Colette Kress's first earnings call, and the numbers held
Yesterday I told you what to watch: NVIDIA's own forecast of about $108 billion, and the mood for interim CEO Colette Kress's first call in charge. The numbers came in last night, and they beat the forecast by a lot. Revenue was $115.8 billion for the quarter, up 103% from a year ago. The data center business, the part that sells the chips AI runs on, brought in $107.6 billion of that.
Kress also raised the outlook. She guided next quarter to $130 billion, give or take 3%, and raised next fiscal year's growth outlook to about 75%, up from the 70% forecast in August. Two footnotes: that $130 billion assumes no data center chip sales to China at all, and the quarter has an extra week in it. She called the new Vera Rubin chips "off to an extraordinary start," with nearly $20 billion in revenue in their first shipping quarter, the fastest ramp in company history.
The other question hanging over the call was whether founder Jensen Huang might come back. Board chair Harvey Jones answered it at the top: the board does not intend to reinstate him. So the picture this morning is a company that lost its founder a week ago and posted its biggest quarter ever anyway. If you are planning a big AI spend, here is the part that matters: the supply line held through a CEO change, and the company is telling buyers the growth continues from here.
A serious flaw in Java, found by an AI, patched the same day
This is the security story to act on today. Security firm Calif disclosed a flaw on Tuesday in OpenJDK, the open source code behind most Java software. Java runs a big share of the world's business systems, and the flaw sits in the part that decodes Base64, the format apps use to send login details and files as text. Calif named it Java4Shell.
Here is why it matters. In Calif's own testing, the flaw let them take over default setups of Jira, Confluence, Jenkins, and Keycloak, tools a lot of companies run, without logging in. Those results are Calif's, and they have not been independently confirmed. The flaw affects Java versions 17.0.6 and newer, including 21, 25, and 27, and only on servers with newer Intel or AMD chips. Machines running ARM chips are not affected. Oracle scores it 8.1 out of 10; Calif rates it 9.2 under a newer scoring system.
The good news? Oracle and Amazon shipped fixed builds the same day, and no attacks using it have been reported yet. Other Java vendors, including Microsoft and Red Hat, had not published their fixes as of Tuesday night, so check which Java you run and where you got it. And here is the part that should get your attention: Calif says its researcher found the flaw with an AI coding tool in about an hour, and had a first working exploit two hours later. Yesterday we talked about Google catching attackers who let AI run the whole attack. This is the same tooling, pointed the other way.
OpenAI shipped GPT-6.2, and the headline is cyber
OpenAI released two models on Tuesday: GPT-6.2 Astra, its strongest model for hard professional work, and GPT-6.2 Sol, the cheaper one for everyday tasks. Both are live in ChatGPT Work and Codex for paid plans, and in the API, the connection developers build on. Astra's API price dropped 20%, to $8 per million input tokens and $40 per million output. Sol stays at $2 and $10.
The rating is the part to slow down on. OpenAI says Astra is its most capable model ever for cybersecurity work, and both models hit the company's "Critical" threshold for cyber capability, so they shipped with tighter safeguards on higher-risk security help. Read that next to yesterday's stories, the California suit and Google's attack report. The models are getting better at the exact skills that cut both ways.
For your team, two practical notes. If you are on an Enterprise or Edu plan, an admin has to switch the new models on, so nothing changes until somebody approves it. And when the model under your agents changes, treat it like any other change: run your checks again before it touches real work. Cheaper and stronger is good news. Untested is not.
At Ignite, Microsoft wants to give every agent an ID
Yesterday we talked about Microsoft's plugin catalog, the approved list for Copilot add-ons. Today at Ignite, the other shoe dropped: finding the agents you did not approve. Microsoft's Entra system, the part that handles sign-ins and access, can now discover 60 kinds of local agents running on people's machines, including Claude Code, OpenAI's Codex, and GitHub Copilot CLI, and give each one an Agent ID, its own identity your IT team can govern. Most of this is in preview, which means it is rolling out, not finished.
