£38m to cut media AI costs and latency via edge inference hardware
This article covers Runware, a media startup, which has raised a £38m Series A led by Dawn Capital to build a unified API for AI media generation. The funding will support development of its Sonic Inference Engine and a platform designed to centralise model access and reduce inference cost and latency for product teams, content platforms and developers.
Runware, a media startup with offices in London and San Francisco, has raised a £38 million Series A led by Dawn Capital to build a unified API for AI media generation. The round funds product development of its Sonic Inference Engine® and a stated push to support millions of new open-source and closed-source models — a bet that matters as businesses try to embed media AI without ballooning costs or latency.
Why it matters
AI-driven media — images, video and audio — is moving from experimental features to product requirements for many consumer and enterprise apps. That shift is exposing three bottlenecks: fragmented access to models, latency that harms user experience, and inference unit costs that do not scale.
Runware positions itself as a single API that removes fragmentation while shifting inference to a bespoke hardware and software stack it says reduces capital and operating costs. The company cites the AI inference market rising from an estimated $26 billion in 2025 to $68 billion by 2028, underlining why cheaper, faster inference matters for product teams and for investors backing AI infrastructure.
Customers named by the company include Wix, Together.ai, ImagineArt, Quora and Higgsfield. These examples represent content platforms and developer tools where media generation is core to product value or user engagement.
Inside the product
Runware aggregates what it calls almost 300 model classes and 400k+ model variants behind a single schema and endpoint, enabling teams to A/B test, route or swap models with minimal code changes. The company says its Sonic Inference Engine® combines software and purpose-built inference hardware to deliver significantly lower cost per inference and lower latency than conventional cloud GPU deployments.
Key performance and pricing claims the company provides:
- Up to 10x lower inference pricing and up to 9x lower cost compared with typical GPU datacentre approaches, according to Runware.
- For open-source models, Runware reports 30–40% faster inference than other media inference platforms and 5–10x better price-performance than incumbents; for closed-source models it reports 10–40% savings.
- A roadmap to deploy more than 2 million new AI models from Hugging Face to its platform by the end of 2026.
Runware also highlights a distributed inference approach with small, modular "inference PODs" placed closer to users. The company says this lets it deploy capacity faster and in locations where power is available and affordable, which it frames as both a cost advantage and a response to evolving national AI regulations.
Who are the investors?
Runware’s Series A is led by Dawn Capital. The company has not disclosed other participants in the round.
In the announcement, Shamillah Bankiya, Partner at Dawn Capital, said:
Runware is already proving a hit with global companies building AI applications that require media inference. Flaviu and Ioana have built the rare platform that delights developers, satisfies enterprise checklists, and bends the cost curve in the customer’s favour. Runware has the right product at the right layer, built by the right team, and we’re thrilled to be on the journey with them as they take on this huge and urgent market.
If you're researching potential backers in this space:
- Explore our list of media venture capital firms in the UK
- Browse our directory of media angel investors in the UK
Founder perspective
The founders frame the product as solving engineering and commercial barriers to wider AI adoption in media workflows.
In the announcement, Ioana Hreninciuc, Co-founder & CEO at Runware, said:
We give clients the best price and developer experience in a single API, so they can roll out any new model in minutes - without integrating dozens of providers, managing RPMs, or negotiating huge commitments. Through our API, they offer unlimited AI features to end-users, and we see them hit repeated growth peaks as a result.
In the announcement, Flaviu Radulescu, Co-founder at Runware, said:
I believe that in the future, most of the products and services will be enhanced by AI. We are building a platform that can run AI faster, more cost-effectively, with higher redundancy and lower latency. Our inference PODs are 100x cheaper and faster to build, and deployed near users, in alignment with each country’s evolving AI regulations. We’ve designed our architecture so that we can place inference capacity wherever power is available and affordable, rather than constructing large data centres that require years of approvals, construction, and new power infrastructure. We can have an inference POD up and running in 3 weeks, not 3 years.
The company says the new capital will be used to expand the inference platform, invest in the Sonic Inference Engine® and hire for product and infrastructure engineering roles.
The bigger picture
Runware’s raise highlights two broader trends in the UK and European AI ecosystem. First, investors are continuing to back infrastructure plays that aim to reduce the marginal cost of AI features. Second, there is increasing focus on deployment models that align with national regulation and regional data or power constraints, rather than relying solely on hyperscale datacentres.
For UK founders and product teams, Runware’s approach — centralising model access behind one API and deploying inference capacity closer to users — is one route to manage cost and latency as media AI moves into mainstream products. The deal also underlines continued VC interest in companies that tackle the practical economics of large-scale AI deployments across Europe and beyond.
They've invested in Runware
Get to know these media investors
Click here for a full list of 7,526+ startup investors in the UK