GenAI Chronicle
2010
2011
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2026

April

6 events

Editors' summary

Cut off in Washington, Anthropic draws more money from the giants

On 24 April, Google announced an investment of up to $40 billion in Anthropic, with 5 gigawatts of compute alongside it. On the 20th, Amazon put in a further $5 billion and Anthropic committed to spending over $100 billion with AWS across ten years. Money from the largest players kept flowing to a company the federal government had cut off in February.

On the 22nd, Google announced its eighth-generation TPUs, splitting training and inference across separate chips.

On the 23rd, OpenAI released GPT-5.5, stronger at carrying out work on a PC as an agent. Anthropic had released Claude Opus 4.7 on the 16th.

This block is written by the editors. It is kept separate from the sourced record below.

Record6 events
  1. 7
    ResearchAnthropic

    Anthropic withholds Mythos Preview and starts Project Glasswing

    Anthropic published its assessment of Claude Mythos Preview and said it would not make the model generally available. The unreleased model had found vulnerabilities in every major operating system and every major web browser and had written working exploits for them without human guidance, among them a 27-year-old bug in OpenBSD, a 16-year-old flaw in FFmpeg, and a 17-year-old unauthenticated remote code execution in FreeBSD's NFS server (CVE-2026-4747). More than 99% of the vulnerabilities it had found were still unpatched, the company said, which was why it would not ship the model. Alongside the assessment it launched Project Glasswing, a joint effort to secure critical software, with AWS, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA and Palo Alto Networks. Around fifty organisations were given access, backed by $100 million in model usage credits and $4 million donated to open-source security groups. At the time, deciding not to ship a frontier model because of what it could do, and handing it only to the defenders, was a new shape of decision.

  2. 16
    ModelAnthropic

    Claude Opus 4.7 targets the hardest engineering work

    Anthropic released Claude Opus 4.7, foregrounding gains in advanced software engineering and especially on the hardest tasks — framing it as a model you can hand your toughest coding work to. It handled long-running, complex tasks with more consistency, followed explicit instructions more literally, and gained stronger file-system-based memory for work spanning multiple sessions. Image input grew to 2,576 pixels on the long edge, about 3.75 megapixels and more than triple prior Claude models, and a new xhigh effort level gave finer control over the tradeoff between reasoning depth and latency. Internal benchmarks showed a 13% lift in coding and 21% fewer errors in document reasoning. Pricing matched Opus 4.6 at $5 and $25 per million tokens.

  3. 20
    BusinessAnthropic

    Amazon adds $5 billion; Anthropic commits over $100 billion to AWS

    Amazon announced an additional $5 billion investment in Anthropic, with up to $20 billion more possible against commercial milestones, on top of the $8 billion it had already put in; the new money came at Anthropic's $380 billion valuation. In return Anthropic committed to spend over $100 billion with AWS over the next decade and to take up to 5 gigawatts of AWS Trainium capacity for training and serving Claude. With the spending commitment an order of magnitude larger than the investment, the deal made plain the shape of these arrangements: cloud providers converting stakes in AI companies into demand for their own infrastructure.

  4. 22
    ProductGoogle

    Google splits its eighth-generation TPUs into training and inference chips

    At Google Cloud Next in Las Vegas, Google announced its eighth generation of TPUs, split for the first time into distinct chips for training and serving: TPU 8t for training and TPU 8i for inference and reinforcement learning. TPU 8t connects up to 9,600 chips and was said to improve training price-performance by up to 2.8x over the seventh-generation Ironwood. TPU 8i puts 1,152 chips in a pod with triple Ironwood's on-chip SRAM, optimized for running autonomous AI agents in parallel. Both target TSMC's 2nm process, with availability to customers outside Google aimed at late 2027. With Anthropic contracting for up to a million TPUs and Google investing up to $40 billion in that company, TPUs had become not just the substrate for Google's own models but a business set against NVIDIA.

  5. 23
    ModelOpenAI / Greg Brockman

    GPT-5.5 released with stronger agentic computer-work capabilities

    OpenAI released GPT-5.5, billed as its smartest and most intuitive model to use. The company emphasized agentic work capabilities: writing and debugging code, researching online, analyzing data, creating documents and spreadsheets, operating software, and moving across tools until a task is finished. It rolled out to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex, with GPT-5.5 and GPT-5.5 Pro becoming available in the API the next day, April 24.

  6. 24
    BusinessAnthropic / Google

    Google to invest up to $40 billion in Anthropic

    Google announced it would invest up to $40 billion in Anthropic, starting with $10 billion at a $350 billion valuation and releasing the remaining $30 billion as performance milestones are met. Structured as cash plus compute, the deal has Google Cloud delivering five gigawatts of capacity over five years with room for several more later. It came four days after Amazon's $5 billion investment with up to $25 billion total — underscoring again that Google, which builds Gemini in direct competition with Claude, was simultaneously pouring money into its rival. Anthropic's run-rate revenue had gone from $1 billion at the end of 2024 to $9 billion at the end of 2025 and $30 billion by early April 2026.

↑ 2026