Forbes: Qualcomm’s Boom Highlights AI Shift To The Edge
I encourage you to read industry observer Scott Raynovich's article about the AI Edge. This was a great read for me, and the article quotes my...
9 min read
Marc Austin : Apr 29, 2025 8:00:00 AM
When I look back in my personal database of training data, I realize that I had some interesting OG exposure to artificial intelligence way back in 1995. My first startup job was at a pre-Amazon internet retailer called Internet Shopping Network. I bet on Fox Network founder Barry Diller instead of the unknown Jeff Bezos to add a new digital channel to multi-channel retailing. That was obviously a bad bet, but I did learn something about neural networks.
ISN had a software download store, and my first assignment was to get a handle on credit card chargebacks our bank was assessing from from on our software download store. Our credit card processor was First Data (now Fiserv). They referred me to our bank, and when I asked how they identified and scored suspected fraud, the bank responded with “Neuralware.” Yes, there was a company called Neuralware back then. My boss BD Goel had a penchant for incubating startups, so rather than licensing Neuralware, we invited a group of data scientists to work with us for the better part of a year. We provided desks, workstations, meals, a couple racks of Sun UltraSPARC servers, what seemed like a ton of storage at the time, and access to all our transaction data. I suppose you could say this was an AI training workload well before such things were en vogue.
Fast forward 30 years, and sure enough, AI is still relevant in banking and financial services. Below are a handful of use cases, served up of course by generative AI. Fascinating that fraud detection & prevention is still at the top of the list after all these years.
AI algorithms can detect unusual patterns or anomalies in transactions (like sudden large withdrawals or spending in odd locations) and flag or block potentially fraudulent activity in real-time. Machine learning models get smarter over time by learning from past fraud cases.
While related to fraud detection, cybersecurity use case is more about fraud prevention. Your customer data is under constant attack by a range of cyber threats. AI has proven useful in cyber threat detection. Threat detection in turn prevents data breaches that give fodder for fraud. Better to prevent fraud than detect it!
I’m personally not a fan of the customer experience, but AI-powered chatbots promise to reduce operational costs. Chatbots can address common customer engagements like balance checks, loan statuses, or transaction issues 24/7, improving response time and reducing the workload on human agents. Hopefully they also improve customer satisfaction.
AI can analyze a broader datasets (not just credit scores) to assess someone's creditworthiness—things like transaction history, mobile data, or even social behavior. This opens up lending to more people, especially those with little credit history.
AI can track spending habits and income patterns to offer real-time budgeting advice, investment suggestions, or savings plans. Some platforms even use AI to help people invest based on their risk tolerance and goals (robo-advisors).
AI tools can quickly analyze vast amounts of data to identify potential risks—like market crashes, regulatory compliance issues, or investment losses—so that financial institutions can act before things spiral.
Tasks like data entry, KYC (Know Your Customer), onboarding, data analytics and report generation can be automated using AI, saving time and reducing errors. This is often referred to as Intelligent Process Automation (IPA). Full life-cycle workflows can be automated with the integration of advanced AI-driven technologies, enabling seamless coordination across various stages of a process.
AI models can process financial news, social media sentiment, and market data in real time to make trading decisions much faster than any human could. Hedge funds and financial institutions use this to gain an edge in the markets.
AI helps detect and report suspicious transactions more effectively by learning the subtle signs of money laundering across multiple channels—reducing false positives and improving investigation times. Anti-money laundering (AML) regulatory compliance involves a comprehensive set of procedures, laws, and regulations designed to prevent criminals from disguising illegally obtained funds as legitimate income.
AI models can analyze vast amounts of data like historical prices, trends, and global events to predict things like stock prices, economic shifts, or currency fluctuations, giving analysts and investors better tools to streamline decision-making.
As much as AI promises to automate and change banking operations, most banks continue to rely on legacy applications as the backbone of their core operations. Many of these applications were built years ago, prior to cloud computing. These apps were designed to run on virtual machines or bare metal, and their authors may have left the company. The time, cost and effort to rewrite them as cloud native micro-services is prohibitive. Most IT shops just want stable, reliable on-premises infrastructure with an effective lifecycle that matches the duration of their applications.
Infrastructure duration typically does not match application duration when banks rely on legacy infrastructure vendors. Legacy vendors with market saturation need renewals and upgrades to continue reporting revenue growth. That means they have policies that prematurely end the life of switches and routers. When this happens, they stop supporting the platforms even well before they expected lifecycle actually matures. Worse yet, some vendors stop licensing proprietary software so that customers no longer have the legal right to use their products.
