A new investigative report has reignited debate over artificial intelligence and the planet.

Consequently, environmental NGOs now accuse Big Tech of large-scale Greenwashing.

The analysis, released 17 February 2026, reviewed 154 statements promising AI would cut Climate emissions.

Joshi found nearly three quarters lacked credible evidence, labelling the optimism a potential Hoax.

This article unpacks the data, competing narratives, and policy implications for enterprise leaders.

Moreover, rising data-centre electricity use suggests AI might expand, not shrink, the sector's Carbon load.

Industry leaders counter that smart algorithms optimise grids, logistics, and materials discovery.

Nevertheless, absent transparent reporting, investors struggle to separate legitimate savings from marketing spin.

Therefore, understanding the evidence gap has become a boardroom priority.

Read on for a data-driven tour of AI's contested Climate record.

In contrast, regulators in Europe and Asia already draft disclosure rules that could change corporate incentives.

Such frameworks may shift how quickly Greenwashing narratives collapse under scrutiny.

Report Uncovers Industry Hoax

Joshi's study was backed by Friends of the Earth U.S. and five partner groups.

Analysts checked 154 published statements touting AI as a net Climate benefit.

They judged 74% unproven, 26% partially supported, and zero fully validated.

• 36% of statements cited no sources.
• Only 26% linked peer-reviewed research.
• Most reused a 5–10% Carbon reduction claim from BCG.

In addition, only four statements supplied replication code or raw datasets.

Consequently, NGOs called the AI climate promise a dangerous Hoax that distracts from urgent decarbonisation.

These findings reveal sizable evidence gaps.

However, understanding the sources behind those gaps requires deeper inspection, explored next.

Evidence Behind Bold Claims

Corporate sustainability pages lean heavily on consultancy estimates, chiefly BCG's 2021 client experience note.

BCG suggested AI could trim global Carbon emissions by 5–10%.

However, the figure originated from anecdotal projects, not a peer-reviewed model, exemplifying Greenwashing.

The International Energy Agency later modelled a Widespread Adoption Case, projecting a 1.4-gigaton cut by 2035.

IEA simultaneously warned that data-centre electricity demand may almost double in the same period.

In contrast, independent academic work estimated AI already emitted up to 79.7 million tonnes Carbon last year.

Moreover, the IEA model flags rebound effects that could erase half the projected savings.

Evidence therefore appears mixed and fragmented.

Consequently, stakeholders should scrutinise AI's actual Footprint, addressed in the following section.

Growing AI Emissions Footprint

Generative models such as ChatGPT demand extensive parallel compute during training and inference.

Digiconomist estimated AI's 2025 Footprint at 32.6–79.7 million tonnes CO2 and 312–765 billion litres of water.

Moreover, most operators hide workload level data, complicating verification.

That opacity fuels Greenwashing across sustainability reports.

Meanwhile, grid planners fear surging datacentre loads will crowd out renewable integration.

• Training a large model can consume as much electricity as 100 U.S. homes yearly.
• Water usage often equals thousands of households per day.
• Cooling demands rise in warmer regions, worsening local strain.

Researchers note that water stress hotspots align with planned hyperscale campuses in arid regions.

Nevertheless, companies still argue efficiency gains will offset growth.

Current trajectories suggest the opposite.

However, narrative framing remains powerful, as the next section shows.

Corporate Narratives Face Scrutiny

Marketing materials often merge low-energy optimisation successes with high-energy generative pipelines.

Consequently, many executives present blanket Greenwashing claims without disclosing model mix or load factors.

Google cites the BCG percentage while building massive GPU clusters for generative research.

Microsoft repeats similar language yet remains silent on water consumption at desert campuses.

Furthermore, press interviews reveal some executives conflate efficiency case studies with speculative generative breakthroughs.

Jon Koomey warns journalists to distrust self-interested projections without granular data.

Sasha Luccioni meanwhile urges firms to distinguish between model classes before promising sweeping Climate benefits.

In contrast, watchdogs urge separating model classes during any public emissions disclosure.

These voices expose credibility gaps.

Therefore, pressure for transparent metrics keeps building, as outlined ahead.

Transparency Demands Gain Momentum

Investors, regulators, and researchers now push for AI-specific energy, water, and Carbon disclosures.

The European Commission has drafted rules that would mandate workload level reporting by 2028.

Similarly, California lawmakers consider bills tying data-centre permits to transparent Climate metrics.

NGOs recommend adopting verifiable protocols like the Green Web Foundation's open method.

Professionals can enhance their expertise with the AI Developer™ certification.

Moreover, standard setting bodies propose including rebound effect calculations within assurance audits.

Global pension funds now demand science-based targets and threaten divestment when claims lack evidence.

Opaque filings risk regulatory penalties for Greenwashing.

Disclosure rules would shrink informational asymmetry.

Subsequently, organisations could pursue genuine Sustainability strategies, discussed in our next section.

Consequently, several cloud vendors pilot real-time energy dashboards for enterprise customers.

Toward Verifiable AI Sustainability

Achieving real Sustainability outcomes demands concurrent efficiency improvements and workload restraint.

Energy researchers advise targeting smaller, task-specific models, alongside 24/7 renewable matching.

Additionally, teams should adopt efficient coding patterns that cut redundant compute cycles.

IEA scenarios suggest such measures could halve projected data-centre growth within a decade.

Furthermore, companies should disclose training hours, inference calls, and regional grid mixes.

Rebound effects must enter planning forecasts to prevent unintentional Carbon growth.

• Publish AI energy dashboards quarterly.
• Provide water consumption normalized per 1,000 inferences.
• Commit to independent audits against Sustainability targets.

Nevertheless, corporate culture must also reward researchers for efficiency, not only model scale.

Otherwise, Greenwashing will persist unchecked.

Real progress pairs transparency and restraint.

Consequently, the final section recaps major lessons and next steps.

Final Thoughts And Action

Joshi's analysis underscores how easily optimistic rhetoric morphs into Greenwashing when evidence remains thin.

Most public AI Climate claims rely on consultancy web articles or corporate blogs, not peer-reviewed science.

Meanwhile, emissions and water Footprint already rival those of mid-sized nations.

Stakeholders must demand granular metrics, adopt verifiable standards, and resist hype.

Therefore, executives should prioritise smaller models, renewable procurement, and audited reporting.

Professionals exploring strategic roles can validate their skills via the earlier linked AI Developer certification.

Ultimately, rigorous disclosure will convert aspirational Sustainability messaging into measurable progress.