Nowhere is cloud use being more dynamically adopted at the moment than in finance – just ask any cloud vendor, from hyperscaler to cloud service provider to SaaS provider, about how strategically important the sector is to them and their current growth.
Big, massive, deals in financial services offer huge prestige when successful.
Financial services cloud use cases have unique historical dynamics and differing levels of maturity. Overall, the sector has been late to the cloud party. On the sell-side in particular, there has been slow take up in the front office, partly for latency limtations of the sector’s highest frequency needs and partly for regulatory compliance edicts. However, the sector is now “modernizing” at full steam. In the middle office, cloud proved to be a great playground in
developing updated risk algorithms that risk researchers collaboratively worked on. These days, most production risk algorithms are cloud-centric, particularly anything batch, but like as with the front office, post-GFC regulatory cautiousness has led to delays in adoption. Back office functions have probably shifted faster sooner to cloud, in part because vendors of, say, commoditized solutions, e.g., ERP systems, have increasingly insisted on cloud-centric SaaS/AaaS delivery. There remain data privacy and governance issues here also, but the vendor/buy dependency rather than build has meant no easy alternative. We’ll return to their vendors and suppliers later, as they, interestingly, are ahead of the cloud adoption curve and perhaps paying some of the penalities of cloud dependency at this current moment in time.
The buy-side has arguably adopted cloud in a more organized fashion. Yes, there are many desks, like on the sell-side, but a more cohesively integrated workflow from research to portfolio manager to decision making which means, I think, more business collaboration and integration. I was peripherally involved with a terrific, early stage machine learning cloud project on model portfolio building at Aberdeen Asset Management, which if you’ve followed Aberdeen Asset Management over the last few years, has likely strategically impacted their
expansion. I see direct cause and effect between their early innovation in cloud and machine learning technologies leading the way to not just their success but the mainstream rollout of cloud data pipelines delivered by the hyperscalers and their partners. Other buy-side providers are following Aberdeen’s lead, while benefitting from commoditized and democratized tooling.
Regulators and central banks too have, perhaps paradoxically, flocked to cloud given their concerns over PII, data security and transparency. Collaboration across such organizations, for example in areas like economic and financial stability research, offers a perfect cloud use case. That being said, regulators and central banks also, paradoxically perhaps, warn of cloud concentration risks which I see as alive and real today among the current round of tech layoffs in and beyond Silicon Valley, to which we will return.
In sum there is some way to go for Financial Services to catch up with their internet and software services providers peers – often those same vendors which they have used in their back office business – which have been tech led, service-oriented, and largely regulator-free, cloud-first for not far off a decade. Any time you use the internet – to fill in your timesheet for work, buy something online, surf the internet, chat to your friends, post pictures, that’s cloud in action. Cloud is fully mainstream.
All this means that financial services cloud adoption is increasing at an escalating rate as it catches up. Yet they will also encounter – and probably are encountering – problems like those experienced by the tech early adopters of cloud. This presents an opportunity for Financial Services to learn from Tech sector mistakes, but also appreciate and overcome its distinct market drivers which perhaps exacerbate the acuteness of the sector’s challenges. Worst case if not dealt with, in the medium term, this may help force a round of systemic risk-inducing Wall Street and Main Street job reductions even more harmful than those impacting Silicon Valley today.
The Cloud Paradox Challenge
In 2021, Wang and Casada of Andreessen Horowitz wrote a brilliant, thought-provoking paper entitled the
Cost of Cloud, a Trillion Dollar Paradox. Their focus was the tech leaders, some the same firms providing services to our financial services sector, commonly predicated on rapid growth in the early/mid 2010s delivered by cloud scale. The result – the sector got addicted to spending more on cloud, particularly public cloud. Every $1 more spent on cloud, the authors argued, meant $20 less on market cap, leading to pain felt during trying times of lower growth later in the decade as business consolidated. In aggregate, this amounted to trillions of dollars lost, with an addictive need to use more cloud services to maintain current revenues, hence the paradox.
As a result, some tech firms, not all, took action. FinOps-led initatives focussed on cost reduction, cost optimization and, so the paper argues, beginning to take some tech back to in-house data centers. I see the current swathe of tech layoffs affecting Silicon Valley coming directly and indirectly from Cloud Paradox cost inflation.
