Gemini 2.5 Pro, AI Agents, and a Legendary Wizard of Oz Makeover: The Biggest Reveals from Google Cloud Next '25

Google Cloud just wrapped its Next ‘25 event in Las Vegas, unveiling a jaw-dropping 229 announcements, spanning everything from advanced AI models to new ways of connecting your favorite tools with Google’s agentic ecosystem.

 You’d think that would be enough to grab headlines on its own, but Google also teamed up with The Sphere to reimagine The Wizard of Oz using AI. And that might just be the most mind-bending demo of them all.

To unpack what happened at Next ‘25, I spoke with Marketing AI Institute founder and CEO Paul Roetzer, who attended the event, on Episode 144 of The Artificial Intelligence Show. 

A Quick Snapshot of Google Cloud Next ‘25

While Google unveiled far too many new products and updates to cover in detail, here’s a sampler of what dominated the conversation at Next '25:

• Gemini 2.5 Pro. Google’s latest heavyweight AI model, now in public preview. It claims advanced reasoning and coding capabilities that outpace previous generations—and currently ranks #1 on the Chatbot Arena leaderboard, according to Google’s stats.
• Gemini 2.5 Flash. A speedier, more cost-efficient variant of Gemini that Google says still packs enough punch for many tasks.
• Generative media. Major upgrades to Google’s text-to-image, text-to-audio, and text-to-video models (Imagen 3, Chirp 3, Veo 2, and a new text-to-music model called Lyria). The focus? High-quality outputs and faster, more precise editing capabilities—whether you’re generating images or entire video scenes.
• AI infrastructure. Google flexed its muscle in massive-scale AI training and inference. Think new GPUs, next-gen TPUs, ultra-fast networking, and storage—basically, an industrial-strength AI backbone.
• Agentspace. Google made big updates to its “AI control center,” which integrates your everyday work apps and data with its most powerful models and newly improved AI agents. The idea: let you build, customize, and manage AI-driven workflows—and do it all within a single ecosystem.

Sounds like enough news to last the year, right? Well, Google had one more trick up its sleeve…

Google AI Brings 1939’s Wizard of Oz into the Future at The Sphere

Imagine stepping into one of the most advanced venues on the planet, only to witness The Wizard of Oz from 1939 getting an ultra-HD, 360-degree AI makeover—complete with out-of-this-world visuals, seats that rumble with thunder, and an immersive environment that puts you right in Dorothy’s ruby slippers.

That’s exactly what happened on the first night of Google Cloud Next 25. Roetzer was there, and according to him, it was nothing short of “crazy.”

The Sphere is a massive spherical 360-degree event space where Google previewed its work to use AI to turn The Wizard of Oz into a fully modern, immersive experience.

So how do you bring a classic film—shot in a small, rectangular frame—onto a 160,000-square-foot dome without it looking horribly stretched and blurry?

According to Roetzer, Google took multiple AI models (including versions of Veo and Imagen) and, with a huge team of human professionals, pioneered three key techniques:

1. Super resolution. Sharpening those old-school frames into ultra-high definition imagery that can match The Sphere’s enormous display.
2. Outpainting. Filling in the gaps between scenes to seamlessly expand the original frame.
3. Performance generation. Compositing characters and details that simply never existed in the original film, so the viewer sees a continuous, immersive scene in 360 degrees.

The end result is a mesmerizing preview of what’s possible when you mix cutting-edge AI with the next generation of immersive experiences. The full reimagined Oz film debuts at The Sphere in late August. If you’re planning a trip to Vegas, you might want to add this to your list.

“The thing I took away from it was the human-machine collaboration,” says Roetzer.

This wasn’t a matter of handing the entire film to Gemini and having it figure all of this out. Dozens of the top minds within Google DeepMind and Google Cloud worked on this, pushing the limits of the models and creating entirely new techniques to make this possible.

Says Roetzer:

“They interviewed one guy from Google DeepMind and said, ‘Hey, when this project started [two years ago], what did you think was impossible?’ And he answered: ‘Everything. There was nothing we were doing that the models at that moment could actually achieve.’”

Agentspace: The Real Star of Google Cloud Next ‘25

Wowing the crowd with The Wizard of Oz aside, the real star of Next ‘25 for many attendees was Agentspace, Google’s hub for building, training, and orchestrating AI agents.

“It’s a single space that enables you to, in a no-code environment, build agents to do whatever you want to do,” says Roetzer. “And it connects to third-party software and data. So it basically becomes a platform where you live and do everything you need to do [with AI].”

That means everyday users can use Agentspace to build their own AI agents for tasks like:

• Researching any topic (with a “Deep Research” mode that synthesizes sources and details)
• Transforming lengthy documents into presentations
• Generating an audio overview of your reports or strategic plans
• Creating an “agent gallery” where your entire organization can discover and deploy new AI helpers

Basically, it’s the holy grail of letting AI handle busywork so you can stay focused on higher-level tasks. 

“The vision for it is powerful,” says Roetzer. “You could see how it becomes like a control panel basically for a knowledge worker to just have all the tools they need right there.”

The only catch? Agentspace isn’t broadly available yet. Google’s letting potential users request access, but a timeline for general availability is still unclear.

Sundar Pichai’s Big AI Prediction

In a session at Next ‘25, Alphabet and Google CEO Sundar Pichai made a statement that got folks buzzing.

He expects the pace of model advancements to continue for at least the next 12 to 18 months, says Roetzer, with new major models every three to four months.

Let that sink in: Every few months, we may see leaps in capabilities, new specialized models, and expansions of existing AI frameworks—on top of everything Google just announced. 

“That is just crazy to think about,” says Roetzer.

If you’re feeling overwhelmed, you’re not alone. But you’re also about to witness an era of unprecedented AI growth, specifically from Google’s ecosystem, if Pichai’s prediction holds.