The solution to matching duration is to choose open source infrastructure that is not controlled by a vendor. Smart banks will network like hyper-scalers. They purchase white-box equipment that meets open standards established by members of the Open Compute Project. They will run open-source software and fully automate operations. When banks and insurance companies network like hyperscalers, they will maximize the effective life of their assets with no forced upgrades or EOL notices.
Hedgehog enables banks and financial services companies to network like hyperscalers. Our software appliance is open source. We test, certify and support our software on a wide range of white box network equipment that meets OCP specifications. Our customers prefer to get support from our talented team, but Hedgehog infrastructure will never stop working, preventing it from being used as a lever just to squeeze licensing revenue out of customers.
We talk a lot about businesses and governments needing to have intentional, well-conceived AI infrastructure strategies. Most every major bank or financial service provider runs private cloud data centers. Initiatives to gain cost savings through cloud migration have resulted in hybrid cloud deployments for certain banking services like digital banking. Fintech startups are more likely to use public cloud computing because they lack the capital or operational efficiency to invest in their own computing resources. The exception in fintech of is cryptocurrency. The computing power required to mint new coins makes it unprofitable for miners to use AWS. Crypto miners have built their own computing resources to run blockchain workloads. Miners typically have not needed to invest in the foundational cloud services, and security measures that multiple data scientist tenants need in GPU cloud solutions for their AI technologies.
The reality is that most players in the banking industry and financial services continue to operate on-premises infrastructure for most of your banking services. They do this for regulatory compliance. If you think about it for second, who else has more personal identifiable information (PII) about you than your bank or insurance company? In most countries, strict regulatory or data sovereignty requirements require your bank, broker, or insurer to keep your data private in a private cloud running in an on-premises data center. You want this for common banking operations like the software that manages your bank statement, and you also want this for new technologies like generative AI. Banks ensure your data privacy with private cloud for:
Tighter data control – Crucial for data security, customer privacy, sensitive data, or proprietary models.
Regulatory pressure – Regulatory requirements in many regions including the EU Middle East include data residency laws.
Legacy integration – Easier to connect with older core banking systems.
The banking system is considered critical infrastructure in most countries. Controlling this critical infrastructure is required for national security, particularly in the new world order of tariffs, quotas and the threat of a global trade war. Outsourcing critical infrastructure to a 3 member cloud computing cartel based in the United States creates strategic weakness for many countries.
Some of the biggest U.S. banks have "data bunkers.” These are underground or ultra-hardened data centers designed to survive natural disasters, cyberattacks, and even limited physical attacks. Most are Tier III or Tier IV (high redundancy, 99.995%+ uptime). Security is high 24/7 armed guards, mantraps, and biometric access. Some facilities are over 500,000 square feet, and most have full replication at geographically distant sites for disaster recovery.
Most banks and financial services companies—especially mid-sized or fast-moving fintechs—are using public cloud providers like AWS, Microsoft Azure, or Google Cloud Platform (GCP) to rapidly develop digital transformation initiatives and new customer experiences.
Many banks take a hybrid approach—keeping sensitive or critical systems and data management in-house on private cloud, while using public cloud to experiment with new customer experiences. A bank might store sensitive customer transaction data on-prem for compliance, but experiment with training fraud detection models on cloud infrastructure. Another bank may use the cloud to run chatbot services or mobile app features, while core banking services remain on tightly secured private servers.
As the banking sector moves from experimenting with advancements in artificial intelligence on public cloud, they are learning the scalability metrics for AI workloads and the need to optimize cloud technology to manage IT costs.
Hedgehog offers an AI cloud networking solution uniquely suited for hybrid multi-cloud infrastructure strategies. Our Open Network Fabric is built from the ground up for AI workloads. Our high-performance network minimizes congestion, and we offer a cloud user experience for on-premises workloads. Our Gateway is built to connect private cloud and on-premises workloads with virtual private clouds running in AWS, Azure or Google Cloud. This means that Hedgehog banking customers can run workloads wherever they want with a common, consistent operational model optimized for cost and performance.
A lot of banks and financial services companies operate (or heavily rely on) their own data centers because of how critical data, security, and uptime are in that industry. Here are a few case studies and examples:
JPMorgan Chase – One of the biggest — they have multiple large-scale data centers around the world, and are heavily investing in private cloud infrastructure too. Examples of JPMorgan Chase data centers include their Westerville, Ohio facility (just outside Columbus). This is one of their largest, high-security facilities handling critical banking operations. Their Plano, Texas data center is part of their larger corporate campus and tech hub expansion in Texas.
