- OpenAI
- 1. ChatGPT rolls out new imagegen model and creation tool
- 2. OpenAI releases best model yet for professional work
- NVIDIA
- 3. NVIDIA acquires open-source workload manager SchedMD
- Google
- 4. Google upgrades Translate with Gemini-powered features
- Fundraises
- 5. Databricks raises $4B+ at a $134B valuation, AI products hit $1B+ run-rate
- 6. OpenAI-backed Chai Discovery raises $130M in Series B at a $1.3B valuation
- In other news
- 2 new AI tools to try
- Recommended reading
The CFO AI Digest: December 17, 2025
OpenAI continues to double down on ChatGPT, launching new models for professional work and for image generation. NVIDIA makes a major acquisition, and Google leans into its undefeated distribution advantage with Translate upgrades.
Plus, Databricks’ jaw-dropping valuation and Chai Discovery’s raise with a promise for the next frontier in AI-led drug discovery.
Here’s the rundown:
OpenAI
1. ChatGPT rolls out new imagegen model and creation tool
OpenAI released GPT Image 1.5, a major upgrade to its image generation capabilities in ChatGPT and the API. The new model enables quicker, more precise image edits, preserving key details like lighting, facial likeness, and composition, while following instructions more reliably. It supports complex edits like swapping styles or objects, renders dense text better, and generates images up to 4x faster than before, OpenAI says. A new "Images" sidebar in ChatGPT offers preset styles and trending prompts, simplifying creative workflows. Early users like Wix, Canva, and Figma report strong results for marketing, ecommerce, and design use cases. API pricing is 20% lower than the previous version.
CFO takeaway: A week after integrating Adobe's creativity apps into ChatGPT, OpenAI has rolled out its own image generation model as it potentially both competes against, and partners with, creative incumbents. ChatGPT Images, as an OpenAI design director put it, distinguishes itself from other models by focusing on exploring, tweaking, and iterating images without breaking what already works. This fits the revision-heavy nature of professional design, and may push companies to consolidate creative software spend into ChatGPT, moving OpenAI beyond enterprise knowledge work deeper into creative production.
2. OpenAI releases best model yet for professional work
GPT‑5.2 launched last week, establishing itself as OpenAI’s most advanced model yet for professional work and long-running agents. It matched or outperformed industry professionals on 70.9% of business tasks, such as creating spreadsheets and presentations, across 44 occupations, while working 11 times faster and at under 1% of the cost. On junior investment banking tasks like leveraged buyout analyses, it scored 68.4%, up from 59.1% in GPT‑5.1. The model has improved long-context understanding for analyzing lengthy documents like contracts and reports, and enhanced tool-calling capabilities that let it autonomously select and use external tools and APIs to complete complex workflows. GPT‑5.2 is available in three versions: Instant, Thinking, and Pro, via ChatGPT and OpenAI’s API.
CFO takeaway: GPT‑5.2 is OpenAI’s strongest case yet for enterprise adoption, with an explicit goal to unlock more economic value for people. The company is also signaling its plans to automate tasks and not just enhance productivity. By benchmarking the model against professional experts on common business tasks, OpenAI is showing that ChatGPT can deliver better or comparable output at a fraction of the cost and time. This aligns with OpenAI’s Project Mercury plans, which aims to replace junior investment banker tasks.
NVIDIA
3. NVIDIA acquires open-source workload manager SchedMD
NVIDIA acquired SchedMD, the company behind Slurm, the leading open-source job scheduler for high-performance computing (HPC) and AI. Terms of the deal are not public. Slurm is used on more than half of the world’s top 100 supercomputers to manage AI training and inference workloads. NVIDIA will continue to develop and distribute Slurm as an open-source, vendor-neutral software. SchedMD’s customers include cloud providers, AI companies, and research labs across industries such as autonomous driving, healthcare, energy, and financial services.
