Philips

Philips is a Dutch health-technology company headquartered in Amsterdam, founded in Eindhoven in 1891, with shares listed in Amsterdam and New York (NYSE: PHG). Its best-known products are imaging systems, including magnetic resonance imaging (MRI), computed tomography (CT) and ultrasound, alongside the Azurion image-guided therapy platform, patient monitors, and consumer lines such as Sonicare toothbrushes. The lighting business that once carried the name is a separate company, Signify, which still licenses the Philips brand for products such as Hue. It appears here as a large health-technology AI adopter whose AI reaches customers as features inside its scanners and imaging software, not as a standalone model or service.

Its defining pattern is the embedded-AI device maker. The learned models sit inside hardware it already sells, reconstructing images or guiding a procedure, so a hospital buys a scanner or an upgrade rather than an AI subscription. In July 2025 the US Food and Drug Administration (FDA) cleared SmartSpeed Precise, deep learning MRI reconstruction that pairs one AI model for denoising with a second for sharpening. Philips says it delivers up to three times faster scanning and up to 80 percent sharper images than its earlier techniques, and that it is cleared across its current MRI portfolio, including machines already installed. In March 2026 the FDA cleared DeviceGuide, built with Edwards Lifesciences, which tracks a heart-valve repair device in real time by fusing live ultrasound and X-ray images. In April 2026 it cleared Verida, a spectral CT scanner with a deep learning reconstruction engine designed to cut image noise.

Where it needs outside models or compute, it partners. In November 2024 Philips expanded its Amazon Web Services collaboration, moving its radiology, pathology and cardiology software onto HealthSuite Imaging on AWS and planning generative AI features built on foundation models from Amazon Bedrock. In May 2025 it announced a foundation model for MRI with NVIDIA, aimed at automated scan planning and image enhancement, and in March 2026 the pair showed a research preview that predicts an image before a scan begins. In November 2025 it previewed generative AI for radiology tasks such as patient summaries, and agentic AI for prioritizing studies, both labeled work in progress.

The backdrop is a costly quality failure. After a June 2021 recall of sleep apnea and ventilator devices over degrading foam, Philips Respironics entered a consent decree with the US government in April 2024, and it will not resume selling new sleep and respiratory machines in the US until the decree's requirements are met. It is not an AI story, and the company's AI work is concentrated in imaging, monitoring and image-guided therapy, which the decree does not cover.

The honest framing: Philips's value rests on an installed base of scanners, monitors and interventional systems and the service contracts around them, and AI is a lever that makes that hardware faster, sharper and easier to run. The measured results so far are the company's own claims against its older techniques, not independent outcome data, and the NVIDIA foundation model and the generative and agentic tools are still intentions. That places Philips beside Siemens Healthineers and GE HealthCare, which also embed AI in the scanners they sell, and near Medtronic, whose AI rides inside regulated devices, rather than with the drug makers that buy or build models for discovery.