Bank of America – Operates private data centers and has been consolidating and modernizing them in recent years. One key operations center handling core banking workloads is Chesterfield County just south of Richmond, Virginia.
Wells Fargo – Runs several dedicated data centers and also partners with cloud providers. Their Winston-Salem, NC data center is a critical operations center, with backup and redundancy features.
Citigroup (Citi) – Owns and operates global data centers but is also shifting toward a hybrid model (own data centers + public cloud). Their primary data center facility is Florham Park, New Jersey, close to Citi's NYC operations. It plays a significant role in hosting core banking platforms and trading systems. Citi’s Dallas, Texas facility is a newer center used for disaster recovery and critical cloud operations.
Goldman Sachs – Historically Goldman has their own data centers, and while they are moving some workloads to cloud (like AWS), they still maintain critical infrastructure internally. Goldman operates several critical Tier III/Tier IV data centers around the New York metro area, hosting a lot of their risk management and trading workloads. Their London (Surrey area) facility is the primary data center for European operations.
Morgan Stanley – Similar strategy: in-house data centers for core systems, hybrid cloud for non-core workloads.
BNY Mellon – Operates key data centers for custody and asset management services.
Deutsche Bank - Historically heavy on internal infrastructure. In a strategic shift: large-scale partnership with Google Cloud, but continues operating critical data centers. Several large-scale facilities near Frankfurt continue supporting European banking services.
HSBC - Major in-house data centers in and around Hong Kong, critical for supporting HSBC's massive Asia-Pacific operations. The United Kingdom data center near London is one of Europe’s larger banking private data centers. HSBC’s hybrid cloud environment beginning to take shape.
Barclays - Operates its own facilities with critical banking systems kept in-house and ancillary services moved to cloud. Their Manchester data center focuses on resiliency and disaster recovery for UK operations.
BNP Paribas - Runs several data centers across Europe, balancing internal hosting with emerging cloud adoption. Their Paris data center hosts core IT operations, including private cloud hosting infrastructure.
PNC Financial Services - Operates private data centers and is investing in modernizing infrastructure toward a hybrid environment.
U.S. Bank - Runs in-house facilities, but is increasingly collaborating with cloud providers (Google Cloud, AWS).
TD Bank - Maintains proprietary data center operations in North America.
Royal Bank of Canada (RBC) - Operates private data centers and has major internal cloud initiatives.
Visa – Owns and operates ultra-secure, highly redundant data centers for transaction processing. Visa Operations Center East (OCE) located in “Data Center Alley” near Ashburn, Virginia. This highly secure data centers handles major transaction volumes across the U.S. Visa Data Center West in O’Fallon, Missouri is a key redundancy and backup center to OCE.
Mastercard – Same — mission-critical transaction processing runs through Mastercard-owned centers. Their Global Technology and Operations Center is also located in O'Fallon, Missouri processing billions of transactions per year.
American Express – Has its own infrastructure and data centers, although also embracing cloud transformation.
PayPal – Operates its own data centers and also uses cloud services.
FIS (Fidelity National Information Services) – One of the world's largest fintechs, operates multiple secure data centers. The FIS Data Center in Little Rock, Arkansas is still a key fintech processing hub for many small banks and credit unions.
Fiserv – Major provider of financial tech solutions, with heavy investments in secure, compliant data center infrastructure.
Charles Schwab – Runs internal data centers for brokerage and investment services, although moving to hybrid models.
BlackRock – Operates and manages some internal data centers (especially for Aladdin, their risk management platform).
NASDAQ – Maintains its own data centers for stock exchange and trading systems where ultra low-latency is a key requirement. The NASDAQ data center in Carteret, New Jersey is the main matching engine for the stock exchange. Ultra low-latency trading happens here.
If you want to sustain your core banking operations or build new hybrid cloud infrastructure for artificial intelligence and digital transformation projects, schedule a call with the Hedgehog team. We would love to help you reach your project goals.
I encourage you to read industry observer Scott Raynovich's article about the AI Edge. This was a great read for me, and the article quotes my...
In my “AI Needs a New Network” post last week, I noted that NVIDIA reported $13 billion in networking ARR on $18.4 billion of annual data center...
In case you missed it, Juniper produced a web event last week called AI-Native NOW. I thought it might introduce Juniper as a Hedgehog competitor for...