For payments, fraud detection and identity verification, the milliseconds lost in “round-tripping” data to a distant server represent more than just lag — they are a structural vulnerability. As the industry matures, the competitive frontier is shifting toward edge AI, moving the point of decision-making from the data center to the literal edge of the network — the ATM, the point-of-sale (POS) terminal, and the branch server.

From Batch Processing to Instant Inference

At the heart of this shift is inference, the moment a trained model applies its logic to a live transaction. While the cloud remains the ideal laboratory for training massive models, it is an increasingly inefficient theater for execution.

Financial workflows are rarely “batch” problems; they are “now” problems. Authorizing a high-value payment or flagging a suspicious login happens in a heartbeat. By moving inference into local gateways and on-premise infrastructure, institutions are effectively eliminating the “cloud tax” — the combined burden of latency, bandwidth costs and egress fees. This local execution isn’t just a technical preference; it’s a cost-control strategy. As transaction volumes surge, edge deployments offer a more predictable total cost of ownership (TCO) compared to the variable, often skyrocketing costs of cloud-only scaling.

Coverage from PYMNTS highlights how financial firms are transitioning from cloud-centric large models toward task-specific systems optimized for real-time operations and cost control.

From Cloud-Centric AI to Decision-Making at the Edge

The first wave of enterprise AI adoption leaned heavily on cloud infrastructure. Large models and centralized data lakes proved effective for analytics, forecasting and customer insights. But financial workflows are not batch problems. Authorizing a payment, flagging fraud or approving a cash withdrawal happens in milliseconds. Routing every decision process through a centralized cloud introduces latency, cost and operational risk.

Edge AI moves inference into branch servers, payment gateways and local infrastructure, enabling systems to decide without every query circling back to a central cloud. That local execution is especially critical in finance, where latency, privacy and compliance are business requirements.

Real-time processing at the edge trims costly round trips and avoids the cloud bandwidth and egress fees that accumulate at scale. CIO highlights that as inference volumes grow, edge deployments often deliver lower and more predictable total cost of ownership than cloud-only approaches.

Banks and payments providers are identifying specific edge use cases where local intelligence unlocks business value. Fraud detection systems at ATMs can use facial analytics and transaction context to assess threats in real time without routing sensitive video data, keeping customer information on-premise and reducing exposure.

Edge AI also supports smart branch automation, real-time risk scoring and adaptive security controls that respond instantly to contextual signals, functions that centralized cloud inference cannot economically replicate at transaction scale.

Edge AI delivers clear operational and governance advantages by reducing bandwidth use, cloud dependency and attack surface. Keeping decision logic local also simplifies compliance by limiting unnecessary data movement, a priority for regulated financial institutions.

Edge AI Stack Is Coalescing Across the Tech Industry

The broader tech ecosystem reinforces this trend. As reported by Reuters, chipmakers such as Arm are expanding edge-optimized AI licensing programs to accelerate on-device inference development, reflecting growing conviction that distributed AI will capture a larger share of enterprise compute workloads. Nvidia is advancing that shift through platforms such as EGX, Jetson and IGX, which bring accelerated computing and real-time inference into enterprise, industrial and infrastructure environments where latency and reliability matter.

Intel is taking a similar approach by integrating AI accelerators such as its Gaudi 3 chips into hybrid architectures and partnering with providers including IBM to push scalable, secure inference closer to users. IBM, in turn, is embedding AI across hybrid cloud and edge deployments through its watsonx platform and enterprise services, with an emphasis on governance, integration and control.

In financial services, these converging moves make edge AI more than a deployment option. It is increasingly the infrastructure layer for enterprise AI, enabling institutions to embed intelligence directly into transaction flows while maintaining discipline over cost, risk and operational continuity.

Republishing guidelines

by Brandon Jarvis

Gov.-elect Abigail Spanberger has tapped Del. Mark Sickles, D-Fairfax, to serve as her Secretary of Finance.

Sickles has been in the House of Delegates for 22 years and is the second-highest-ranking Democrat on the House Appropriations Committee.

“As the Vice Chair of the House Appropriations Committee, Delegate Sickles has years of experience working with both Democrats and Republicans to pass commonsense budgets that have offered tax relief for families and helped Virginia’s economy grow,” Spanberger said in a statement Tuesday.

Sickles has been a House budget negotiator since 2018.

“We need to make sure every tax dollar is employed to its greatest effect for hard-working Virginians to keep tuition low, to build more affordable housing, to ensure teachers are properly rewarded for their work, and to make quality healthcare available and affordable for everyone,” Sickles said in a statement. “The Finance Secretariat must be a team player in helping Virginia’s government to perform to its greatest potential.”

Sickles is the third member of the House that Spanberger has selected to serve in her administration. Del. Candi Mundon King, D-Prince William, was tapped to serve as the Secretary of the Commonwealth, and Del. David Bulova, D-Fairfax, was named Secretary of Historic and Natural Resources.

This work is licensed under CC BY-NC-ND 4.0

Stories posted on Virginiascope.com are available for publications to republish in their entirety for free.

View the guidelines

SEOUL, Dec 23 (Reuters) – South Korea’s central bank needs to remain wary of financial stability risks, such as heightened volatility in the won currency and upward pressure on house prices, a board member said on Tuesday.

“Volatility is increasing in financial and foreign exchange markets with sharp fluctuations in stock prices and comparative weakness in the won,” said Chang Yong-sung, a member of the Bank of Korea’s seven-seat monetary policy board.

Sign up here.

The won hit on Tuesday its weakest level since early April at 1,483.5 per dollar. It has fallen more than 8% in the second half of 2025.

Chang also warned of high credit risks for some vulnerable sectors and continuously rising house prices in his comments released with the central bank’s semiannual financial stability report.

In the report, the BOK said it would monitor risk factors within the financial system and proactively seek market stabilising measures if needed, though it noted most indicators of foreign exchange conditions remained stable.

Monetary policy would continue to be coordinated with macroprudential policies, it added.

The BOK’s next monetary policy meeting is in January.

Our Standards: The Thomson Reuters Trust Principles., opens new tab