What Happened

Three significant developments landed in the AI space this week. OpenAI is internally testing a second-generation image model referred to as Image V2, an upgrade to the image generation capabilities already embedded in GPT-4o . Google pushed a new version of Jules, its AI- powered coding agent, now labeled Jules V2. Meanwhile, financial analysts are digging into the mechanics of OpenAI's recently closed $40 billion funding round, which values the company at approximately $340 billion — bringing cumulative disclosed funding to roughly $122 billion when earlier rounds are included.

Technical Deep Dive

OpenAI Image V2

OpenAI's Image V2 is being tested as an evolution of the image synthesis pipeline integrated into ChatGPT and the GPT-4o API. The original GPT-4o image generation launched in March 2025 and quickly became one of the most-used features on the platform, generating over 700 million images in its first month. Image V2 is expected to improve prompt adherence, text rendering accuracy inside images , and multi-object compositional consistency — three areas where the current model shows measurable degradation. No public release date has been confirmed .

Developers using the current image generation endpoint via the API can access it with:

curl https://api.openai.com/v1/images/generations \   -H "Authorization: Bearer $OPENAI_API_KEY" \   -H "Content- Type: application/json" \   -d '{     "model": "gpt-image-1",     "prompt": "A photorealistic render of a red  cube on a white table",     "n": 1,     "size": "1024x1024"   }'

Image V2 would slot into this same endpoint structure once released, likely under a versioned model identifier.

Google Jules V2

Jules is Google's asynchronous AI coding agent, announced at Google I/O 2025 and built on top of the Gemini 1.5 Pro model. Jules V 2 brings improved context handling for large codebases, better GitHub integration, and a reduced rate of hallucinated function calls in multi -file edit tasks. The agent operates by cloning a repository, spinning up a virtual machine, making code changes, and submitting a pull request — all without synchronous developer input.

Jules V2 is accessible via the Jules dashboard and integ rates directly with GitHub repositories. Setup requires connecting your GitHub account and granting repository-level permissions. Jules then accepts natural language task descriptions and resolves issues or implements features autonomously.

Competing tools in this space include Devin by Cognition AI, which targets a similar fully autonomous workflow, and GitHub Copilot Workspace, which takes a more collaborative, step-confirmation approach. Jules differentiates on speed and Google 's infrastructure backing.

Dissecting the $40B OpenAI Round

The funding round, led by SoftBank, closed at a $ 340 billion post-money valuation. Analysts note the structure includes a conditional tr anche: SoftBank's total commitment of up to $40 billion is split, with $10 billion funded at close and the remaining $30 billion contingent on OpenAI completing its transition from a c apped-profit LLC to a Delaware public benefit corporation by year -end 2025. Failure to restructure could reduce SoftBank's commitment.

Revenue figures provide the valuation context: OpenAI reported approximately $3.7 billion in annualized revenue as of late 2024, projecting $11 .6 billion for 2025 and $29.4 billion for 2026. At $340 billion, the implied 2025 revenue multiple is roughly 29x — aggressive but not unprecedented in high-growth infrastructure-layer companies. By comparison, Palantir trades near 35x forward revenue, though with prof itability that OpenAI currently lacks.

OpenAI reported a net loss of approximately $5 billion in 2024, primarily driven by compute costs. The company's compute bill is projected to exceed $7 billion in 2025, underscoring why the capital raise is operationally necessary rather than purely strategic.

Who Should Care

  • Product developers building image-heavy applications on the GPT-4o image API should monitor the Image V2 rollout. Improvements in text rendering specifically could unblock use cases like automated graphic design, document generation, and localized marketing asset creation.
  • Engineering teams evaluating AI coding agents should benchmark Jules V2 against Devin and Copilot Workspace on their specific stack . Jules V2's async PR-based workflow fits teams that want AI to handle background tasks without interrupting synchronous developer flow.
  • Enterprise buyers and procurement teams negotiating OpenAI contracts in 2025 should understand the structural dependency on the capped-profit transition. If the restruct uring stalls, it could affect OpenAI's capitalization timeline and indirectly influence pricing flexibility and roadmap commitments.
  • Investors and startup founders operating adjacent to OpenAI's core products should note the implied burn rate and capital requirements. A company spending $7B+ annually on compute is both a dominant buyer in the GPU market and a company whose pricing pressure on API consumers is structurally const rained.

What To Do This Week

  • If you're using GPT-4o image generation in production, document current prompt behaviors and output quality as a baseline. Image V2 may introduce behavioral changes that require prompt adjustments at release.
  • Sign up for Jules V2 access at labs.google.com/jules and test it on a low-risk open-source repository before evaluating it for internal codebases. Specifically test multi-file refactors and issue resolution tasks to assess context fidelity.
  • Review your OpenAI API spend against the company 's revenue trajectory. At $11.6B projected 2025 revenue with continued losses, price adjustments on API tiers remain a realistic near -term scenario — build budget flexibility accordingly.
  • For teams evaluating AI coding agents, run a structured comparison: assign the same GitHub issue to Jules V2, Devin, and Copilot Workspace and measure PR quality, time-to-completion , and required human correction rounds.