CASE STUDY: Inside Anta’s AI365 and Its Artificial Intelligence Strategy

The strategic question

Anta Sports grew into China’s largest sportswear group on scale: a multi-brand portfolio, thousands of stores and a decade of steady share gains against Nike and Adidas at home. In October 2025, the group added a second axis to that story. At the second conference of the Sporting Goods Industry Innovation Consortium, an industry body Anta convened and leads, the group formally launched AI365, a plan to route artificial intelligence through design, marketing, merchandise, supply chain, retail and operations, with a stated target of creating more than RMB5 billion (€630m) in value within three years.

What separates AI365 from much of the AI language circulating across the sporting goods sector is Anta’s willingness to attach operating metrics to the initiative, even where many of those metrics fall short of filing-grade disclosure. This case study asks a narrower question: how much of that reported value has actually shown up in Anta’s own numbers, and how much still rests on figures the group has not put in a filing.

1. The strategic logic: to take AI across the whole business
2. The build: implementing the framework
3. How the model works
4. The market tailwind
5. Competitive implications
6. The open question, and a takeaway

1. The strategic logic: to take AI across the whole business

Anta designed AI365 as a systematic effort to route AI through the group’s entire operating structure at once, rather than a pilot confined to one department that might later expand. The plan sets three directions: AI aimed at cutting costs and raising efficiency, AI aimed at driving growth, and AI aimed at improving the user experience. Those directions run across six declared domains, marketing, design, merchandise, supply chain, retail and operations, underpinned by five support commitments covering mechanism, team, culture, budget and outside collaboration.

The group set a three-year roadmap for execution. 2025 is the “Good Product” phase, concentrated on AI in design. 2026 shifts to “Good Retail,” pairing AI marketing with AI-enabled retail operations. 2027 is scheduled as “Good Operations,” aimed at reshaping internal business processes for scale efficiency. Chairman Ding Shizhong has called the plan an extension of an R&D budget that has already exceeded RMB20 billion (€2.52bn) over the past decade, with a further RMB20 billion planned through 2030.

2. The build: implementing the framework

Anta’s AI effort did not start in October 2025. The Sporting Goods Industry Innovation Consortium, the body that hosted the AI365 launch, was itself convened by Anta in October 2024 together with Tsinghua University, Donghua University and several state research institutes, initially to pool resources on materials science. AI became a second pillar of the consortium’s work over the following year, as Anta expanded its internal AI team with hires from major technology companies and universities.

The October 2025 event paired AI365 with Linglong, a proprietary design model Anta describes as the sporting goods industry’s first vertical AI design tool, trained on more than three decades of its own footwear and apparel data. By the end of 2025, Anta had attached a first set of commercial figures to the strategy, covered below. In May 2026, the group added a second named model, Lingxi, extending AI365 into retail styling and cross-border livestream commerce, formally launched at the consortium’s Third Council meeting in Xiamen.

3. How it works

Linglong, product design.

Anta’s design model can turn a sketch or written brief into a finished design draft in roughly 15 seconds, according to the group’s own interim report. Business press coverage of the October 2025 launch describes a specific case: a FILA tennis shoe that historically took about three months from brief to final design, compressed to under 40 days once designers began working with Linglong, with the model generating 56 concept options from a single brief and a dozen colorway variations in five minutes. Anta has also disclosed an 80 percent reduction in the cycle for expanding a hit product into a fuller size and colorway range, and a 30 percent increase in the rate at which first-round design concepts are approved.

Lingxi, retail styling.

Launched in May 2026 and trained on roughly six years of Anta’s own product photography, Lingxi generates outfit recommendations and a virtual try-on rendering from a scenario description or an uploaded photo. In a trial at Tmall stores, Anta has reported a more than 10 percent year-on-year drop in returns, a more than 10 percent increase in conversion, and a 20 percent rise in click-through rate. 

AI digital-human livestreaming.

Across Anta’s retail network, AI-generated hosts ran more than 100,000 hours of livestreaming in 2025, watched by more than 30 million viewers, concentrated between midnight and 7am, the window when human staff are off shift. By early 2026, more than 80 stores were running the format, including a FILA FUSION flagship store that Alibaba’s Taobao training center has cited as a benchmark case.

