Hedge funds struggle to turn AI adoption into alpha

Hedge funds and other institutional investors are rapidly embedding artificial intelligence into their operations, but relatively few can yet demonstrate that the technology is improving investment performance, according to new research from Neudata.

Just 3% of data buyers surveyed for The Future of Alternative and Market Data 2026 said their firms were not using AI or large language models, showing the extent to which adoption has moved into the mainstream.

More than half, 56%, said they had measured a tangible return from AI deployment over the previous 12 months. However, efficiency rather than alpha accounted for most of the benefits: 41% cited efficiency gains and 19% cost savings, while only 17% credited AI with improving investment performance.

The proportion using AI-processed data to optimise investment strategies actually slipped from 31% in 2025 to 28% this year. Neudata said the findings suggest investment firms are increasingly using AI to support workflows rather than generate alpha.

Larger funds lead AI adoption

The survey also points to a divide between larger and smaller investment firms.

Buyers managing more than $1bn selected an average 2.3 of the five AI initiatives covered by Neudata, compared with 1.7 among firms managing less than $1bn.

Larger firms were particularly likely to use AI for internal task efficiency and chatbots, although more resource-intensive applications such as building proprietary models in-house were more evenly distributed.

LLMs remain the most widely adopted model type, used by 73% of respondents, while 34% use reasoning models. Some 35% license commercial foundation models to build proprietary applications.

The results are based on responses from 191 data industry professionals between June and August, including 69 data buyers and 122 providers. Quantitative funds represented the largest group of investment buyers, followed by multi-strategy, discretionary and macro managers, while 60% of buyer respondents managed more than $1bn.

Alternative data budgets rise

AI adoption is taking place alongside growing expenditure on alternative data.

Some 62% of buyers expect their alternative-data budgets to increase between 2026 and 2027, up from 57% last year. Another 35% expect spending to remain broadly unchanged, meaning 97% anticipate flat or higher budgets and none expects a significant reduction.

Buyers now subscribe to an average 21.4 alternative datasets annually, although the median is considerably lower at nine. They are also onboarding an average 10.1 new alternative datasets each year, suggesting managers continue to experiment extensively with new sources.

But managers remain selective. Most buyers converted fewer than a quarter of the datasets they trialled into paid subscriptions, with 39% identifying a lack of discernible investment signal as the main reason for rejecting a dataset, ahead of price at 25%.

Signal strength was also the factor data providers believed mattered most to investment managers, cited by 72%, up from 64% last year.

AI creates new data challenges

The rapid adoption of AI is also creating concerns over the underlying information being consumed by investment firms.

A third of buyers said they had seen data quality deteriorate over the previous one to two years. Among the 22 respondents reporting deterioration, 14 attributed it wholly or partly to AI-generated content, equivalent to roughly one in five buyers overall. Neudata cautioned that the relatively small sample means the results show divided opinion rather than a clear consensus.

Compliance is also becoming a bigger obstacle to acquiring new data. Some 30% of buyers identified compliance concerns as a challenge when trialling and onboarding datasets, up from 13% last year, while 23% separately cited compliance and legal review delays.

The research suggests that as AI becomes embedded across hedge fund workflows, the industry’s next challenge will be demonstrating whether those efficiencies can translate into sustainable investment returns.