On August 24, 2026, fitness technology firm Keep unveiled its interim results for the period ending June 30, revealing a set of figures that capture a company in transition. Total revenue reached 825 million yuan, a marginal 0.4% increase year-on-year, while net losses narrowed sharply from 35.43 million yuan to 12.19 million yuan—a substantial 65.6% reduction. On a non-IFRS basis, adjusted net profit stood at 5.88 million yuan. Although revenue remained almost flat, the significant narrowing of losses masks a deeper structural overhaul taking place within the business. Revenue from online memberships and paid content dropped 26.9% to 247 million yuan, whereas revenue from self-branded sports products surged 21.7% to 483 million yuan, lifting its share of total revenue from 48% to nearly 60%. Concurrently, average monthly active users fell from 22.49 million in the prior-year period to 18.58 million—a decline of over 3.9 million users. This half-year report reads less like a scorecard of past performance and more like a declaration of strategic intent. Keep is accelerating its transformation from a subscription-driven online content platform into a dual-engine sports technology company powered by AI and consumer goods.

Consumer products take center stage: the retreat of high-margin digital services and the rise of physical goods

The revenue mix for the first half displays a clear divergence across segments. Self-branded sports product revenue climbed 21.7% to 483 million yuan, approaching 60% of total revenue, while online membership and paid content revenue fell 26.9% to 247 million yuan, shrinking to roughly one-third of the total. Advertising and other revenue edged up 9.3% to 96 million yuan. In the same period last year, these two core businesses were much closer in size. Within just twelve months, Keep has pivoted its operational focus from selling content to selling goods. This shift is driven directly by momentum in the consumer goods segment: sports equipment revenue grew 49%, accounting for more than 60% of consumer product sales, while gross merchandise value for the three core categories—muscle building, body shaping, and yoga—expanded by 63%, 49%, and 33% respectively. Sports nutrition revenue increased 39%, with new products contributing over one-third of that figure. Distribution channel revenue rose 35%, while Douyin channel growth exceeded 50%. Overseas markets also delivered initial results, generating over 22 million yuan in revenue during the half. Gross margin for consumer products improved by 5.3 percentage points from 34.8% to 40.1%. Yet even with that improvement, the 40.1% margin for physical goods remains well below the 73.0% margin achieved by online memberships and paid content. The declining share of high-margin digital revenue combined with a rising mix of lower-margin physical products dragged overall gross margin down from 52.2% to 51.1%, with gross profit falling 1.6% year-on-year to 422 million yuan. More telling is the quality of earnings. Although the interim loss narrowed from 35.43 million to 12.19 million yuan, the financial report attributes this primarily to a 23.7 million yuan reduction in share-based compensation expenses. In other words, the narrower loss stems largely from one-off or non-operating cost reductions rather than improved profitability in core operations. Adjusted net profit, which better reflects underlying business performance, slipped from 7.47 million yuan to 5.88 million yuan, a 21.4% decline year-on-year, with an adjusted net margin of just 0.7%, down from 0.9%. These figures expose a fundamental tension: Keep has used a higher proportion of consumer product sales to stabilize revenue, but those products' thinner margins and heavier operating costs—inventory, logistics, and channel investments—are eroding overall profitability. The drop in gross margin from 52.2% to 51.1% is the clearest evidence of this structural change. Meanwhile, declining user numbers cast a shadow over the recovery prospects of the membership business. Average monthly active users fell to 18.58 million, down more than 3.9 million from 22.49 million a year earlier, while average monthly paid subscribers dropped from 2.79 million to 2.17 million, and membership penetration slipped from 12.4% to 11.7%. Compared with 2024 figures, the platform's average MAU has contracted by nearly 40%. Management explained during the earnings call that the user decline primarily reflects the exit of lower-intent exercise users, noting that current membership benefits lean toward content-oriented users and fail to fully serve tool-driven users focused on running or outdoor activities. While the quality of engagement among retained users is improving—ARPU rose from 6.1 yuan to 7.4 yuan, up 21.3% year-on-year, and average monthly exercise time per MAU grew 15.3%—the continued shrinkage of the user base poses a fundamental constraint on long-term membership revenue growth. Keep now stands at a crossroads: relying on consumer products can sustain revenue scale and near-term cash flow, but a low-margin, asset-heavy model struggles to support the valuation logic expected of a technology company, while reviving the high-margin membership business depends on solving user acquisition and retention challenges.

