Two new studies from Substantive Research reveal contrasting commercial pressures across the financial information industry, with market data vendors seeking substantial premiums for AI-enabled access while investment research providers struggle to translate growing machine consumption into additional revenue. The findings highlight how artificial intelligence is reshaping the economics of financial information, raising questions about pricing power, licensing rights and the allocation of institutional data and research budgets.
The first study found that investment research budgets increased by just 0.4% globally in the first half of 2026, despite growing demand for AI-enabled research delivery. The second, conducted in collaboration with BCG Expand and TRG Screen, revealed that changes to AI licensing arrangements could increase some institutional market data agreements by as much as 50% in 2027.
AI Creates New Pricing Pressures
According to Substantive Research, 12 major data vendors have notified clients of pricing changes taking effect in 2027, affecting agreements already worth seven figures annually. Separate AI-related charges identified in 2026 proposals ranged from $30,000 to $1 million, while some initial negotiations reportedly began with demands for increases of up to 300%.The changes affect several areas of the market, including index data, ratings, terminals and ESG information. Requirements for direct data feeds into large language models (LLMs), Model Context Protocol (MCP) connectivity and enterprise-wide AI access are prompting vendors to introduce new licensing arrangements, often without established commercial precedents.
“What we’re seeing is an exceptionally fluid situation, and everyone is trying to figure this out on the fly in real time. You’re not seeing any emerging standardisation of approach,” Mike Carrodus, CEO of Substantive Research, the research and market data discovery and pricing analytics provider, tells Market & Alternative Data Insight. “We’re seeing three levels of inconsistency. The first is provider to provider, with each taking a different approach. The second is each provider being inconsistent in its treatment across its client base. And the third is each provider with each individual client. You might have been fine with a provider for six months, with them understanding that you’re not building products or training models, and then suddenly you get an email saying it’s an extra 80%.”
Vendors are responding partly to pressure from shareholders and financial analysts to demonstrate how they intend to monetise AI. For established providers with significant market share, the technology presents an opportunity to revisit longstanding commercial arrangements and introduce additional usage rights.
Institutional customers are pushing back. Substantive Research found that resistance to new index pricing models has resulted in average negotiated discounts of 34%, although the resulting costs are still rising by nearly 20% year on year.
Testing Vendor Pricing Power
AI could also change competitive dynamics in market data, as institutions explore whether cloud-based infrastructure and more flexible data integration make it easier to substitute alternative suppliers.
Providers of essential datasets, however, retain considerable negotiating power, particularly where switching costs are high or viable alternatives are limited.
“There’s going to be a group of vendors where the consumer market decides that, within certain parameters, there’s an added cost to taking advantage of this new technology, but they can make it work,” predicts Carrodus. “Then there will be a handful of firms where it’s very adversarial. The ones that are going to be more hawkish are those most confident about their ‘have-to-have-ness’. 2027 is going to be fascinating because everyone’s going to test the leverage in this market. There’ll be some big firms exploring alternatives, and if you’re a challenger data provider, you’ll be thinking this is your time to prove yourself.”
The negotiations come against a backdrop of steadily increasing expenditure. BCG Expand estimates that industry market data spending will rise a further 7% in 2026, while the average market data team now manages 34% more vendors than four years ago.
A Different Dynamic in Investment Research
Substantive Research’s survey of 50 major asset managers, representing approximately $20 trillion in assets under management, found little evidence that brokers and independent research providers have secured additional remuneration for delivering content directly into clients’ AI systems.
European research spending has fallen by more than 60% since 2018, while North American budgets have declined by over 50%. Although US research expenditure remains considerably higher, both markets face pressure to accommodate new technology requirements within constrained budgets.Investment research operates under different commercial incentives from market data. Banks may accept limited margins on research because maintaining relationships with portfolio managers supports their wider trading and investment banking businesses.
“Research, to its very core, is a relationship business,” observes Carrodus. “It’s about the rapport between a portfolio manager and a sell-side analyst as the client tries to come to a better decision. If I’m a research provider, I’m more inclined to let some of these industry trends play out and then have the conversation about what’s happening and what the client needs from us. But brokers have good answers to these questions. They’ll say that when we agreed this level of remuneration, you were only able to consume 15% of our output directly. Now your models can consume 100% of it and use it in ways that were impossible before.”
Some brokers are seeking additional payments for machine-readable research, citing delivery costs, intellectual property protection and potential liability associated with third-party information. Others are prioritising access to clients’ AI environments, recognising that exclusion could eventually threaten their relevance as research suppliers.
Carrodus expects negotiations involving a handful of more aggressive brokers to reach some form of resolution towards the end of 2026.
Analytics Spending Accelerates
While overall research expenditure remains largely unchanged, spending on AI-enabled research analytics and tooling increased by 40% over the past year, reaching 7% of average research budgets.
The growth reflects the ability of AI-driven discovery and analysis tools to help portfolio managers extract more value from existing subscriptions, potentially reducing their dependence on traditional research distribution channels.
“I think what some asset managers are discovering is that investing in analytics and tooling leverages their large spend on research enormously,” explains Carrodus. “If you buy the right tools, you leverage that research manifold. But the question is whether it’s a zero-sum game. If you’re a massive asset manager, you often put a lot of that tooling and analytics outside the research budget. You don’t necessarily want to take money away from the brokers because you want them incentivised to make you an early phone call and maintain those relationships.”
The growing overlap between research content, proprietary datasets and analytical technology is also complicating the distinction between expenditure on research and the tools used to consume it.
This has implications for commission sharing agreements (CSAs), particularly in Europe, where questions remain over which AI-enabled services qualify for commission-funded research expenditure.
Pressure on European Research Funding
Rising market data costs could place further pressure on European research budgets, potentially undermining regulatory efforts to reverse the decline in investment research following MiFID II.
The UK’s Financial Conduct Authority has introduced greater flexibility for asset managers to fund research through trading commissions, but adoption of the revised arrangements has been slow. Uncertainty over the eligibility of AI-enabled analytics and other services adds another complication.
“If I’ve got a market data budget that’s going up, and I’m having to find money from somewhere, my research budget is going to start to become endangered, especially if there’s ‘have-to-have’ stuff on the data side that I just have to pay for. Something’s got to give, and maybe I’ll take a bit more money out of the research budget,” says Carrodus. “Well, Europe’s politicians have decided they don’t want that. They want SME research back, and they want dynamism in the research markets. If people embraced the new regulations and ring-fenced research funding through CSAs, you wouldn’t end up with this diminishing fundamental equity research world.”
For data vendors, the challenge is determining how aggressively to pursue AI-related revenues without encouraging customers to seek alternatives. Research providers face a different calculation, balancing potential licensing premiums against the importance of remaining accessible within clients’ increasingly automated investment processes.
Carrodus compares the risk of brokers being excluded from clients’ LLMs to businesses failing to establish a website in the mid-1990s. As AI becomes an established channel for research consumption, providers may find that maintaining distribution and client relationships offers greater long-term value than maximising immediate licensing revenues.
Subscribe to our newsletter
