It’s a week heavy on OpenAI news. ChatGPT embeds itself further into consumer purchase behavior, ChatGPT Enterprise rolls out to Accenture professionals, and Sam Altman reportedly declares a code red in response to Gemini 3’s power.

Plus, Anthropic acquires an all-in-one JavaScript, and AWS makes big announcements at its annual conference.

Here’s the rundown:

OpenAI

1. ChatGPT adds shopping research to guide purchase decisions

ChatGPT now includes shopping research, a visual interface to help with complex purchase decisions that regular ChatGPT responses aren’t optimized for. Users can access the feature by asking a question, like “I need a laptop under $1,000,” or by selecting it from the tools menu (+). The tool asks follow-up questions, pulls real-time product data from the web, and generates a personalized buyer’s guide with top options, comparisons, and links to retailers. It adapts to user feedback and uses memory to further tailor results, like factoring in a user’s preference for gaming when recommending laptops. It performs especially well in detail-heavy categories like electronics, home and kitchen, beauty, and outdoor gear. The feature is available across all ChatGPT plans, with nearly unlimited usage offered through the holidays.

CFO takeaway: OpenAI continues to commercialize consumer use cases with this launch, which follows its Instant Checkout rollout and PayPal wallet integration. Detailed purchase needs, like budget and product features, are rich intent data that could support advertising and eventually help drive sales through Instant Checkout. This feature also deepens ChatGPT’s reinforcement learning, making recommendations more personalized to the user to increase conversion rates.

Read more

2. OpenAI partners with Accenture to scale ChatGPT Enterprise and AI agents

Accenture is giving tens of thousands of its employees access to ChatGPT Enterprise to use across consulting, operations, and delivery work. Accenture employees will complete OpenAI Certifications, the program’s largest upskilling effort yet. As part of the partnership, OpenAI and Accenture are launching a flagship client program to help joint clients integrate OpenAI’s agentic capabilities across functions like customer service, supply chain, finance, and HR. For example, Accenture will use OpenAI’s AgentKit to help clients design and deploy custom AI workflows faster.

CFO takeaway: Similar to Anthropic’s Claude rollout for Deloitte, OpenAI is scaling enterprise adoption with Accenture. With this partnership, Accenture becomes both a customer, by deploying ChatGPT Enterprise, and a distribution channel, by implementing AI agents for joint clients. The focus on agentic capabilities specifically positions OpenAI to capture recurring revenue as enterprises increasingly start experimenting with, and adopting, agents.

Read more

3. Altman declares 'code red,’ doubles down on ChatGPT improvement

OpenAI CEO Sam Altman has directed the company to prioritize improving ChatGPT, delaying plans for advertising and AI agents, as competition intensifies from Google's Gemini, according to a report from The Information. Altman called this a "critical time for ChatGPT" as Google's chatbot has surged to 650 million monthly users, though still trailing ChatGPT's 800 million weekly users. The company will focus on personalizing the chatbot, improving image generation, and boosting speed and reliability. OpenAI plans to release a new reasoning model next week that it claims will outperform Google's Gemini 3.

CFO takeaway: OpenAI’s renewed focus on ChatGPT is aligned with the new feature and partnership outlined above. OpenAI has stood out for its broad, aggressive investments — including on data centers and chips — as covered in our comparison on revenue projections with Anthropic. Altman’s declaration now marks both a sharp focus and growing competitive pressure, a reminder that even the market leader must make tradeoffs to keep up in the AI race.

Read more

4. OpenAI takes ownership stake in Thrive Holdings, embeds AI capabilities into service ops

OpenAI has taken an ownership stake in Thrive Holdings, a VC that owns and operates businesses in the services industry. As part of the partnership, OpenAI will embed its research, product, and engineering teams into Thrive Holdings’ companies. The project starts with accounting and IT services, sectors that are high-volume, rules-based, and workflow-heavy, making them especially well-suited for AI-driven efficiency gains. The goal is to build a repeatable model for applying AI across established industries and enable domain experts in these fields to adopt AI as a native tool.

