On October 28, 2025, Jensen Huang stood at a podium in Washington, DC, at Nvidia's GTC conference and announced a partnership with Palantir. He chose his words with care. "We work with Palantir," he said, "to accelerate everything Palantir does so that we can do data processing at a much much larger scale and more speed — whether it's structured data of the past, human-recorded data, unstructured data — and process that data for our government, for national security, and for enterprises around the world, process that data at speed of light and find insight from it." Not process data. Not analyse information. Accelerate everything Palantir does.
The supply chain runs like this. In 2012, Matt Danzeisen co-founded Crescendo Equity Partners in Seoul. In 2013, Crescendo made its first investment in Hanmi Semiconductor. In 2016, Thiel and Crescendo invested again in convertible bonds worth 75 billion won in Hanmi. In 2021, Crescendo and the Hanmi chairman jointly invested in HPSP — a company that describes itself as the world's only manufacturer of high-pressure hydrogen annealing equipment, which is essential for advanced semiconductor processing. HPSP was listed on KOSDAQ in 2022. By 2024 its sales had grown to 181.4 billion won, a return of 639% on the original investment. Korean financial press called it "Korea's ASML." Crescendo held a 39.42% stake.
On March 12, 2026, Palantir and Nvidia jointly announced the Palantir AI OS Reference Architecture (AIOS-RA): a complete, production-ready AI infrastructure running Palantir's full software suite — AIP, Foundry, Apollo, Rubix, and AIP Hub — on Nvidia Blackwell Ultra systems. Eight Blackwell Ultra GPUs per unit. The product is explicitly marketed to "clients with data sovereignty requirements, existing GPU infrastructure, latency-sensitive workflows, and wide geographic distribution." Those clients include the US government, NATO-aligned defence agencies, and foreign governments. [SOURCE: NVIDIA Newsroom · Palantir blog · BusinessWire March 12 2026]
