![[Google Cloud Next '26 Recap #4] Live Report from the Two Keynotes](https://media2.dev.to/dynamic/image/width=1000,height=420,fit=cover,gravity=auto,format=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fwdscx7wfkq5uqkli1n5d.jpeg)
This is the fourth post in my Google Cloud Next '26 (Las Vegas) recap series. You can find the...
This is the fourth post in my Google Cloud Next '26 (Las Vegas) recap series.
You can find the previous posts here 👇
This time, I'd like to share a live report on the two keynotes from Next '26:
…with a focus on the atmosphere of the venue itself. When you attend a tech conference, the keynotes are the one thing you simply can't skip.
You may have already seen them, but the keynotes are also available on YouTube:
According to Next's Session Types, the definition of Keynotes is as follows:
Join Google Cloud and industry leaders as they make big announcements, showcase the latest products and customer successes, and set the stage for everything else at the event.
In other words, Keynotes are the foundational sessions where major announcements, the latest products, and customer stories all land at once — setting the direction for the entire event.
The keynote venue, as in past years, was the Michelob ULTRA Arena inside Mandalay Bay. This was my first time experiencing it in person, and with a capacity of 12,000 people, the sheer scale of the place hit me the moment I walked in.
Watching the seats gradually fill up before the show, with the lighting and sound design slowly building the atmosphere, gave me a kind of anticipation you simply can't get from a livestream.
Before the Opening Keynote kicked off, there was a music and visuals performance segment. The visuals appeared to be generated by AI, and an operator was switching between visual patterns by giving instructions through hand gestures in time with the music — a very unique production.
Even before the keynote actually started, the warm-up alone made it clear: "We're about to enter the era of agents."
The contents of the Opening Keynote have already been covered widely, so I'll just touch on the highlights here.
The industry is undergoing a major shift from generative AI to the "Agentic Era", where AI agents capable of autonomously reasoning, acting, and scaling are now being deployed across enterprises.
Google's strengths were framed as a fully owned full stack, end to end:
Custom silicon (Ironwood / 8th-gen TPU) → Frontier models (Gemini) → Cloud platform → Enterprise distribution channel (Workspace with 3 billion users)
The emphasis was on the fact that Google owns the entire stack from chip to app, in-house.
Vertex AI has been rebranded to the Gemini Enterprise Agent Platform. The employee-facing AI assistant Agentspace is being folded in as well, consolidating everything into a single product called Gemini Enterprise — positioned as a platform that handles agent build, scale, governance, and optimization end-to-end.
A core business interface that lets even non-technical users build and use agents through natural language.
A complete redesign of the data foundation.
Integrating technology from the Wiz acquisition, autonomous Red / Blue / Green agents detect and remediate vulnerabilities at machine speed.
Agents are being embedded across Gmail / Docs / Sheets / Drive / Meet / Chat.
The moment that stood out to me the most was the segment titled "State-of-the-art models in Gemini Enterprise Agent Platform", where various models were introduced.
Alongside Google's four models — Gemini 3.1 Pro / Gemini 3.1 Flash image / Lyria 3 Pro / Veo 3.1 Lite — Claude Opus 4.7 was featured as one of the Anthropic Models as well.
There may have been an element of wanting to avoid coming across as a closed ecosystem, but personally, I couldn't help but read it as a sign that Google is genuinely acknowledging the strength of Opus 4.7's model performance.
Before the Developer Keynote, there was the same kind of music and visuals performance segment as the Opening Keynote. Same format, but it still had that "Oh, here we go again" feel of building anticipation.
The Developer Keynote has also been covered extensively on the official Google Cloud blog and elsewhere, so I'll just hit the highlights here.
The Developer Keynote was a developer-focused session centered around live demos and live coding.
The overarching theme was a multi-agent system that plans and simulates a fictional Las Vegas marathon, walking through the journey from the very first prototype all the way to production in a clear, story-driven progression.
For the record, the Las Vegas Marathon Simulator was made up of three agents:
The Developer Keynote presented the following 7 demos:
The flow — "build agents → make them collaborate → give them memory → debug them → handle infrastructure declaratively → distribute them no-code → secure them" — covered the entire agent development journey end-to-end. Even from a developer's perspective, it was packed with takeaways.
What I really appreciated was that all the solutions were already published on GitHub, with the demos available as Codelabs. The QR codes shown on stage during the keynote pointed to:
Having a clean path of "watched the keynote → curious → let me try it myself right now" set up like this is genuinely one of the best things a developer-focused conference can do.
By the way, a couple of slightly odd things happened during the Developer Keynote.
The person sitting diagonally in front of me — apparently some kind of insider — started taking selfies with their phone before the keynote began. Their phone had a very bright dedicated phone light attached, which was honestly a bit dazzling from where I was sitting.
Then, near the end of the keynote during one of the demos, the person sitting directly in front of me turned around, stared right at me for a moment, and then silently walked away. I'm not sure what they wanted…?
Apparently this kind of thing is fairly common at tech conferences, so I'll just leave it be.
After attending both keynotes in person, here's what I came away with:
"You can just watch the keynotes on YouTube" is technically true — but the value of being there in person is genuinely something else, and these sessions reminded me of that all over again.
To be continued in #5 (the final post). [Google Cloud Next '26 Recap #5] How I Prepared for the Trip — and How It Actually Played Out
aiMost of us have seen a coding agent fail to complete a task we know it can do. We just don't...
googlecloudWhen building Generative AI applications, developers often encounter a massive bottleneck: sequential...
discussI’ve been thinking about sharing some electronic circuit posts on Dev.to — small circuits, DIY...
agentsWhat nobody tells you about exporting your multi-agent prototype to a local workspace. Every...
agenticarchitectAutonomous agents are genuinely good at answering messy business questions. Give one an LLM and a set...
aiPR volume went up, ticket quality didn't, and the gap got filled with LLMs on both sides of the review: bots reviewing, bots replying, bots occasionally arguing with bots about priorities that only existed in a teammate's head. Our CEO named the actual problem, and it's bigger than code review.
Workflows from the Neura Market marketplace related to this Stable Diffusion resource