A new analysis by the Sierra Club, released on 28 October, casts a stark light on the environmental cost of artificial intelligence’s data infrastructure. The US environmental group found that the planned surge in fossil-fuelled power plants to run AI data centres could push the US power sector’s carbon emissions to 2.1 billion metric tons by 2030, a 42% increase from 2025 levels ([1]). That jump would effectively wipe out 15 years of emissions reductions from the grid. The findings underscore a hidden cost of the AI boom: the massive energy consumption and carbon footprint of the data centres required to train and run large AI models. As companies race to scale up their AI capabilities, this climate impact could invite stricter regulations or community opposition, adding new risks and costs to AI projects.
Those community frictions are already emerging. A University of Houston survey of likely voters in Texas, where data centre construction has become a campaign issue, found 79% support limits on new data centres in their state ([2]). A near-unanimous 99% of respondents believe any new facilities should pay for required electric grid upgrades ([3]). Texans are worried that the rising power and water demands of huge AI-focused server farms will increase utility costs and strain local resources ([4]). Candidates in the state’s upcoming November elections have taken note of these concerns, tempering their support for unbridled data centre expansion ([5]). For tech leaders, these warnings highlight that the path to AI at scale runs through not just technical milestones, but also sustainability and public trust. Failing to address environmental and infrastructure impacts could stall critical data projects or lead to expensive compliance requirements as regulators and communities push back.
Even as companies invest heavily in AI tools, most are constrained by a lack of data readiness. A global survey of 10,000 businesses found that 97% of organisations have active AI initiatives, but only 5% say their data is ready to support them ([1]). This Dun & Bradstreet survey, released in May, reflects the messy reality behind the AI hype: enterprises are eager to experiment with generative AI and other advanced tools, yet 95% admit their data, from quality and integration to governance, isn’t fully prepared for scalable AI. Over half of those polled cited data access and quality issues as top barriers to moving from experimentation to real operational AI ([2]). The result is a growing ‘readiness gap’ where AI projects proceed on shaky data foundations, increasing the likelihood of errors, bias and compliance failures.
Other research confirms the disconnect. In a study of 2,000 companies by Accenture this year, a mere 7% of firms (dubbed “Data Reinventors”) have built the necessary AI-ready data capabilities to scale advanced AI across the enterprise ([3]). Meanwhile, a survey of 600 data leaders found widespread overconfidence: 88% believed their data was AI-ready, yet 43% identified data readiness as a primary obstacle to achieving AI outcomes ([4]) ([5]). The implication for executives is clear: a significant number of organisations are engaging with AI without the solid groundwork of accurate, well-governed data. This overestimation of readiness can lead to wasted investments and underwhelming results, as AI initiatives falter due to unaddressed data quality issues and fragmented data ownership.
The surge of autonomous AI and generative models is also testing corporate governance and data accountability, with regulators increasingly alarmed at the gap. In mid-October, IBM released new research showing 80% of surveyed executives in Europe and the Middle East experienced an AI-related incident reaching their board level in the past year ([1]). Yet only 40% of those organisations updated their AI governance policies in response ([2]), revealing a troubling inertia in adapting risk controls. This suggests that many companies lack clear ownership and oversight for AI and data issues, even as such incidents, from model errors to data leaks or compliance breaches, become more common.
Regulators are moving to enforce stronger accountability. On 26 October, the UK’s Financial Conduct Authority (FCA) reiterated that it expects financial firms to treat major AI failures and ‘near misses’ like any other operational incident, reporting them for review and remediation ([3]). The FCA’s Chief Data, Information and Intelligence Officer warned that “poorly governed AI can create consumer harm, cyber vulnerabilities, and operational failures”, urging that greater autonomy must come with clear responsibility ([4]). Similarly, earlier this month the UK Information Commissioner’s Office secured commitments from ten leading AI firms, including OpenAI, Google, Apple and Meta, to tighten their handling of personal data and be more transparent about how they use it ([5]). For enterprise leaders, these signals mean that data governance and stewardship cannot be an afterthought. Ensuring someone is accountable for the quality and ethical use of data is now essential to avoid regulatory penalties and to maintain trust when deploying AI at scale.