Institute briefing
Open-Weight AI Models to Dominate Enterprise Workloads
New analysis predicts that the high costs associated with proprietary large language models will drive 80% of enterprise AI workloads towards more financially viable open-weight models, shifting greater governance and ethical responsibility onto businesses.
What happened
New analysis suggests that the economic burden associated with proprietary large language models (LLMs) will compel a significant majority of businesses to adopt open-weight AI models for their enterprise workloads. This shift is primarily driven by the escalating "token bills" – the costs incurred from processing data through commercial AI services. These expenses are projected to become prohibitive for many organisations, making open-source alternatives a more financially viable option for deploying AI at scale within a business context.
Why it matters
This projected migration towards open-weight models carries substantial implications for AI governance, ethics, and business strategy. While proprietary models often offer perceived advantages in terms of support and potentially refined performance, their cost structure could inadvertently push businesses towards solutions that require greater internal expertise for deployment, customisation, and ongoing ethical oversight. For leaders implementing responsible AI, this necessitates a re-evaluation of resource allocation, focusing on developing in-house capabilities for model evaluation, bias detection, and security, rather than solely relying on vendor assurances. The increased prevalence of open-source AI also underscores the growing importance of community-driven standards and collaborative efforts in establishing ethical AI practices.
By the numbers
- 80% of enterprise AI workloads are projected to shift to open-weight models.
The Institute take
The anticipated mass adoption of open-weight AI models due to cost pressures is a watershed moment, often framed purely as an economic decision. However, this trend fundamentally alters the landscape of AI risk and responsibility, shifting a greater burden onto the adopting organisations themselves.
While the financial savings of open-weight models are clear, businesses must recognise that this choice is not merely about reducing expenditure; it's about internalising more of the governance and ethical responsibility. Relying on open-source does not absolve an organisation of its duty to ensure fair, transparent, and secure AI. Responsible AI leaders must proactively invest in the internal expertise and processes required to rigorously evaluate, monitor, and manage these models, rather than assuming the 'open' nature equates to 'problem-free'.
Briefing notes
Questions this story answers
01Why are businesses expected to shift to open-weight AI models?+
Businesses are expected to shift to open-weight AI models primarily due to the escalating 'token bills' and prohibitive costs associated with proprietary large language models, making open-source alternatives more financially viable for enterprise workloads.
02What percentage of enterprise AI workloads are projected to move to open-weight models?+
It is projected that 80% of enterprise AI workloads will shift to open-weight models.
03What are the implications of this shift for businesses?+
This shift implies that businesses will need to develop greater internal expertise for deploying, customising, and overseeing open-weight models, taking on more responsibility for governance, ethical oversight, model evaluation, bias detection, and security, rather than relying solely on vendors.