There’s a gap between ambition and infrastructure when it comes to deploying AI in gaming. Writing for iGB, Cristopher Kuehl, chief data, information & AI officer at Continent 8 Technologies, offers practical advice on the questions operators should be asking of solution providers.
“We have an AI strategy.”
That is now fairly standard across iGaming boardrooms. The problem? Many of those strategies are already failing at the infrastructure level.
Having a strategy and having the infrastructure required to execute it are two very different things. Across the industry, there is a growing gap between ambition and operational reality. Operators are investing in heavy large language models (LLMs) and complex predictive tools. Yet many are still relying on infrastructure originally designed for traditional workloads.
If you are an operator, the question is no longer whether a provider can ‘host AI’. Instead, the focus should be on how their network handles the specific, high-velocity and highly regulated demands of iGaming.
Here are the core technical questions to bring to your next vendor review.
Where is the compute located?
Legacy iGaming infrastructure was built around CPUs and RAM. Modern AI requires parallel processing, graphics processing units (GPUs) and Neural Processing Units (NPUs).
If your provider is quoting standard cloud instances without detailing specialised hardware clusters, this may limit AI performance and scalability. Operators should look to providers that can offer access to high-performance GPU infrastructure within regulated environments.
Running workloads on a general-purpose cloud in a different country can create latency. Localised, high-density compute is often needed to handle real-time fraud detection and in-play personalisation.
Is the model in-country or in-state?
Data residency is a strict, known regulatory requirement in iGaming and online sports betting. Player data cannot cross certain borders. However, AI creates additional complexity: not just where data is stored but where it is processed.
For example, if your player database sits in Ontario, Canada, but the AI model that is performing calculations lives in a data centre in Virginia, US, this may raise compliance concerns depending on how data is processed and transferred. Regulators across multiple jurisdictions, including the UK and North America, are increasing scrutiny of data residency, cross-border processing and third-party dependencies.
As a result, operators should ensure that both data and the associated AI processing remain within approved regulatory boundaries. In practice, this means working with providers that support in-country or regionally compliant AI deployments.
iGB has published a report on the evolution of AI in gaming, produced alongside managed IT solutions provider Continent 8. Sign up to read for free.
How do we prove the logic?
Regulators require transparency. For example, if an algorithm flags a VIP player for problem gambling or blocks a withdrawal, the operator must be able to prove the reasons why it happened.
Standard logging alone is insufficient. You need telemetry that captures model inputs, the exact weights used, and the confidence score of the output.
Asking the provider what kind of audit logging they support at the model layer is important. Without a regulator-ready audit trail of a specific AI decision, operators cannot deploy that model in a live environment. Operators should be able to recreate the exact logic path the AI took at any historical timestamp.
What is the exact latency of inference?
In our sector, milliseconds translate directly to revenue. Research across digital platforms shows that even a one-second delay in mobile load time can reduce conversion rates by up to 20%. This highlights how sensitive user engagement is to performance, particularly in mobile-first environments that closely mirror player behaviour.
At the same time, the cost of disruption is immediate. Industry estimates suggest downtime in gaming environments can exceed $6,000 per minute, underlining the direct financial impact of even short service interruptions.
Therefore, in use cases such as real-time odds generation or live bet limits triggering, the processing time should be near zero. Operators should assess the proximity between their operational data and AI compute environments. Architectures relying on public internet connectivity may suffer in terms of user experience. A private, high-speed connection between your raw operational data and your AI models is needed, ideally with the compute sitting adjacent to the data.
Continent 8 Technologies has become a thought leader in deploying AI in igaming
How are you securing and orchestrating the models?
Operators should evaluate orchestration and security. Scaling AI capacity based on a Saturday night player load requires managed Kubernetes specifically tuned for AI workloads. Your provider needs to scale GPU allocation up and down dynamically.
Furthermore, the security perimeter for AI is fundamentally different. Standard firewalls may not be able to stop prompt injection attacks or model manipulation.
Threat actors are no longer just trying to steal data. They are trying to disrupt the models behind the business decisions. Providers need to be able to demonstrate to operators their AI-specific defense mechanisms and how they protect the integrity of the model itself.
Moving out of the experimental phase
You should not be treating AI as a plugin. It is becoming a foundational component of modern technology stacks.
The industry is moving out of the experimental phase of chatbots and into the scale phase of adoption. Even a brilliant predictive model relies on the quality of the infrastructure supporting it.
The opportunity is clear: ask better questions, demand better answers, and choose partners built for regulated AI at scale.
Operator AI infrastructure checklist
Before selecting an AI infrastructure provider, operators should pressure-test capabilities against a core set of non-negotiable requirements.
- Is compute local and compliant?
- Are GPUs available and scalable?
- Can decisions be fully audited?
- Is inference latency sub-millisecond?
- Are models secure and protected?
- Are costs predictable at scale?
Infrastructure built for iGaming demands
Continent 8 has built its infrastructure specifically for the demands of regulated iGaming and online sports betting, not retrofitted from general-purpose cloud environments.
With a global footprint of licensed locations spanning over 100 datacentres globally, private high-speed connectivity, and in-country compute capabilities, Continent 8 enables operators to deploy AI workloads where they need to run – securely, compliantly, and with minimal latency.
From GPU-enabled environments to fully auditable architectures and AI-aware security, the focus is on delivering infrastructure that is aligned to the operational realities of modern iGaming. As AI moves into core production systems, having a provider that understands both the regulatory landscape and the performance requirements becomes essential.
To see how leading operators are deploying AI securely, compliantly and at scale using purpose-built infrastructure, explore iGB’s report on the evolution of AI in the gambling industry here.