AI promises and payoffs: faster technical deployment, commercialization still unproven

If consumer products are the ballast keeping Keep stable today, AI is the new engine it is betting on for the future. The first half of 2026 saw intense activity in the AI arena. In April, Keep launched Keepace.ai, its proprietary vertical large language model for sports and health, built on a decade of accumulated movement data, sports science literature, and expert knowledge. The model focuses on three core capabilities: training course generation, sports knowledge Q&A, and exercise data interpretation. By August, Keepace.ai had completed national and Beijing municipal filing for large model registration. Around the same time, Keep released MoveBench, a benchmark for evaluating sports and health large models, in which Keepace.ai ranked first among all tested models in both curriculum generation and sports Q&A. On the deployment front, Keepace.ai has been integrated across the app's pre-exercise, in-exercise, and post-exercise workflows, with over 8,000 AI-generated customized courses now available, and an AI voice-guided running companion delivering real-time interactive feedback. Daily token consumption across the app more than doubled from 8.4 billion in March to 18.5 billion in July. By technical metrics, AI adoption and usage frequency are accelerating rapidly. However, a significant gap remains between technology rollout and commercial monetization. The most striking contradiction: AI usage has doubled, yet membership revenue continues to contract—online membership and paid content revenue fell to 247 million yuan, down 26.9% year-on-year. The abundance of AI-generated personalized courses and intelligent coaching features has yet to translate into stronger renewal intent or increased paying users. As market analysis has observed, the sports large model now extends into training plans, Q&A, voice coaching, and data interpretation—no longer confined to product launches—but commercial outcomes have not yet followed in step. Another avenue attracting expectations is B-end commercialization. Keep states that its sports health large model is well suited to meet the AI service needs of sports hardware manufacturers, insurance companies, and healthcare providers. The combination of its sports science resource library, an already-filed vertical model, and proven agent capabilities positions it as a potential AI solutions provider for the sports health industry. Yet this B-end business remains in its infancy, and its actual contribution will take time to validate. One notable detail: while AI strategy has been elevated to a core priority, R&D expenses fell 23.2% in the first half, raising questions about the substance of the AI narrative. If AI is genuinely the next-phase growth engine, does the decline in R&D spending signal ongoing tension between short-term profitability and long-term investment? Institutional sentiment reflects cautious acknowledgment of the transformation. Huatai Securities maintained a buy rating but slashed its target price from HK$7.40 to HK$3.32, citing the transition phase and the need for time to verify steady-state margins. The target price-to-sales multiple was cut from 1.5x to 0.8x—a notable discount against the 2.9x peer average for 2026. This divergence between rating and target price underscores the market's wait-and-see stance: directionally constructive, but uncertain about timing. Keep has indicated that in the second half it will continue advancing Keepace.ai across customized courses, in-exercise guidance, and post-workout reviews, expand multimodal capabilities, and explore monetization for AI memberships and B-end services. The company has set a goal of achieving a qualitative leap by 2027. This suggests that 2026 serves as a crucial bridge period from technical validation to commercialization, with the market required to exercise patience before AI investments convert into tangible revenue and profit. Keep's transformation is, at its core, a genetic recombination from a content platform into an AI-plus-consumer-goods platform. Consumer products underpin the revenue base, while AI carries the long-term narrative—but whether these two pillars can move in concert will determine whether Keep can truly emerge from its current predicament. Slight revenue growth, user attrition, and margin pressure are the short-term numbers behind a company mid-course correction, and the pain that accompanies such a shift. Whether the AI bet pays off by 2027 will be the decisive factor in regaining market confidence.