Businesses still struggling to manage data budgets, deliver ROI when it comes to AI

Enterprises are pouring cash into infrastructure to support their AI efforts, but are still struggling to contain data storage costs and even to access their data, research from Wasabi has shown.

The cloud storage vendor found that two thirds of the investment companies were making in AI went on infrastructure for data and compute, with a third going on software/SaaS. And the majority of companies – 60 percent – are planning to increase their spending. 

Perhaps unsurprisingly, cloud storage accounts for a large chunk this spend, with almost two thirds of companies saying they use hybrid storage to support AI workflows. Meanwhile, 81 percent of companies said they use more than one public cloud provider for their object storage capacity.

Andrew Smith, director of strategy and market intelligence at Wasabi, said the focus on infrastructure was “the complete opposite” of the traditional cloud market, where the focus is on software and services.

“It’s a great illustration of the critical role cloud storage and cloud infrastructure services play in this generational buildup of AI-enabled solutions and services.”

However, while tech leaders are clearly well versed in the use of the cloud to underpin their AI efforts in general, and storage efforts in particular, they’re still struggling to maintain costs.

Almost half, worldwide, said they exceeded their budgets for cloud storage last year. Moreover, this didn’t necessarily equate to more capacity. 

That’s because fees and add-ons are eating up their budgets, accounting for half of their cloud storage spend. This ranges from egress, and API-based data operations fees for reads, writes, and lists, though Wasabi also flagged up “lesser-known fees” for operations like data retrieval, object lock, and replication requests.

If anything the problem is getting worse, with fees accounting for 47 percent of spend in 2024 and 48 percent in 2023.

This contributed to 49 percent of respondents fessing up to exceeding their budget, with over 15 percent saying they did so “massively”.

Higher than expected data ops fees payments were on a par with higher than expected storage usage as a reason for busting the budget, the research showed, cited by over 40 percent of respondents as a reason they busted their budget.

Despite all this spend on data services – and accompanying fees – in the pursuit of AI, companies still struggle to access, analyze or fully utilize their data. This can range from logs, call recordings, and chat, to email, video, and document archives. 

Just 3 percent believe they have full access to their data, with 25 percent saying up to a quarter of their data is “dark”. Over a third say a quarter to half of their data is dark, while a quarter struggle to access between half and three-quarters of their data. And a quarter say 75 to 99 percent of their data is dark.

Which may in part explain why just under a third of respondents are seeing a positive return on their AI investments right now. However, just over a half expect to deliver a positive ROI on AI projects in the next 12 months.

Wasabi’s own service promises no egress fees, an offer that was compelling enough for it to close a $70 million E round funding in January, bring total investment in the firm to $600 million, and valuing it at $1.8 billion.