Accels has launched a new platform designed to combine AI infrastructure with financial services for businesses building autonomous AI agents. The company is betting that as agents increasingly plan, execute tasks and transact independently, the financial infrastructure supporting AI will need to evolve alongside the underlying models—with token consumption, compute financing and payments becoming core components of the agentic economy.
The economics of artificial intelligence have historically been built around human users.
People purchase subscriptions, organisations buy software seats and developers pay for access to computing resources. AI agents introduce a different model. An autonomous agent can operate continuously, consume model tokens, access computing resources and potentially initiate transactions without a person being involved in every step.
Accels is building its business around that emerging model.
The company has launched a platform that combines AI model infrastructure with financial services, aiming to provide businesses and developers with the tools needed to build, operate and monetise agentic applications without assembling separate AI infrastructure, financing and payment providers.
Its proposition is based on a relatively simple premise: if AI agents become economic actors, they will require infrastructure designed around their own consumption and payment patterns.
The first component of Accels’ platform is AI model routing.
Through a unified API gateway, developers can connect applications to multiple large language models rather than integrating separately with individual providers. Accels says the platform currently covers 15 leading model providers and more than 150 AI models, including both international and Chinese model families.
For developers, model routing can provide a layer of abstraction between an application and the underlying AI models. That can make it easier to switch models, manage workloads and potentially optimise cost or performance depending on the task.
This is becoming increasingly relevant as AI applications move toward multi-model architectures.
A single agent may use one model for reasoning, another for summarisation and another for specialised tasks. Instead of hard-coding every integration into the application, an orchestration layer can manage those connections.
Accels is adding a financial layer on top of that infrastructure.
Its token lending and financial products include payment terms that allow customers to pay for token consumption on account rather than entirely upfront. The company is also introducing what it calls “compute loans”—credit lines intended to provide businesses with access to AI compute and token capacity without requiring equivalent upfront cash expenditure.
That represents an unusual intersection between cloud infrastructure and financial technology.
Traditional cloud-financing models generally revolve around physical or virtual computing capacity. AI introduces another variable because model usage can generate rapidly changing token consumption costs.
For an AI startup whose application suddenly scales from thousands to millions of agent interactions, its infrastructure bill can rise before customer revenue has fully caught up. Access to working capital tied specifically to AI consumption could therefore become an important financial product for companies operating at the edge of the agentic economy.
The opportunity is particularly relevant to startups.
An AI application can be technically ready to scale while still facing a cash-flow constraint caused by infrastructure costs. A financing product linked to token and compute usage could allow developers to scale capacity before receiving corresponding customer payments.
The third part of Accels’ platform is merchant acquiring.
The company says its acquiring infrastructure allows businesses to accept payments, manage complex billing and expand internationally through a unified platform.
At first glance, merchant acquiring may appear separate from AI infrastructure. In an agentic economy, however, the two could become increasingly interconnected.
An autonomous software agent might eventually purchase data, computing resources, digital services or other products on behalf of a user or business. That creates a need for payment infrastructure capable of handling machine-initiated transactions, usage-based billing and potentially much higher transaction volumes.
The distinction between an AI platform and a fintech platform could therefore become less pronounced.
Accels is effectively trying to connect three layers of that emerging stack: the models that power agents, the financial capacity that allows them to consume infrastructure, and the payment rails that allow businesses to monetise their activity.
The company’s thesis is reinforced by forecasts it cites from Goldman Sachs Research and other industry sources. Accels says Goldman Sachs Research expects global AI token consumption to grow 24-fold by 2030, while other industry forecasts estimate annual token-related spending could increase from roughly $200 billion today to as much as $2.5 trillion.
Those figures are forward-looking forecasts, not current market measurements, and the eventual scale will depend on how quickly agentic AI adoption develops, how token economics evolve and whether inference becomes more efficient.
There is also an important technology variable.
AI models are becoming more efficient, while inference costs continue to change as hardware, model architectures and optimisation techniques improve. If models require fewer tokens or cheaper computation to perform the same task, total spending may not increase proportionally with the number of AI agents.
That makes the financial infrastructure opportunity more nuanced than simply assuming token consumption will rise indefinitely.
Still, the direction of travel is significant.
AI agents could create new forms of software usage in which applications operate continuously rather than waiting for users to initiate individual requests. That changes not only infrastructure demand but also billing, credit, settlement and risk management.
Financial institutions and fintech companies are likely to encounter similar questions as agentic systems enter banking, commerce and enterprise workflows.
Who authorises an agent to spend money? How are spending limits established? How are machine-to-machine payments authenticated? How are transactions reconciled? Can an agent access credit? How can businesses distinguish legitimate automated activity from compromised agents?
These questions are still emerging, but they point toward a potentially new category of agentic financial infrastructure.
Accels is entering an ecosystem that already includes major cloud and AI infrastructure providers. AWS, Microsoft and Google provide large-scale compute, model and enterprise AI infrastructure, while specialised fintech companies provide payment, acquiring and embedded-finance services.
Accels’ differentiation is its attempt to combine those categories around the economics of AI agents.
That strategy could prove valuable if developers increasingly want a single infrastructure layer spanning model access, usage financing and payments. It could also face the complexity that comes with operating across multiple highly competitive markets.
The company’s launch nevertheless reflects an important change in how AI infrastructure is being conceived.
The next generation of AI applications may not simply be software that humans use. They could be systems that continuously consume resources, interact with other software and initiate economic transactions.
If that happens at scale, the infrastructure supporting them will need to account for more than inference.
It will need to support identity, credit, payments, billing and financial controls for machine-driven activity.
Accels is positioning its platform around that possibility.
Market Landscape
AI infrastructure is becoming increasingly modular. Model routing, inference optimisation, data infrastructure and agent orchestration are emerging as distinct technology layers, while fintech infrastructure is evolving around embedded payments, automated billing and machine-driven commerce.
Accels is attempting to connect these markets.
Its model-routing infrastructure competes indirectly with cloud and AI platforms from AWS, Microsoft and Google, while its financial products intersect with fintech providers specialising in payments, merchant acquiring and embedded finance.
The emerging opportunity is agentic commerce. As software agents begin to act on behalf of people and businesses, payment and credit infrastructure may need to accommodate machine-initiated transactions rather than assuming every transaction begins with a human.
That could create demand for new controls around authentication, spending permissions, transaction monitoring and settlement.
The major uncertainty is adoption. The agentic economy remains an emerging market, and forecasts for token consumption and spending depend heavily on the pace of enterprise adoption and the economics of increasingly efficient AI models.
Top Insights
- Accels has launched an infrastructure platform combining AI model access with financial services, targeting businesses building and scaling autonomous AI agents.
- Its model-routing layer connects developers to 150+ AI models, providing a unified interface across 15 model providers.
- Token financing and compute loans introduce a fintech layer to AI infrastructure, helping businesses fund usage without equivalent upfront expenditure.
- Merchant acquiring extends the platform into payments, potentially supporting the growing need for machine-driven billing and agentic commerce.
- The broader opportunity is agentic financial infrastructure, where AI systems may increasingly consume resources, access credit and initiate transactions autonomously.
Get in touch with our fintech expert
