Happening Now: ICONIQ Debuts Annual ‘State of AI Research' Report at the Cerebral Valley AI Summit London
The survey of 300 software execs shows trends in spending, R&D AI
In London, we’re in the midst of our mid-year Cerebral Valley AI Summit with leading AI founders, investors, and operators all in attendance.
On stage, ICONIQ General Partner Seth Pierrepont and Principal Vivian Guo are taking us through their 2025 State of AI Research report, which published just this morning. ICONIQ surveyed 300 executives at software companies building AI products. The market caps from these companies range from under $10 million in revenue to over $750 million last year.
You can check out the slides from the presentation happening now. Paid subscribers can see the full deck.
Some key takeaways:
Companies big and small are boosting their AI R&D. Startups with under $100 million in revenue and those with more than $1 billion in revenue are both allocating over 25% of their R&D budgets to AI development.
Leaders are still most concerned about the accuracy of LLM outputs, but cost is quickly becoming a concern as well. 74% of ICONIQ’s survey respondents ranked accuracy as one of their top three concerns, while cost came up high for 57%.
Builders are increasingly focused on AI agents and applications, rather than AI infrastructure: 67% of respondents mentioned they were building agents, and 59% were creating end-user apps, while only 44% were putting work into infra.
As AI tools become more standardized in the workplace, more budget is being allocated to infrastructure and less to talent. In the early pre-product days, AI talent accounted for 57% of spending, and infrastructure only 13%. But by the scaling phase, infrastructure jumps to 22% of costs while talent falls to 36%.
Employees still aren’t utilizing all of the AI tools they have access to, unless they are externally motivated to do so. The disparity is biggest inside companies doing $1 billion or more in revenue last year, where only 44% of employees appear to be actively using their AI tools.
The two biggest challenges around AI deployment are no longer technical. Instead, finding the right use cases for AI tools and measuring the ROI are the biggest hurdles.
See the highlights from ICONIQ’s report below:
Note: this post is too long for your email. Click here to see the full post.
