NVIDIA’s D.C. Blitz: Lilly’s AI Factory, Nokia 6G, Uber AVs & DOE’s “Solstice”

October 29, 2025

In partnership with

Introducing the first AI-native CRM

Connect your email, and you’ll instantly get a CRM with enriched customer insights and a platform that grows with your business.

With AI at the core, Attio lets you:

  • Prospect and route leads with research agents

  • Get real-time insights during customer calls

  • Build powerful automations for your complex workflows

Join industry leaders like Granola, Taskrabbit, Flatfile and more.

  • Lilly × NVIDIA: Pharma’s largest AI factory, >1,000 Blackwell-Ultra GPUs, to accelerate discovery-to-delivery.

  • Palantir × NVIDIA: CUDA-X + Nemotron models embedded in Palantir AIP for decision intelligence at the edge.

  • Nokia × NVIDIA: Strategic AI-RAN/6G pact + $1B NVIDIA equity investment at $6.01/share.

  • Uber × NVIDIA (+ Stellantis): Uber to build one of the world’s largest AV networks on NVIDIA DRIVE; initial 5,000 L4 vehicles from Stellantis.

  • DOE × NVIDIA (+ Oracle): DOE’s largest AI supercomputer (“Solstice”) using 100,000 Blackwell GPUs; DOE has published the partnership.

  • Quantum × NVIDIA: NVQLink introduced to directly connect QPUs with GPU supercomputers for microsecond-class hybrid workflows.

1) Eli Lilly x NVIDIA — Building Pharma’s Most Powerful AI Factory

What’s official:

  • Lilly is deploying an NVIDIA Blackwell-based DGX SuperPOD with >1,000 Blackwell Ultra GPUs to power an “AI factory” for discovery, clinical, manufacturing and enterprise agents.

  • Lilly’s own release frames it as “the industry’s most powerful AI supercomputer” for medicine discovery and delivery.

Why it matters: Brings foundation-model scale (molecules, modalities, imaging) inside a regulated, proprietary environment—shortening hypothesis-to-trial loops and enabling federated collaborations (via Lilly TuneLab) without exposing IP.

2) Palantir x NVIDIA — Operational AI From Data to Decisions

What’s official:

  • Palantir is integrating NVIDIA accelerated computing (CUDA-X) and open-source Nemotron models into the Palantir AI Platform’s Ontology core.

  • Palantir’s blog confirms the partnership unveiled in D.C. and how NVIDIA models will be delivered through AIP and the edge.

Why it matters: It marries Palantir’s structured, mission-critical Ontology with NVIDIA’s training/inference stack for supply chains, logistics, and real-time ops—with early adopters like Lowe’s highlighted.

3) Nokia x NVIDIA — AI-RAN on the Path to 6G (+ $1B Equity)

What’s official:

  • Strategic partnership to add NVIDIA-powered AI-RAN products to Nokia’s RAN portfolio for AI-native 5G-Advanced → 6G.

  • NVIDIA will invest $1B in Nokia via 166,389,351 new shares at $6.01 each (subject to customary conditions).

  • NVIDIA’s own newsroom mirrors the 6G/AI-RAN framing and the investment headline.

Why it matters: Positions AI as a first-class workload inside the RAN—enabling programmable, inference-capable basebands and a glidepath to 6G with distributed edge AI.

4) Uber x NVIDIA (with Stellantis) — Scaling a Global L4 Robotaxi Network

What’s official:

  • Uber IR: building one of the world’s largest AV networks on NVIDIA DRIVE AGX Hyperion 10; Stellantis to supply 5,000 L4 vehicles.

  • NVIDIA newsroom: partnership to support Uber’s global expansion; DRIVE AV + Hyperion 10 is the reference stack.

Note on “100,000 fleet”: The 100k figure is widely reported by media, but not in the official releases above; treat it as reported guidance, not committed inventory. (Press for official)

5) U.S. Department of Energy x NVIDIA (+ Oracle) — “Solstice” & Friends

What’s official:

  • DOE newsroom confirms a public-private partnership with NVIDIA and Oracle to build the DOE’s largest AI supercomputer.

  • NVIDIA newsroom: “Solstice” to feature ~100,000 Blackwell GPUs; also “Equinox” (~10,000).

About “seven supercomputers”: “Seven” has been stated in media coverage, but DOE has formally published the flagship partnership; additional systems have not yet been detailed on DOE’s site.

What’s official:

  • NVIDIA introduced NVQLink, a high-speed interconnect and open platform to link quantum processors (QPUs) with GPU supercomputers (Brookhaven, Berkeley Lab, Oak Ridge, PNNL, etc.).

Why it matters: Hybrid quantum-classical workflows (error mitigation, VQE-style solvers, surface-code research) need µs-latency, synchronized data paths between QPUs↔GPUs—NVQLink is the missing bus. (Industry press echoes the positioning.)

Status Checks on Items You Listed

  • Samsung partnership (new): No new Samsung×NVIDIA press release in the last 48 hours. Samsung’s official most recent NVIDIA-adjacent network item is AI-RAN work (MWC’25 timeframe). I’ve linked Samsung’s newsroom entry for that context.

  • Hyundai partnership (new): No new Hyundai×NVIDIA release in the last 48 hours. Hyundai’s official newsroom shows a Jan 2025 NVIDIA AI collaboration announcement; NVIDIA’s partner page reflects ongoing SDV work.

  • “$500B in expected revenue through 2026”: Official channels don’t state this as revenue. Reuters attributes $500B to chip bookings/backlog, not top-line revenue. Treat as orders, not P&L guidance.

Strategic Takeaways

  1. AI Factories Are Crossing Sectors: From Lilly’s drug discovery to DOE science and Nokia’s AI-RAN, the same NVIDIA compute stack (Blackwell + CUDA-X + NIM) is becoming a common substrate across healthcare, telecom, and government.

  2. Edge & Network-Native AI Rise: Palantir AIP at the edge and AI-RAN in the network make inference a built-in capability of physical infrastructure, not an add-on.

  3. Hybrid Quantum-Classical Is Getting Real: NVQLink formalizes the QPU↔GPU data path required for practical, near-term quantum workflows.

“What to Watch Next”

  • Lilly: commissioning timeline + model zoo details on TuneLab; validation studies that show cycle-time compression from hit-to-lead.

  • Palantir: sector-specific AIP blueprints with Nemotron; performance benchmarks vs. prior AIP stacks.

  • Nokia: initial AI-RAN field trials and CSP rollouts; 6G research milestones tapping NVIDIA’s stack.

  • Uber: pilot deployments, safety KPIs, cost/ride vs. human drivers; expansion beyond the first 5,000 Stellantis units.

  • DOE: “Solstice” acceptance tests, workload mix (climate, fusion, nuclear stewardship), energy envelope.

  • NVQLink: first lab demos with named QPU vendors (latency, fidelity, throughput).