire May 12, 2026 — developers on either endpoint need to migrate before that date.

Direct API access was not open at launch. It is expected in early May 2026.

Business Impact: For product teams building content pipelines, marketing automation, or document generation, gpt-image-2 closes the gap between what AI can generate and what brand teams will actually approve. The reasoning layer is the meaningful change: it reduces prompt iteration cycles on complex compositions, which is where earlier models required the most human correction. Teams running DALL-E 3 in production have a hard deadline of May 12 to migrate.

3. OpenAI Launches ChatGPT for Clinicians With Free Access for Verified Healthcare Providers

ChatGPT for Clinicians

OpenAI announced ChatGPT for Clinicians on April 23, offering free access to verified U.S. physicians, nurse practitioners, physician assistants, and pharmacists. The product targets three clinical use cases: care consultation, medical writing and documentation, and medical research. Features include trusted clinical search with citations from peer-reviewed sources, a deep research mode across medical journals, reusable skills for recurring workflows (referral letters, prior authorization, patient instructions), and HIPAA-compliant configuration options.

OpenAI simultaneously released HealthBench Professional, an open benchmark for evaluating AI performance on authentic clinical tasks across the same three use cases. According to a 2026 American Medical Association survey, 72% of physicians now use AI in clinical practice, up from 48% the prior year.

Business Impact: The free access model is a clear adoption-first play in a market where clinical AI products typically carry enterprise licensing costs. For healthcare organizations, the relevant questions are credential verification at scale, EHR integration, and workflow fit. The introduction of HealthBench Professional also signals that OpenAI intends to define the evaluation standard for clinical AI, which shapes how competitors and buyers benchmark alternatives.

4. New York Finalizes RAISE Act Amendments, Setting January 2027 Deadline for Frontier AI Developers

RAISE Act Amendments

New York Governor Kathy Hochul signed amendments to the Responsible AI Safety and Education (RAISE) Act on March 27, 2026. The amendments were finalized and published by legal counsel in April, clarifying compliance requirements ahead of the January 1, 2027 effective date.

The amended law applies to "large frontier developers," now defined by a revenue threshold matching California's SB-53: companies with more than $500 million in annual gross revenue. Covered developers must publish a frontier AI framework covering catastrophic risk thresholds and mitigations, third-party evaluation practices, cybersecurity controls, internal governance structures, and critical safety incident reporting. If a safety incident occurs, developers have 72 hours to notify the New York Department of Financial Services, which now holds primary enforcement authority. The DFS is authorized to issue rules and enforce the act.

The amendments align New York's requirements with California's Transparency in Frontier Artificial Intelligence Act, reducing compliance complexity for organizations subject to both laws.

Business Impact: Any AI developer or company using externally developed frontier models in New York with more than $500 million in annual revenue should treat January 1, 2027 as a planning deadline, not a compliance deadline. The documentation requirements, including risk frameworks and third-party evaluations, take months to build. Legal and compliance teams that have not started this process should start now.

Quick Bytes

  • Google announced that 75% of new code inside Google is now written by AI and reviewed by human engineers, up from an earlier reported figure. (Google Cloud Next 2026 keynote, April 22)
  • OpenAI acquired personal finance AI startup Hiro Finance on April 20, its seventh known acquisition in 2026. The Hiro team joins OpenAI; the Hiro product shuts down. (various reports, April 2026)
  • Q1 2026 set global venture records: four of the five largest VC rounds in history closed in Q1, with OpenAI ($122B), Anthropic ($30B), xAI ($20B), and Waymo ($16B) accounting for 65% of global venture investment in the quarter. (Crunchbase, April 2026)
  • Microsoft released three new foundational models in early April: MAI-Transcribe-1 (speech recognition), MAI-Voice-1 (text-to-speech with custom voice cloning), and MAI-Image-2 (image generation). These are priced below equivalent Google and OpenAI offerings. (Microsoft AI, April 2026)

Industry Impact Analysis

Google Cloud Next 2026 makes the enterprise platform picture clearer. Google is building vertically: custom inference silicon (TPU 8i), an agent orchestration and governance layer (Gemini Enterprise Agent Platform), a data layer (Agentic Data Cloud), and a security layer (Agentic Defense). Organizations that standardize on Google Cloud get a more integrated stack for agentic deployment than was available six months ago.

OpenAI's two launches this week show different growth vectors. ChatGPT Images 2.0 targets developers and creative production pipelines with a technically stronger model. ChatGPT for Clinicians targets a vertical market where AI adoption is accelerating but dedicated tools are still scarce. Both follow OpenAI's consistent pattern: release free or competitively priced access to expand installed base, then monetize through enterprise tiers and API consumption.

The RAISE Act finalization is a practical reminder that regulatory timelines are moving faster than most enterprise governance programs. Nine months to January 1, 2027 is enough time to build a compliance framework, but not if the work starts in November.

About Azumo

Azumo builds AI and software systems for product teams that need to move from prototype to production. Our LATAM-based senior engineers work in your time zone across 300+ projects, with SOC 2 certification and a 95% NPS.

The announcements this week, particularly the Gemini Enterprise Agent Platform and the maturation of agent infrastructure at Google Cloud, are the kinds of developments where teams need partners who have already worked through the integration layer. Our team has production experience across agentic orchestration, enterprise AI pipelines, and the governance tooling that keeps deployments stable.

If you are evaluating agent infrastructure, planning an OpenAI migration, or mapping your RAISE Act compliance path, we are ready to help.

Learn more: 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.