H&M Group has been making a lot of noise about artificial intelligence — and some of it is real. The group has confirmed investments in data-driven decision-making, generative AI for creative production, and tech infrastructure upgrades. But between a confirmed deployment and a press release about "exploring" a technology lies a significant distance, and that distance matters if you are trying to understand where one of the world's largest clothing retailers actually stands.
This piece audits what H&M Group has publicly disclosed, cross-references it against its own financial reporting, and separates what is live from what is still in the lab.
Key takeaways
- H&M Group has publicly committed capital expenditure of SEK 9–10 billion for tech infrastructure and store investment, with AI cited as a core driver of more accurate decision-making.
- The group's digital twin programme for models moved from announcement to first public image release, but the company describes its broader application as still being "explored."
- Generative AI is being used to amplify creative production, not replace it — H&M's own framing positions human teams as co-creators.
- The McKinsey State of Fashion report identifies AI adoption and geographic diversification as defining forces for the industry, a context in which H&M's disclosures read as cautious but credible.
- Investors and brand strategists should distinguish between infrastructure spend (confirmed) and specific AI product deployments (partially confirmed, partially exploratory).
What has H&M Group actually confirmed?
The financial commitment is real
In its full-year report published in January 2026, H&M Group stated that capital expenditure in comparable currency is planned at SEK 9–10 billion for 2026, with investments allocated primarily to the store portfolio and tech infrastructure. The report explicitly names "more data-driven decision-making and increased use of AI" as improving accuracy and giving the group "more tools for expressing creativity." That is a direct quote from the group's own financial disclosure — not a marketing brief.
What the report does not do is itemise specific AI systems, name vendors, or quantify the share of capex going to AI versus physical store upgrades. The commitment is real; the granularity is not there.
Demand forecasting: described, not detailed
H&M Group has spoken publicly about using AI to improve demand forecasting and reduce overstock — a chronic problem for volume fashion retailers. The full-year report frames this as improving "accuracy," which is consistent with machine-learning applications in inventory planning. However, the group has not published methodology, error-rate improvements, or before-and-after comparisons in any source available for this audit.
What you can reasonably infer: a retailer operating at H&M Group's scale, with the capex it is committing to tech infrastructure, is almost certainly running ML-assisted forecasting in some form. What you cannot infer from public disclosures alone is how mature that system is, whether it operates group-wide or brand-by-brand, or how it compares to peers.
Returns reduction: the gap between claim and evidence
Reducing returns is one of the most commercially and environmentally significant applications of AI in fashion retail — better size recommendations, more accurate product imagery, and smarter fit data all reduce the rate at which customers send items back. H&M Group has referenced sustainability and accuracy improvements in its communications, but has not published a specific returns-reduction figure attributable to AI in any source reviewed here.
This is worth noting because the returns problem is industry-wide. The McKinsey State of Fashion report, co-produced with Business of Fashion, identifies consumer value shifts and experience-led commerce as defining pressures — both of which make returns friction a strategic liability. H&M's silence on specific numbers here is not unusual; most retailers treat returns data as commercially sensitive. But it does mean the claim cannot be independently verified.
The digital twin programme: from announcement to first drop
What was announced
In March 2025, H&M told CNN it planned to create 30 digital twins of its models that year — AI-generated replicas built with the consent of the models themselves, who would be compensated as if they had appeared in a physical shoot. The group described itself as still "exploring" how these avatars would be used and said it was working with agencies and models to carry out the initiative responsibly.
What was delivered
By July 2025, H&M had moved from announcement to execution. The group released its first set of images featuring digital twins, set against fashion-capital backdrops and showcasing seasonal denim. A behind-the-scenes film accompanied the drop. H&M's own language is careful: the initiative is framed as "exploring emerging technologies like generative AI to amplify creativity and reimagine how we showcase fashion" — amplify, not automate.
That framing matters. It signals that H&M is positioning generative AI as a creative tool operated by human teams, not a replacement for traditional production. Whether that positioning holds as the technology matures — and as cost pressures intensify — is an open question.
The labour question
The digital twin programme attracted criticism from modelling agencies and worker advocates who raised concerns about the long-term impact on human model employment. H&M's response — consent, compensation, collaboration — addresses the immediate ethical question but does not resolve the structural one. If AI-generated imagery scales to replace a meaningful share of physical shoots, the compensation model for individual models does not offset the aggregate reduction in work.
This is not a reason to dismiss the technology, but it is a reason to read H&M's "responsible" framing with some scrutiny. The group is navigating a genuine tension between cost efficiency and worker impact, and its public communications lean toward the former while emphasising the latter.
Image search and customer-facing AI: what's in the product?
H&M's apps and website have incorporated visual search features — the ability to photograph an item and find similar products in the catalogue — for several years. This is a relatively mature application of computer vision in retail, and H&M is not unusual in offering it. The group has not published specifics about the underlying technology stack or the performance of these features in its public disclosures.