Agent 365, Microsoft's control center for agents, also widened to cover other platforms, including Anthropic, OpenAI, and Salesforce, and Microsoft says more than 50 million agents are already managed through it. That number is Microsoft's own. The idea underneath is sound, whoever sells it to you: you cannot govern what you cannot see. An agent with an ID can be limited, watched, and shut off. An agent nobody knows about cannot.
If you run IT, the homework is the same one from yesterday, one step earlier. Before you argue about which agents are allowed, get the list of which ones are already running.
Also today
Anthropic prices its shares tonight, Wednesday evening, and starts trading Thursday on the Nasdaq under the ticker ANT. Bloomberg reported the order book closed Tuesday with demand close to seven times the shares on offer. Tomorrow we will know the price, and whether daily use carries the first day of trading.
The House is scheduled to hold its closed-door briefing today on the NVIDIA matter, the next date on the calendar from last week's coverage.
Google will now keep Gemini's temporary chats, and chats with history off, for about two months, up from 72 hours, TechCrunch reports. If your team uses temporary chats to keep things private, that window just got longer.
Microsoft also released its own coding models at Ignite, MAI Code 2 and MAI Code 2 Flash, and both are already inside GitHub Copilot, Reuters reports.
The big picture
So what ties today together? The systems held, and the checks are the story. NVIDIA's business ran straight through losing its founder. A flaw that could have been a disaster got found, disclosed, and patched in a day. And the agent argument moved from which ones to buy to which ones you can see. Before Friday, do two things: check the Java version on anything that faces the internet, and ask who owns the list of AI agents running on your machines. Those two lists are where the risk lives now.
SOURCES
- NVIDIA, Financial Results for Third Quarter Fiscal 2027 (Nov 17, 2026): https://www.sec.gov/Archives/edgar/data/0001045810/000104581026000088/q3fy27pr.htm
- CNBC, Nvidia earnings live updates (Nov 17, 2026): https://www.cnbc.com/2026/11/17/nvidia-nvda-earnings-report-q3-2027.html
- Censys, Nov 17 Advisory: OpenJDK Remote Code Execution (Java4Shell) (Nov 17, 2026): https://censys.com/advisory/cve-2026-114938/
- Tenable via Security Boulevard, Java4Shell: Frequently asked questions (Nov 17, 2026): https://securityboulevard.com/2026/11/java4shell-cve-2026-114938-frequently-asked-questions-about-openjdk-remote-code-execution-vulnerability/
- OpenAI Community, GPT-6.2 Astra and Sol are now available (Nov 17, 2026): https://community.openai.com/t/gpt-6-2-astra-and-sol-are-now-available/1417531/1
- Microsoft Entra Blog, Microsoft Entra news announcements from Microsoft Ignite November 2026 (Nov 17, 2026): https://techcommunity.microsoft.com/blog/microsoft-entra-blog/ai-is-changing-access-see-how-entra-helps-you-stay-ahead-/4556456
- Microsoft Copilot Blog, Copilot at Microsoft Ignite (Nov 17, 2026): https://www.microsoft.com/en-us/copilot/blog/2026/11/17/copilot-at-microsoft-ignite-expanding-whats-possible-with-intelligence-built-for-work/
- CNBC, Anthropic IPO guide (Nov 16, 2026): https://www.cnbc.com/2026/11/16/anthropic-ipo-date-price-range.html
About this brief: This brief was researched and drafted by my personal AI assistant and approved by me before publishing. Product details change fast; check primary sources before you build on them.
Want more with the Tech Simplified crew? Subscribe on Substack.
Get the week's briefs in one email. The Briefcase packs Monday to Friday's AI Briefs into one Sunday email.
Follow me on LinkedIn for project updates, ideas for using AI, and lessons you can bring to your own work.