“We use tools like Terraform, so that we can stand up our stack automatically and ship it into different regions,” Acosta explains. “We moved up to the G4 architecture. We saw our median render times drop from 100 milliseconds to about 40 milliseconds, and we virtually have to change no code.”

By shifting image and video workloads directly onto Google Cloud’s GPU hardware, Imgix eliminated severe capacity constraints during heavy traffic spikes and boosted throughput per node by 6x. Beyond raw performance, transitioning to Google Cloud also fundamentally reshaped the engineering team's day-to-day operations. "In the old way, we had to react to all kinds of hardware failures," Acosta reflects. "In the new way, Google handles all of that for us. We still get paged but it's way less than once a week now, which is amazing.” Behind the scenes, database operations rely on Cloud Spanner to deliver consistent single-digit millisecond response times during high-volume requests while keeping infrastructure costs predictable. Meanwhile, BigQuery consolidates data from multiple pipelines into a single, unified view, allowing Imgix to analyze system performance and customer usage patterns in real time.

“Spanner actually gives us single-digit millisecond response times, which we need in a live throughput environment,” Acosta says. “It's very fast. It’s also very consistent and that helps us save costs.”

By pairing high-density compute with automated cloud orchestration, Imgix established a high-velocity foundation capable of powering visual media at scale with unprecedented speed.

To deliver sophisticated AI capabilities—like smart video cropping and automated highlight extraction—without inflating customer bills, Imgix uses Gemini to select and route requests to the best model for the task. Gemini evaluates whether a given image or video actually requires an AI feature before triggering heavy GPU processing jobs.

“If we're using the most expensive model every single time, that's a big bill for our customers,' Acosta explains. “So, we use Gemini to figure out, ‘Hey, does this image or video even need this AI feature?’ and for a fraction of a penny, we know not to apply it.”

Beyond platform features, Gemini has also supercharged Imgix’s development workflows. Engineering teams utilize Gemini to plan and implement new software features, dramatically accelerating product development cycles.

"Being able to use Gemini to both plan and implement new features has just really increased the speed of feature development,” Acosta reflects. "The speed is still on a hockey-stick climb."

Looking forward, Imgix is expanding its automated pipeline to serve custom, generative AI capabilities—including real-time background replacement, object removal, and image upscaling—enabling brands to personalize visual experiences dynamically at global scale.

"Being able to solve more complicated problems, maybe even specific for each customer at scale, even faster, that's where we're going," Acosta says. "With Google Cloud, Imgix is building a new way to power visual media."