Anthropic’s expected mega-IPO later this month raises questions about ESG disclosure, corporate governance and public shareholder rights. A new report from Integrum ESG, drawing on Minerva Analytics’ governance data and research, examines potential issues arising from the listing.

This includes more than US$900 billion of disclosed compute and infrastructure commitments with no reported figures for emissions, energy consumption or water use. The report also finds that governance arrangements would leave effective control with Anthropic’s founders and Long-Term Benefit Trust after listing, including a structure under which a single Class F share held on behalf of the founders would carry 50.1% of the general vote.

“Anthropic PBC is set to become 1% of global equity market cap, so plenty of investors will soon own Anthropic through index funds and pension portfolios whether they have chosen to or not,” Shai Hill, Integrum ESG’s Founder and CEO told Minerva Analytics. “Its prospectus goes into significant detail on compute power, with U$900 billion of capex commitments. Yet it gives no figure for the energy it will take to run, or the water it will take to cool, its data centres.

“It names both as risks, along with local moratoriums on new data centres, but does not attempt to quantify them, even as a new legislative landscape in the US is starting to bring more opposition to new data centre construction,” he added. “Moreover, our partner Minerva Analytics' deep dive research clearly shows how much control stays with the founders and the trust after listing, disempowering the new institutional shareholders. Investors need to consider and price these ESG risks before the IPO.”

Anthropic’s environmental disclosure gaps: emissions, energy and water

Integrum ESG’s report found that Anthropic's IPO prospectus discusses AI infrastructure risks, including power availability, rising electricity costs, water consumption, climate-related impacts on cooling requirements and restrictions on data-centre development.

However, while those risks are described in detail, many of the metrics investors would typically use to assess them are absent. According to the report, Anthropic has not reported Scope 1 or Scope 2 emissions, quantified energy-consumption figures or water-use data. The analysis also notes that references to emissions are absent from the prospectus despite extensive discussion of infrastructure expansion and compute capacity.

The omission is notable given the scale of the company's disclosed commitments. Including purchase agreements and leases, Anthropic has reported approximately U$900.9 billion of compute and infrastructure obligations, with roughly U$451.6 billion expected before the end of 2030.

The report does not argue that Anthropic's environmental footprint is necessarily larger than that of its peers. Rather, it suggests investors currently lack the information needed to determine how the company's infrastructure impacts compare with those of other businesses operating across the AI ecosystem.

How Anthropic’s ESG disclosures compare with AI peers

To provide context, the report compares Anthropic's ESG disclosure practices with those of 14 organisations across the AI value chain, including chip manufacturers, cloud providers and data-centre operators. According to the analysis, 13 of the 14 peers report both Scope 1 and Scope 2 emissions, while most also publish energy-use and water-use data. Anthropic and SpaceX are identified as the exceptions.

Investors are seeking greater disclosure on electricity demand, grid capacity, power pricing and development constraints as public policy adds to uncertainty over AI expansion. Energy demand and data-centre development featured in the recent US midterm elections, putting the direction of policy in focus for companies reliant on large-scale computing.

Against that backdrop, Anthropic's reported valuation expectations have eased. Some investors earlier this year were reported to be seeking a valuation of as much as U$2 trillion; more recent expectations point to a listing below U$1.5 trillion, less than SpaceX's record valuation earlier this year. The change does not establish whether disclosure, governance or execution risks are driving investor sentiment.

Anthropic’s IPO governance: founder control and shareholder rights

Alongside its assessment of infrastructure-related disclosures, the report examines Anthropic’s post-IPO governance, including founder control and public shareholder rights.

Minerva has previously reported on Anthropic's plans for a multiple-class share structure ahead of its expected listing, highlighting the implications for founder control and shareholder accountability. At the same time, concerns over whether governance arrangements are keeping pace with AI-related risks have become increasingly prominent across the sector, echoing warnings raised in research from UK pension fund Railpen.

The report’s governance analysis, informed by Minerva Analytics' research and governance expertise, finds that a single Class F share held on behalf of Anthropic's seven founders will carry 50.1% of the general vote regardless of how many Class A shares are issued. Meanwhile, the company's Long-Term Benefit Trust will retain the right to elect four of the seven directors.

As a result, public shareholders will be unable to outvote the founders on matters submitted to shareholders generally and will have no say over the directors elected by the Trust. The report also notes that neither of these control mechanisms is subject to a time-based sunset provision.

The governance structure is significant because shareholder engagement is often one of the principal routes through which investors seek greater disclosure and accountability. In this case, the report suggests that shareholders will have relatively limited formal influence over both governance outcomes and future reporting practices once the company enters public markets.

What Anthropic’s IPO means for public shareholders

For investors following what is expected to be one of the year's most closely watched listings, the unresolved question is whether Anthropic will provide more quantitative information on its infrastructure-related impacts. The report concludes that investors currently have less of that information than is typically available from comparable listed organisations.

Until it is clear whether the disclosure gap will narrow after listing, investors face two assessments: how to judge substantial AI infrastructure commitments from narrative discussion rather than reported environmental metrics, and how to weigh a governance structure that concentrates control among the founders and the Long-Term Benefit Trust.

Taken together, the findings suggest that Anthropic's market debut is about more than investor appetite for AI. It is also an early test of how public-market investors approach the combination of limited quantitative disclosure and concentrated control that may increasingly accompany the next generation of AI companies.