Banks and financial firms could now already have such balance sheet challenges, perhaps unknowingly, maybe hidden deep in some obscure cost centers. Even if not, they should still proactively look for them as such costs could bite in future. With financial firms, including large investment banks, porting their IT portfolios onto cloud, costs may well swell, particularly among the compute/data-sensitive and hitherto-resistant front offices where only the very fastest microsecond latency processes need now to be conducted on-premises.
Financial Services Opportunities
Here is an opportunity for financial services firms to not make the same mistakes as their tech peers, and truly optimize their functions. Four possible opportunities to lower costs and improve performance include:
i) Consider microservices-based infrastructures. They may offer more efficiency than lift & shift as firms move applications to the cloud. Development teams in banks are more astute than most and can develop if not acquire DevOps skills to appropriately refactor, rearchitect and release with agility their applications across microservices. Slick cloud-native architectures mean optimized infrastructures.
ii) Right-size your compute. Think carefully about re-using and optimizing IT stacks on and off-cloud. When working with programming languages developed pre-cloud, e.g. Java, issues such as garbage collection can help vertical big data applications but also hinder as they tie up machines and instances. Also, onerous compilation can require dedicated cloud resource when only needed some of the time. Take action to manage and optimize. Java is a great language that offers great scale in big data and on cloud, but use newer versions designed for cloud and consider approaches to alleviate language challenges that cloud throws up, for example navigating compilation overheads by using ahead-of-time or outsourced just-in-time compilation. Savings on and off-cloud mean more re-use opportunities.
iii) Right-size costly queries. Focus on data, memory and compute intensive operations in particular. Some calculations – the ingestion, joining and combining of large time series data sets for example – can be incredibly onerous when run in relational databases, warehouses and data lakes. Consider using time series databases and analytics engines to reduce compute and save memory when running computationally onerous yet conceptually simple time series operations – aggregating and combining datasets, for example, or when bucketing time series data and trends for, say, trading strategies or defining and deploying time-oriented features into MLOps feature stores.
iv) Run the analytic where the data is and take advantage of efficient in-memory computation. Duplicating data into and from, say, cloud data warehouses means more memory use and higher costs.
In a customer case that combined iii and iv, a particular time series query running in a popular cloud data warehouse that hitherto cost them $50 per query (they run billions of that query over 300 billion records, servicing 220 million customers so you can imagine the scale) became just 50 cents per query using a time series in-memory engine, running on a 5th of the infrastructure and 5x faster. That has led to millions of dollars annual savings for them, infrastructure re-use opportunities, and more timely customer service. It also happens to be part of a cloud data pipeline so delivering FinOps, DevOps and, yes, MLOps advantages.
Financial Sector-Specific Challenges
Motivations for financial firms differ to their tech service providers and tech peers. Where costs directly correlate to specific service and product revenues in tech, in finance it is more complex. Siloed business units have different drivers and the aggregate cost to revenue picture gets blurred. In some critical parts of the industry, the need for speed and winning at all costs outweighs dull cloud accounting protocols, so Cloud Paradox addiction could unwittingly set in. While other sectors of the business could be more aware, good cloud cultures may not extend across the whole organization. The bank CFO and CIO has complexity to navigate, giving FinOps teams more challenges. Opportunities to aggregate costs, leveraging economies of scale across IT systems, and apply optimizations are non-trivial. But that also is a reason to pay more attention to it.
And above all, financial services sectors, especially those public complanies. have a reputation for being short term-focussed. Long term planning across financial businesses can be harder therefore than for tech firms with founders in it for the long term, some of whom I’ve been lucky enough to work with.
Overall, the Cloud Paradox likely impacts financial services today though it may not yet be appreciated. With that in mind, learn from those early tech adopters suffering from – and sometimes overcoming – cloud cost addiction. If and when they do, financial services firms willl realise lower costs while improving performance, and continue to leverage and enjoy the voluminous benefits of elastic cloud scale.
- SEO Powered Content & PR Distribution. Get Amplified Today.
- Platoblockchain. Web3 Metaverse Intelligence. Knowledge Amplified. Access Here.
- Source: https://www.finextra.com/blogposting/23556/the-cloud-paradox-and-four-ways-financial-services-firms-can-solve-it?utm_medium=rssfinextra&utm_source=finextrablogs