Callosum raises £75m led by Atomico with DCVC
This article covers Callosum, an AI infrastructure startup, which has raised £75m in a seed funding round led by Atomico to commercialise a software layer that co-optimises AI models and specialised chips and to launch APIs that let developers adopt frontier hardware. The development aims to support developers and enterprises seeking cheaper, faster and more resilient AI inference by orchestrating workloads across specialised processors.
Callosum, an AI infrastructure startup, has raised £75 million in a seed funding round led by Atomico — one of the largest seed deals in Europe — to commercialise a software layer that co-optimises AI models and specialised chips and to ship a family of APIs that let developers adopt frontier hardware without extensive engineering work. The raise matters because it signals investor conviction in heterogeneous compute and orchestration as the next phase of industrial AI deployment.
Why it matters
The economics of running large AI models are shifting. After years of investment in ever-larger foundation models, attention and budgets are moving towards inference efficiency and operational cost. Callosum’s approach addresses both by decomposing workloads and routing each task to the model and processor best suited to it, rather than running everything on a single type of chip.
If the company’s performance claims hold up in practice — a 4x speed improvement, 70% lower compute costs and a 10% higher task success rate on complex agentic financial workloads compared with a single frontier model on conventional GPUs — the technology could materially change how enterprises deploy AI. It also intersects with concerns about supply concentration and national resilience, since orchestrating across hardware reduces dependency on any single supplier or architecture.
Inside the product
Callosum’s platform sits as an orchestration layer between applications and hardware. It decomposes AI inference work, matches sub-tasks to specialised models and processors, and routes execution to the most suitable hardware to balance speed, cost and precision.
The company launches today with a family of APIs intended to let developers plug in frontier hardware without rewriting applications, collapsing what is often months of engineering into a single integration. Callosum says the APIs are already in production for complex, agentic workloads in financial services.
Callosum’s silicon and infrastructure partners include Cerebras, the US chipmaker known for wafer-scale processors, which the company credits for some of its fastest inference results. Other hardware partners are Rebellions in South Korea, Axelera in the Netherlands and Tendrils in the United Kingdom; infrastructure partner Supermicro supports deployment. As a founding member of the Scaling Inference Lab at ARIA, the UK’s advanced research agency, Callosum is also integrating emergent technologies from firms such as Oriole Networks and Lumai, which are exploring light-based or optical computing approaches.
Who are the investors?
The round was led by Atomico with participation from Plural, DCVC and UK Sovereign AI. The £75 million seed follows a £7.5 million pre-seed less than six months earlier, underscoring rapid investor appetite for infrastructure plays that help bring specialised processors into production.
In the announcement, Sasha Vidiborskiy, Partner at Atomico, said:
AI is entering a new era, where intelligence won't be defined by a single model or a single chip, but by systems that can enable the best setup for the job. Callosum is building the infrastructure layer that will make that possible. The team has already demonstrated world-class technical breakthroughs, and incredible speed of execution, and we believe their vision has the potential to fundamentally reshape how AI is deployed at scale. We're proud to partner with them as they build the foundation for the agentic era.
In the announcement, Jason Pontin, General Partner at DCVC, said:
Orchestration is all you need. Token consumption by users of scientific and enterprise computing is expanding exponentially, but Callosum knows the key to making each token cheaper and more valuable is to treat AI computing as multimodal, not monolithic. Their talent for orchestrating workloads across heterogeneous computing resources is rooted in their deep understanding of which jobs are best for which models and chips.
If you're researching potential backers in this space:
- Explore our list of AI venture capital firms in the UK
- Browse our directory of AI angel investors in the UK
Founder perspective
The company was founded by Danyal Akarca and Jascha Achterberg, who met while doing PhDs at Cambridge. Their academic work argued that biological intelligence arises from specialised circuits working together rather than repeating one unit many times, a thesis they say maps directly to multi-model, multi-hardware AI systems.
In the announcement, Danyal Akarca, Co-founder & CEO at Callosum, said:
The industry has been betting that one model or one kind of chip will rule them all. But nature shows us the opposite. Intelligence is collective and will emerge from many specialised systems working together. We believe AI needs a new axis of scaling, where many models and hardware co-evolve as a single system. That's how we build AI that is not only more capable, but dramatically faster, more affordable and far more energy-efficient, opening the door to applications that were not viable before.
In the announcement, Jascha Achterberg, Co-founder at Callosum, said:
AI sovereignty isn't about owning a chip fab. It's about never being hostage to one. It's about building systems that can harness the best intelligence wherever it exists. By running AI across many different chips and architectures, we create systems that are more resilient, more efficient and less dependent on any one company or country. That's how you build AI infrastructure that can endure.
The founders say the new funding will be used to scale the technology, deepen partnerships across the silicon ecosystem and expand the team. A selection of the tailored-inference APIs are live now, with more customer deployments and partner announcements expected in the coming months.
The bigger picture
Callosum’s raise highlights a broader industry shift: spending and attention are moving from raw model scale to making inference cheaper, faster and more reliable. The proliferation of specialised processors — from wafer-scale to analog and optical designs — challenges the one-chip-fits-all assumption and creates an opportunity for orchestration layers that can span architectures.
The deal also reflects growing interest from AI investors in infrastructure that reduces vendor lock-in and supports national resilience in compute supply. In the UK context, Callosum’s involvement with ARIA’s Scaling Inference Lab and partnerships with local firms like Tendrils underline a domestic thread in an otherwise international hardware landscape.
This funding round will be watched closely across Europe. If Callosum’s orchestration approach proves repeatable at scale, it could influence how enterprises and public bodies think about procurement, sovereignty and the next architecture for deployed AI.
They've invested in Callosum
Get to know these AI investors
| Investors | Investment Focus | Startup Investments | Round Size | Connect |
|---|---|---|---|---|
 | ||||
 | ||||
 DCVC( ) Palo Alto, US  | ||||
| All investors | All investor sectors | All funded startups | All funding rounds |
Click here for a full list of 7,589+ startup investors in the UK