CFO takeaway: Instead of selling models via API, OpenAI is embedding its teams inside operating businesses to build and apply AI directly within day-to-day workflows. As the CEO and founder of Thrive Capital and Thrive Holdings wrote, the next wave of AI transformation will happen "from the inside-out," using owned portfolio companies as the testing ground to build industry-specific models with proprietary data. For OpenAI, this is a way to demonstrate its effectiveness in traditional sectors while also refining its AI through real-world feedback.

Read more

Anthropic

5. Anthropic acquires Bun to strengthen Claude Code

Anthropic has acquired Bun, a high-performance tool that helps engineers work with JavaScript, one of the web's most popular coding languages. Bun combines several developer tools into an end-to-end developer platform that helps with everything from runtime to testing. Anthropic has already been using Bun to support Claude Code's infrastructure behind the scenes and says the acquisition will further improve performance, reliability, and deployment speed for enterprise customers. Bun will remain open-source and MIT-licensed.

CFO takeaway: Bun's founder notes that Bun was already supporting Claude Code. By acquiring Bun, Anthropic brings key infrastructure in-house and eliminates a third-party dependency. The acquisition is yet another signal to competitors on Anthropic’s strategic focus: coding capabilities and the technical infrastructure that powers them.

Read more

NVIDIA

6. NVIDIA and Synopsys partner on AI-powered engineering

NVIDIA and chipmaker Synopsys have announced a multi-year strategic partnership to accelerate engineering and design across industries including semiconductors, aerospace, and automotive. The collaboration will integrate NVIDIA's AI and GPU computing power with Synopsys' engineering software to help R&D teams iterate faster. Key initiatives include accelerating Synopsys applications using NVIDIA's CUDA technology — a platform that allows software to harness GPU processing power — advancing autonomous design workflows, creating digital twins for virtual testing, and moving these tools to the cloud.

CFO takeaway: This partnership addresses time-to-market, a key resource constraint in chip design, automotive, and aerospace. NVIDIA's GPU computing power will accelerate Synopsys' simulation software, allowing engineers to test and verify products faster than traditional CPU-based workflows. With this partnership, NVIDIA also embeds itself into software tools engineers use daily, expanding its presence across the engineering stack.

Read more

Amazon

7. AWS debuts new chip and AI agents at re:Invent

At its re:Invent conference, AWS launched a wave of AI products spanning chips, models, agents, and infrastructure. It rolled out its Trainium3 next-gen AI chip, now generally available, and previewed Trainium4, predicted to improve performance further. It introduced its first frontier agents for development, security, and ops. It also expanded the Nova model family with four new models and launched Nova Forge, a framework for customizing models using proprietary and Amazon-curated data. Finally, it introduced AI Factories, which bring Trainium hardware and Bedrock, AWS’s platform for running foundation models, into customer data centers.

CFO takeaway: As a Wall Street Journal reporter noted following a conversation with the AWS CEO, the cloud services provider is leaning into agentic AI by leveraging its infrastructure, memory architecture, and model orchestration to power agents that can run for days. Nova Forge also opens a new path for enterprise model customization through mid-training, which lets companies resume training from earlier model checkpoints. Amazon says this can deliver 40% to 60% performance gains over traditional fine-tuning.

Read more

See you next week.

In other news

1. Sundar Pichai projects 10-year timeline for space-based data centers (Fortune)
2. Expedia taps ex-Google VP as its first Chief AI Officer (Skift)

2 new AI tools to try

1. Runway Gen-4.5 (Runway)
2. DeepSeek V3.2 (Hugging Face)

Recommended reading

1. The AI power boom has just begun (Coatue)
2. Black Friday data shows online sales strong, store results mixed (Forbes)
3. Alignment research blog (OpenAI)