CFO takeaway: By acquiring the scheduler that manages workloads on over half of the world’s top supercomputers, NVIDIA gains control over how high-value, limited compute capacity is allocated, even on non-NVIDIA hardware. Its commitment to keeping Slurm open-source maintains ecosystem trust but also ensures it can optimize the stack for its own chips first. NVIDIA becomes the orchestration layer for AI compute, making its infrastructure harder to displace. This move reinforces a broader trend: the AI stack is consolidating around a few players who control key bottlenecks.
4. Google upgrades Translate with Gemini-powered features
Google rolled out a major upgrade to Google Translate, powered by Gemini. Text translations deliver greater accuracy and context awareness, translating phrases based on their intended meaning rather than literal word-for-word conversion, especially for idioms, slang, and local expressions. The improvement covers nearly 20 languages including Spanish, Hindi, and Chinese. Google also launched a beta feature that translates live audio from conversations or media into users' headphones in real time while preserving the speaker's tone and cadence across over 70 languages. The app's built-in language learning feature has added consecutive-day tracking and expanded to nearly 20 new countries including Germany, India, and Sweden.
CFO takeaway: Voice AI continues to prove its commercial viability. Last week, we discussed how ElevenLabs has demonstrated market demand for natural-sounding, reliable voice technology. This week, Google is leveraging its unmatched distribution to bring high-context translation into Translate's billions-strong user base. By preserving local meaning, tone, and dialect, Google builds on its already well-established trust globally and maintains product-market fit across languages and regions. For global teams, this will reduce interpreter costs, simplify training, and ease cross-border collaboration.
Fundraises
5. Databricks raises $4B+ at a $134B valuation, AI products hit $1B+ run-rate
Databricks announced a $4B+ Series L funding round led by Insight Partners, Fidelity, and J.P. Morgan, valuing the company at $134B. The company had a revenue run-rate of more than $4.8B in Q3, up 55% YoY, with positive free cash flow, and $1B-plus run-rates from both AI and data warehousing. The new capital will accelerate development of three AI-focused products: Lakebase (a cloud database optimized for AI workloads), Databricks Apps (a platform for building and deploying AI applications), and Agent Bricks (for building production AI agents on enterprise data). The funding will also support AI acquisitions, deepen AI research, and provide employee liquidity.
CFO takeaway: Databricks is betting that the next wave of enterprise software will be built on top of company data, or what it calls “data-intelligent applications.” This reflects the broader market shift: OpenAI embedding company knowledge into ChatGPT, Microsoft folding workflows into Copilot, and enterprises pushing for fewer tools with deeper integration. Databricks’ valuation puts it in rare company and signals that investors see lasting value in platforms that are embedded deep in the enterprise data stack.
6. OpenAI-backed Chai Discovery raises $130M in Series B at a $1.3B valuation
Chai Discovery, a company that uses AI to predict and reprogram molecular interactions to enable computational drug design that compresses multi-year discovery timelines into weeks, raised a $130M Series B at a $1.3B valuation. The round was co-led by Oak HC/FT and General Catalyst. The funding follows two milestones. First, the launch of Chai 2, which achieves double-digit experimental success rates in antibody design, a 100x improvement over previous computational methods. Second, the company’s announcement that its latest models are now capable of designing molecules for historically challenging targets. The funding will support R&D and commercialization of Chai’s molecular design platform.
CFO takeaway: The Google-Yale and Chai Discovery stories show that AI is beginning to reshape the economics of drug discovery. What used to take years of trial-and-error in the lab identifying and ranking viable drug candidates can now happen in weeks using AI models. Chai’s $1.3B valuation on relatively modest funding reflects investor belief that platforms that use AI to design more targeted, higher-quality drug candidates will capture an increasing share of biopharma R&D budgets.
In other news
- OpenAI eliminates ‘vesting cliff’ for new employees to attract top AI talent (The Wall Street Journal)
- Google names Amin Vahdat chief technologist for AI infrastructure (Semafor)
