Daniel Ek: Life After Spotify, Broken Healthcare Incentives, Catching Disease Early & AI's Potential

After spending roughly 12–13 years focused on Spotify, Daniel Ek became its executive chairman on January 1 of the year discussed and stepped away from day-to-day operations. He says he prefers building companies to investing in them. His next major effort is Neko, which he founded in 2018 after questioning why healthcare spending kept rising while outcomes worsened. Following five years of product development, Neko launched in Sweden in 2023, then expanded to the UK and the US.

Ek sees healthcare’s central failure as an incentive problem. A system shaped around infectious disease tends to reward treating acute problems after symptoms become serious, rather than investing in prevention over many years. Because people often change employers and insurers, a payer may not capture savings from preventive care that takes 10–20 years to pay off. Ek argues that making preventive investments cost tens of thousands of dollars rather than millions could make them easier to fund. He cites US healthcare spending of 18% of GDP and the hundreds of billions of dollars spent annually on heart disease, while acknowledging he is not an expert on the US system and that agreement on prevention does not mean agreement on how to deliver it.

Neko’s approach is to combine multiple measurements over time with low-cost sensors and AI, aiming to identify risks early enough for clinicians to intervene. Ek compares the data-to-prediction model with Spotify’s recommendation system, while emphasizing that healthcare requires clinicians as well as AI. Neko’s Gen 2 assessment costs $499 and includes 53 blood markers, more than 6,000 high-resolution skin images to index moles and lesions, and measures of heart health, circulation and grip strength. A visit is designed to take about an hour, including a clinician consultation, though a member without questions may finish the screening in 30–40 minutes. Neko also accepts imported Apple Health data, including some wearable measurements.

For skin checks, AI flags possible mole risks for review by a clinician and expert dermatologists, and Neko catalogs moles across visits to track changes that can be difficult for doctors to remember. Ek says the average member has 950 moles, with wide variation. By the interview, Neko had conducted more than 100,000 scans and reported finding a serious, previously undiagnosed condition in around 1% of members. The company recommends annual screening, but Ek says its scan base is too small to establish population-level health outcomes or conclusively measure any effect on healthcare spending.

Across three years of studies, Neko found that members in the worst health improved the most. Ek links those gains to clear results and clinician guidance; members’ reported changes often concern stress, diet and sleep, and some have quit smoking after prioritizing their health. Neko had completed four clinical trials, with two underway and another four planned, and releases findings annually. Ek expects datasets 10 or 100 times larger to support important conclusions about interventions across populations and countries. The company upgraded its diagnostic infrastructure to Gen 2 within three years and plans to keep adding diagnostics.

Ek says Neko’s roughly $500 annual price has positive unit economics and some operating clinics are already profitable, attributing lower costs to vertical integration. The company was operating only in New York, with its first location at 300 Lafayette, and planned to add New York locations, launch in Miami and Washington, D.C., and attempt a wider US rollout within 12–24 months. On AI more broadly, Ek sees substantial positive and negative effects, but has no strong position on whether development should be slowed; he believes society can still shape how the technology is used. He expects Spotify could eventually soundtrack every moment of a listener’s life, while conceding that users can still make better playlists than the system. He favors open models for the innovation they can enable, while noting that Spotify also uses frontier models and more efficient, customizable fine-tuned models. Compute may help assess capabilities and risks, the speakers argued, but regulation based on training-compute thresholds risks treating a rapidly changing technology as static.