On model architecture: the consortium’s own announcement of the launch states directly that Linglong is built on a foundation model of roughly a hundred billion parameters, trained on Anta’s proprietary design data. 

Anta does not break out AI365 spending as its own line item in any financial filing, so R&D expense is the closest verifiable proxy available for how much the group is actually committing to the initiative, even though it also covers materials science, product engineering and other R&D unrelated to AI. The table above establishes that baseline: what Anta spent and earned in the period AI365 was meant to be running, against which the AI-specific figures below can be read.

More specifically, the AI-specific commercial figures Anta has disclosed alongside that R&D spending run well ahead of the underlying revenue growth rate:

• AI-assisted design order value: more than RMB2.5 billion (€315m) as of the October 2025 launch, rising to more than RMB9 billion (€1.13bn) by the end of 2025
• Digital-human livestreaming: more than RMB300 million (€38m) in 2025 GMV, from more than 100,000 hours of AI-hosted livestreaming
• Three-year AI365 target: more than RMB5 billion (€630m) in cumulative value creation, against more than RMB20 billion already invested in R&D over the previous decade and a further RMB20 billion planned through 2030

None of these AI-specific figures appear in Anta’s annual or interim financial statements. They originate from consortium-linked disclosures and media coverage of consortium meetings, not from filed accounts, and should be read with that distinction in mind.

4. The market tailwind

Anta’s growth also came during a period when several peers reported weaker trends. Li Ning posted low single-digit growth, while Nike’s Greater China business continued to contract. Management has presented AI365 as a way to protect margin and inventory efficiency, more than as a growth engine, at a moment when growth itself is harder to find across the category.

5. Competitive implications

Within China, Li Ning has turned to AI mainly for supply chain scheduling and replenishment, a narrower and less publicized application than Anta’s. Xtep offers virtual try-on at the retail level, 361 Degrees has invested in flexible, small-batch production lines, and Bosideng, the down-apparel giant, announced its own named vertical model in July 2026, described by chairman Gao Dekang as spanning design through fitting and delivery.

Among the top European brands, the pattern looks different in kind, not just in scale. Adidas has deployed AI mainly in commerce infrastructure, an AI-agent-powered store-building service for B2B partners, built on Salesforce’s platform. PUMA has used AI for customer-facing marketing and service, an in-store concierge device and a fan design co-creation tool built openly on a Stable Diffusion base model. Neither has claimed a single proprietary model spanning design through retail the way Anta and Bosideng have.

The structural advantage behind Anta’s approach is data: Linglong and Lingxi both draw on decades of proprietary design and product-image data that a newer or smaller competitor would struggle to replicate. But data is only half of it.

The other half is scope: Anta built AI365 to reach across the entire value chain, design, marketing, merchandise, supply chain, retail and operations, at once, rather than a single consumer-facing feature or an isolated point solution the way most of its rivals, Chinese and European alike, have approached AI so far.

6. The open question, and a takeaway

Interim ANTA brand chief Lai Shixian has said, on the record, that it remains too early to know how much AI will actually improve Anta’s operating efficiency. That is a notably cautious statement for a company that, in the same period, disclosed an 80 percent reduction in its product development cycle, a 30 percent increase in design-approval rates, a 400 percent rise in creative output, and billions of RMB in AI-attributed order value.

It’s possible both statements are true at once. A faster design cycle doesn’t automatically mean lower costs, healthier inventory, or better margins; it only becomes evidence of those things once enough cycles have run through the system to show up in the numbers Anta reports to shareholders. Yet from the point of view of our readership, the substance that matters isn’t whether every disclosed percentage survives an audit.

It’s facing a competitor that can put AI to work at this depth, building its own model rather than licensing one, and deploying it across countries, brands and controlled companies at once.

That is the more durable fact here, and the harder one for rivals to answer. Whether AI365 delivers the specific efficiency gain Anta has promised is still an open question, by the company’s own admission. Whether Anta now has the infrastructure, the data, and the organizational will to keep building on this scale is not.

Eastern views

Asia-Pacific insights for the sporting goods industry

Analysis, insights, and expert perspectives on the sporting goods industry across Asia-Pacific — covering market trends, manufacturing, retail, and brand strategy from China to Southeast Asia to Oceania. With Jakarta-based contributing editor Yohana Belinda.

Explore Eastern views →