Google Cloud’s Three Frontiers of AI: Intelligence, Latency, and Cost

AI models are pushing against three frontiers at once—and understanding this distinction is key to unlocking the potential of agentic AI.

That’s according to Michael Gerstenhaber, a product VP at Google Cloud who runs Vertex, the company’s unified platform for deploying enterprise AI. In a recent interview, Gerstenhaber outlined a framework I hadn’t heard before for thinking about model capabilities.

The Three Frontiers

1. Raw Intelligence

Models like Gemini Pro are tuned for raw intelligence. Think writing code: you want the best code you can get, and it doesn’t matter if it takes 45 minutes. You have to maintain it, put it in production. You just want the best.

2. Latency

Customer support is the classic example. You need intelligence to apply policies: “Are you allowed to transact a return? Can I upgrade my seat?” But it doesn’t matter how right you are if it took 45 minutes to get the answer. For these cases, you want the most intelligent product within your latency budget—because more intelligence no longer matters once the person hangs up.

3. Cost at Scale

Companies like Reddit or Meta want to moderate the entire internet. They have large budgets, but they can’t take an enterprise risk if they don’t know how it scales. They don’t know how many poisonous posts there will be today or tomorrow. So they have to restrict their budget to a model at the highest intelligence they can afford, but in a scalable way to an infinite number of subjects. For this, cost becomes very, very important.

Why This Matters

This framework is particularly valuable for anyone trying to push frontier models in a new direction. It explains why:

The Agentic AI Problem

Gerstenhaber also addressed why agentic systems are taking so long to catch on:

“This technology is basically two years old, and there’s still a lot of missing infrastructure. We don’t have patterns for auditing what the agents are doing. We don’t have patterns for authorization of data to an agent.”

Production is always a trailing indicator of what technology is capable of. Two years isn’t long enough to see what the intelligence supports in production.

Why Software Engineering Is Different

Agentic AI has moved uniquely quickly in software engineering because it fits nicely in the software development lifecycle:

Other professions need to produce similar patterns.

Google’s Vertical Integration

Gerstenhaber explained why he joined Google from Anthropic:

“Google is unique in the world in that we have everything from the interface to the infrastructure layer. We can build data centers. We can buy electricity and build power plants. We have our own chips. We have our own model. We have the inference layer that we control.”

This vertical integration, he argues, is a strength for Google in the AI race.

The Bottom Line

The three frontiers framework—intelligence, latency, cost—is a useful way to think about AI model deployment. Not every use case needs the smartest model. Some need the fastest. Some need the cheapest at scale.

Understanding which frontier you’re pushing against is the first step to building successful AI applications.

FAQ

What are the three frontiers of AI model capability?

Raw intelligence (best quality regardless of time), latency (fastest response within acceptable quality), and cost at scale (cheapest deployment for massive, unpredictable workloads).

Who is Michael Gerstenhaber?

Product VP at Google Cloud who runs Vertex, Google’s developer platform for AI. He previously worked at Anthropic for a year and a half.

Why is agentic AI slow to catch on?

Missing infrastructure for auditing agent actions, authorizing data access, and production patterns. Software engineering adopted it quickly because it fits the existing development lifecycle.

What is Google Vertex?

Google’s unified platform for deploying enterprise AI, giving customers access to agentic patterns, platforms, and inference of smart models.

—

*What do you think about the three frontiers framework? Does this match your experience with AI deployment? Share your thoughts below.*