More broadly, personalisation — using browsing and purchase history to surface relevant products — is standard practice at H&M's scale. Again, the group's disclosures confirm investment in data infrastructure without specifying what is live versus in development.
How to read H&M's AI communications
The language of exploration versus deployment
A useful heuristic when reading any large retailer's AI communications: distinguish between the verbs. "Exploring," "piloting," and "testing" describe work that has not reached production scale. "Investing in," "improving," and "using" suggest something closer to live deployment. H&M's communications mix both registers — which is honest, but requires the reader to do some parsing.
The full-year financial report is the most reliable source, because it is subject to audit and regulatory scrutiny in a way that press releases are not. When the report says AI is improving accuracy and decision-making, that is a claim the group is prepared to stand behind formally. When a press release says the group is "exploring" a technology, that is a statement of intent, not a delivery milestone.
What the capex commitment tells you
SEK 9–10 billion in planned capital expenditure for 2026, split across stores and tech infrastructure, is a substantial commitment. It tells you that H&M Group is serious about the infrastructure layer — the data pipelines, cloud capacity, and engineering talent that AI applications run on. Infrastructure investment tends to precede visible product deployment by one to three years, which means the AI capabilities that H&M is building toward may not be fully visible in its customer-facing products yet.
For investors, this is the relevant signal: the group is building the foundation, not just announcing it.
Where the gaps are
Three areas where H&M's public disclosures leave significant questions unanswered:
- Demand forecasting specifics. How much has forecast accuracy improved? Which categories or markets benefit most? What is the error rate, and how does it compare to pre-AI baselines? None of this is public.
- Returns data. Has AI-assisted size recommendation or product imagery measurably reduced return rates? The group has not said.
- Vendor and technology stack. Which AI platforms, models, or partners is H&M working with? The group's disclosures are silent on this, which is common practice but limits external assessment.
What the industry context tells us
The McKinsey State of Fashion report, co-produced with Vogue Business contributor networks and BoF, identifies AI adoption as one of the defining forces reshaping fashion retail — alongside geographic diversification and experience-led commerce. Low growth is expected to persist across fashion and luxury retail globally, which makes efficiency gains from AI more commercially urgent, not less.
In that context, H&M Group's AI investments look less like optional experimentation and more like a structural response to margin pressure. A retailer that can forecast demand more accurately, reduce overstock, and lower returns rates has a meaningful cost advantage over one that cannot — and at H&M's volume, even small percentage improvements translate to significant numbers.
The question is not whether H&M should be investing in AI. It is whether the investments it has announced are delivering at the pace and scale its communications imply. On current public evidence, the infrastructure commitment is credible, the creative AI work has moved from pilot to first delivery, and the operational AI claims — forecasting, returns — remain asserted rather than demonstrated.
The bottom line for brand strategists and investors
H&M Group is doing more than talking about AI — it is committing capital, releasing work, and embedding AI language into its formal financial reporting. That puts it ahead of retailers whose AI strategy exists only in press releases.
But the gap between infrastructure investment and measurable operational improvement is real, and H&M has not yet provided the data needed to close it publicly. The digital twin programme is the most visible and verifiable initiative; the demand forecasting and returns work remains a stated priority without published results.
Watch the next two annual reports. If AI is genuinely improving forecast accuracy and reducing returns at scale, those improvements should eventually show up in gross margin, inventory turnover, and markdown rates — metrics that are disclosed and auditable. If they do not, the infrastructure spend will need a different explanation.
FAQ
What AI tools is H&M Group currently using? H&M Group has confirmed investment in data-driven decision-making and AI for demand forecasting and creative production, including a generative AI programme producing digital twin imagery of models. Specific tools and technology partners have not been publicly disclosed.
Are H&M's AI model digital twins actually in use? Yes. H&M moved from announcing the programme to releasing the first images featuring digital twins in mid-2025, accompanied by a behind-the-scenes film. The group describes broader application of the technology as still being explored.
Has H&M used AI to reduce product returns? H&M has referenced sustainability and accuracy improvements in its communications, but has not published a specific returns-reduction figure attributable to AI in any public source reviewed here.
What does H&M's capex commitment tell investors? The group has planned SEK 9–10 billion in capital expenditure for 2026, allocated to stores and tech infrastructure, with AI cited as a core driver. This signals serious infrastructure investment, though specific AI product deployments are not itemised in public disclosures.
How does H&M's AI strategy compare to the broader industry? The McKinsey State of Fashion report identifies AI adoption as a defining force for fashion retail amid persistent low growth. H&M's investments align with industry direction, but its public disclosures on operational results — forecasting accuracy, returns rates — are less detailed than its infrastructure commitments suggest they eventually will be.
Further reading
- H&M Group Full-Year Report 2025
- H&M's first generative AI digital twin image drop
- McKinsey State of Fashion