Nobody Uses AI Better Than Meta

Huang’s core thesis is that a direct line connects Nvidia’s GPUs to Meta’s top line. He and Meta both point to the company’s AI-driven ad ranking system, which they claim has delivered 4x more revenue impact than simply increasing ad load . The resulting incremental return on investment (ROI) on that AI spending is reported to be above 20% . For Huang, Meta’s 12% increase in ad pricing is the visible outcome of this AI flywheel, where better targeting and prediction directly translate to higher pricing power .

Huang’s Three-Part Thesis for Meta’s Capex

At Computex, Huang’s defense of Meta’s spending was built on three pillars, which serve as his answer to skeptical investors:

1. Unmatched Monetization of Infrastructure: Huang argued Meta demonstrates the highest monetization of AI infrastructure of any non-cloud company, with a direct, measurable link between AI spend, ad pricing, and impression volume .
2. Proven ROI Above 20%: The ROI on incremental AI investment exceeds 20%, meaning each dollar of capex is projected to generate more than a dollar in advertising profit over its lifecycle, making it a high-return investment rather than a speculative cost .
3. A Deepening Competitive Moat: As Meta’s AI models become more sophisticated at ad ranking and personalization, they create an ever-widening gap in ad effectiveness that rivals will struggle to close, cementing a long-term competitive advantage .

Skeptics of this view maintain that a $125B+ annual capex relative to a ~$225B revenue base is historically unprecedented, and that the useful life and ROI of the vast numbers of new chips are unproven at this scale.

The Broader Computex 2026 Context

Huang’s Meta comment was not an isolated soundbite but part of a landmark keynote where he declared that the AI industry has entered a "full-fledged commercialization phase" . The presentation moved markets, sent individual stocks soaring, and laid out Nvidia’s expanding footprint across the entire computing landscape.

• Nvidia Enters the PC Market with RTX Spark: In a historic announcement, Huang launched the RTX Spark, an Arm-based superchip co-developed with MediaTek and Microsoft. The chip puts a Blackwell RTX GPU and full CUDA stack into Windows on Arm laptops and compact desktops, marking Nvidia’s formal entry into the PC processor market and a direct challenge to Intel, AMD, and Apple . The first RTX Spark devices from major OEMs are set to ship in Fall 2026 .
• Marvell’s Record-Breaking Surge: In a surprise appearance alongside Marvell CEO Matt Murphy, Huang predicted the chipmaker could become the “next trillion-dollar company.” The endorsement sent Marvell (MRVL) shares soaring 32.5% to a record close of $290.79 on June 2, adding roughly $60 billion in market cap in a single day and marking its largest single-day gain ever . Nvidia also committed $2 billion to deepen its partnership with Marvell on custom silicon and optical interconnects .
• Markets Hit Record Highs: The Computex AI-fueled rally pushed all three major U.S. indexes to record closing highs on June 2, with the S&P 500, Nasdaq, and Dow all hitting new records as optimistic AI sentiment swept the market .
• “Agentic AI Has Arrived”: The central theme of Huang’s presentation was that useful, revenue-generating agentic AI is here now. He framed every AI token as a revenue unit, signaling a shift from AI training to inference as a business .
• Dismissing AI Job Fears: Huang addressed head-on the growing concern that AI is eliminating software jobs, calling the idea “complete nonsense” and arguing that agentic AI will in fact create more demand for coders and software tools .

The investor skepticism around Meta's unprecedented capex remains, making it the central tension point in the company's financial story. At Computex, Jensen Huang placed his full weight on one side of that debate, making Meta the headline example in his broader argument that AI has definitively arrived as an engine for real profit.