Jensen Huang’s Big Week: China Trip, AI Demand, New Deals

NVIDIA CEO Jensen Huang had one of his most consequential weeks in recent memory, moving from a Dell Technologies World keynote declaring AI demand “utterly parabolic” to joining President Trump’s China summit at the White House’s direct request — while also announcing a partnership with British AI startup Ineffable Intelligence to pursue what he called “the next frontier” of AI development.

Jensen Huang at Dell Technologies World

Huang took the stage at Dell Technologies World on Monday, May 12, alongside Dell chairman and CEO Michael Dell, to make the case that AI has crossed from experimentation into production.

“We’ve now arrived at the era of useful AI, which is the reason why demand is going parabolic, utterly parabolic,” Huang said, according to NVIDIA’s AI Blog. “What took months now takes weeks. What took weeks now takes days.”

Michael Dell framed the scale of the opportunity: worldwide AI infrastructure spending could reach $3–4 trillion by 2030, with token consumption projected to grow 3,400% in the same window. Dell said more than 5,000 enterprises — including Eli Lilly, Samsung, and Honeywell — are already running AI workloads on Dell AI Factories with NVIDIA hardware.

The keynote showcased NVIDIA’s full hardware portfolio, from a deskside Dell Pro Max with GB10 workstation to a Dell PowerRack with NVIDIA Vera Rubin NVL72. NVIDIA claimed the Vera Rubin NVL72 delivers agentic AI inference at one-tenth the cost per token compared to prior-generation systems, while the Vera CPU runs agent sandboxes 50% faster than traditional CPUs and accelerates enterprise data queries by up to 3x.

The China Trip: Trump Called Huang Directly

The same week, Huang’s name surfaced in a very different context: U.S.-China trade diplomacy.

NVIDIA confirmed to CNBC that Huang would join President Trump’s delegation to China, where Trump was scheduled to meet with Chinese President Xi Jinping on Thursday and Friday. Initial reports had suggested Huang would not be part of the trip, but after seeing media coverage of his absence, Trump called Huang personally, a source told CNBC.

Speaking to reporters on Thursday, May 14, Huang confirmed the account. “President Trump asked me to come,” he said, according to CNBC. He described the summit as “one of the most important summits in human history.”

The trip carries direct commercial implications for NVIDIA. U.S. export controls have blocked the company from selling its most advanced AI chips — including H100 and successor architectures — to Chinese customers without a license, a restriction that has cost NVIDIA billions in potential revenue and pushed Chinese cloud providers toward domestic alternatives. Whether the summit produces any policy movement on chip export rules remains to be seen, and Huang is expected to face pointed questions on the topic during NVIDIA’s upcoming earnings call, according to CNBC.

NVIDIA Backs British Startup Ineffable Intelligence

Separate from the China diplomacy, NVIDIA announced on Wednesday, May 13, a partnership with Ineffable Intelligence, a British AI startup focused on reinforcement learning — a training approach in which AI systems learn from experience rather than from static human-generated data.

“The next frontier of AI is superlearners — systems that learn continuously from experience,” Huang said, according to CNBC’s reporting on the announcement. The companies said they will collaborate to develop new AI systems, though specific product timelines and financial terms were not disclosed.

The deal reflects a broader strategic bet by NVIDIA: that the next phase of AI capability will come not from larger pretraining runs on human text, but from agents and models that improve through interaction with environments — a paradigm where NVIDIA’s simulation and inference infrastructure is already positioned as a core enabler.

Why Reinforcement Learning, Why Now

Reinforcement learning has powered high-profile results in game-playing AI and robotics for years, but applying it reliably to general-purpose language and reasoning tasks has proven difficult. Ineffable Intelligence is among a cohort of startups — alongside companies like Sakana AI and others — betting that recent advances in compute efficiency and reward modeling have made the approach viable at scale.

NVIDIA’s involvement gives Ineffable access to GPU clusters and software tooling, while NVIDIA gains an early stake in a research direction that could drive demand for inference compute well beyond current LLM workloads.

Vera Rubin and the Hardware Roadmap

The Dell keynote was also an opportunity for NVIDIA to put its next-generation hardware in front of enterprise buyers. The Vera Rubin NVL72 — a 72-GPU rack-scale system built around the Rubin GPU architecture and the Vera CPU — is positioned as NVIDIA’s answer to the economics of agentic AI, where inference costs per token matter more than raw training throughput.

Key claimed specifications from the NVIDIA AI Blog:

• 1/10th the cost per token for agentic AI inference vs. prior generation
• 50% faster agent sandbox execution on Vera CPU vs. traditional CPUs
• 3x faster enterprise data queries on Vera CPU
• Available in configurations ranging from deskside workstations (GB10) to full rack deployments (NVL72)

NVIDIA has not publicly disclosed pricing or general availability dates for the Vera Rubin NVL72 beyond what was shown at the Dell event.

What This Means

The past week illustrated how thoroughly Jensen Huang has positioned NVIDIA at the center of three distinct but converging forces: enterprise AI adoption, geopolitical chip policy, and next-generation AI research.

The Dell partnership numbers — 5,000 enterprises in production, $3–4 trillion in projected infrastructure spend — are the kind of figures that shift analyst models, not just press releases. If even a fraction of that projection materializes, NVIDIA’s data center revenue, already its largest segment, has substantial room to grow.

The China trip is the highest-risk variable. Export controls have been the single biggest constraint on NVIDIA’s addressable market outside the U.S., and any softening of restrictions — or any hardening — will move the stock and reshape competitive dynamics with Huawei’s Ascend chips, which have gained ground in China precisely because H100-class hardware is unavailable. Huang’s presence at the summit signals that the chip trade is now explicitly on the diplomatic table, not just a regulatory side matter.

The Ineffable Intelligence deal is smaller in immediate commercial terms but directionally significant. NVIDIA’s long-term moat depends on being the default compute substrate for whatever AI paradigm comes next. Backing a reinforcement learning startup — and framing it publicly as “the next frontier” — is a signal about where Huang thinks the research frontier is heading, and where he wants NVIDIA’s hardware to be indispensable.

FAQ

Why did Jensen Huang join Trump’s China trip?

According to CNBC, President Trump called Huang directly after seeing media coverage that the NVIDIA CEO was not included in the original delegation. Huang confirmed to reporters that Trump “asked me to come” and described the summit as critically important.

What is the NVIDIA Vera Rubin NVL72?

The Vera Rubin NVL72 is a rack-scale AI system combining NVIDIA’s Rubin GPU architecture with the Vera CPU. According to NVIDIA’s AI Blog, it is designed for agentic AI inference workloads and is claimed to deliver inference at one-tenth the cost per token compared to prior-generation NVIDIA systems.

What is Ineffable Intelligence and why is NVIDIA partnering with it?

Ineffable Intelligence is a British AI startup focused on reinforcement learning — training AI systems through experience rather than static human data. According to CNBC, NVIDIA announced a partnership with the company on May 13, 2026, with Huang framing reinforcement learning-based “superlearners” as the next major frontier in AI development.