Descriptions:
Gil Luria, head of technology research at DA Davidson, and Joe Kaiser, CEO of Switchyard Partners, join Bloomberg Technology to dissect Nvidia’s latest earnings and the broader AI infrastructure landscape. Luria leads with Jensen Huang’s unusually direct competitive commentary during the earnings call: Huang explicitly called out AMD and Broadcom as having zero presence in a key segment of Nvidia’s business, dismissed Intel and AMD CPU businesses by announcing Nvidia would grow a standalone CPU business from zero to $20 billion in a single year, and characterized the hyperspeed inference market — where Cerebras competes — as relatively tiny.
The analysts surface several striking data points: Nvidia’s networking business alone runs at a $15 billion annual revenue rate, roughly on par with Qualcomm’s entire company; total projected annual sales are expected to hit $370 billion; and critically, Nvidia consumes nearly two-thirds of all chip packaging capacity coming out of TSMC, creating a supply chain moat rooted in clean-room packaging rather than silicon design. At 14.5x projected sales, the panel argues Nvidia looks cheap relative to SpaceX’s proposed 80x sales multiple.
The conversation turns to SpaceX’s IPO filing and its AI angle through xAI. Luria explains that SpaceX’s massive GPU clusters have excess capacity — since Grok hasn’t yet generated enough inference demand to fill them — which SpaceX is renting to Anthropic for a reported $15 billion annually. SpaceX holds a 90-day termination clause to reclaim that capacity once Grok’s usage grows, making the arrangement a bridge finance mechanism for Elon Musk’s AI buildout.
📺 Source: Bloomberg Technology · Published May 22, 2026
🏷️ Format: News Analysis
