Adobe announced in late June that it’s acquiring Topaz Labs, the AI video and image enhancement company behind Astra (video upscaling) and Wonder (image retouching and enhancement), in a deal expected to close in the second half of 2026. Financial terms weren’t disclosed. Topaz, founded over two decades ago and an Emmy Award winner in 2025 for its production technology, brings a mature toolkit — upscaling, sharpening, noise removal, stabilization, frame interpolation, footage restoration — that Adobe plans to fold into Firefly, Firefly Services, and Creative Cloud apps including Photoshop, Lightroom, and Premiere, while keeping the Topaz tools available standalone.
The asset everyone is under-reading
The headline framing is “Adobe buys an upscaler,” and that’s accurate but incomplete. Buried in the announcement is Neurostream, Topaz’s proprietary technology for running large, complex AI models locally on consumer devices rather than requiring cloud inference or high-end workstation hardware. For a company whose entire generative-AI strategy has run through cloud-hosted Firefly services, acquiring a proven local-inference stack is a meaningfully different kind of capability than acquiring another model.
Why local inference matters for creative work specifically
Cloud-based AI processing carries three costs a professional creative workflow feels acutely: per-generation pricing that scales with iteration count, round-trip latency that breaks the feel of interactive editing, and the requirement to upload client footage or unreleased assets to someone else’s servers. Local inference eliminates all three simultaneously. For video work in particular — where a single upscaling pass on a long timeline can mean enormous amounts of data moving to and from a cloud endpoint — the difference between local and cloud isn’t a nice-to-have, it’s the difference between a usable feature and an unusable one.
What it signals about Adobe’s direction
Read alongside Adobe’s broader 2026 positioning — the Firefly creative agent as an orchestration layer across Creative Cloud, and the agent’s expansion into ChatGPT and Claude — the Topaz acquisition fits a consistent pattern: Adobe is assembling infrastructure rather than competing head-on with frontier generative models. Orchestration, trusted professional-grade operations, and now local model execution are all bets on owning the layer between a creative professional and whatever model happens to be state of the art, rather than betting on being the one that makes it.