Just now! The king of global computing power reveals his hand: input is electricity, output is $BTC, and in the middle is me. Giving up the Chinese market? He says that's the stupidest thing.

Nvidia’s leader summarized the company’s essence in one sentence: input is electrons, output is tokens, and in the middle is Nvidia. This definition was given in response to external doubts about its potential “commodification.” He believes that transforming electrons into valuable tokens inherently involves significant engineering and scientific barriers, making it difficult to replicate simply.

Faced with nearly a hundred billion dollars in supply chain procurement commitments, market analysis indicates this forms part of its moat. But more critically, he continuously communicates with upstream suppliers’ CEOs, demonstrating the future scale of the AI industry to motivate them to invest in expanding production early. The transition from bottlenecks two years ago to abundant capacity today is an example. He believes any bottleneck in the supply chain will not last more than two or three years.

The real long-term constraint lies in downstream energy policies. Without sufficient electricity, no computing industry can thrive.

When asked about competition from Google TPU, Amazon Trainium, and other self-developed chips, he drew a clear line: Nvidia is engaged in “accelerated computing” that covers the entire scientific field, not just AI tensor processing. Its market scope is much larger than any dedicated chip.

He further emphasized that Nvidia’s computing platform offers the best cost-performance ratio globally, with public benchmarks like InferenceMAX standing out, yet competitors have not challenged it. Regarding Anthropic’s extensive use of TPUs, he views it as an exception rather than a trend and admits that failing to make strategic investments in it early was his mistake.

Holding massive cash flows, Nvidia chooses to invest in emerging cloud service providers like CoreWeave rather than directly building its own cloud. This stems from its company philosophy of “doing as much as necessary, as little as possible.” Even if Nvidia doesn’t develop cloud services, others will; but the CUDA ecosystem and accelerated computing platform might not have emerged without Nvidia’s persistent investment.

On GPU allocation, he explicitly denies the “highest bidder wins” bidding model, emphasizing adherence to customer purchase orders and first-come, first-served principles, aiming to be a reliable industry cornerstone.

Regarding the Chinese market, he points out that China does not lack chips and has about 50% of the world’s AI researchers. He believes abandoning the world’s second-largest market won’t lead the US to win the long-term technological race; dialogue and research exchanges might be a safer approach.

The CUDA ecosystem constitutes its core barrier, with hundreds of millions of GPU installations, programmable architecture, and rich application scenarios forming a powerful development flywheel. Coupled with its leading position in cost-performance and energy efficiency, this further drives the flywheel’s operation.

In architecture evolution, Blackwell achieves 50 times better energy efficiency than Hopper, mainly relying on algorithm and computing architecture innovations rather than mere transistor improvements. Its product roadmap maintains annual stable iterations and is the only company capable of handling AI compute orders of any scale, from tens of millions to hundreds of billions of dollars.

Follow me: for more real-time analysis and insights into the crypto market! $BTC $ETH $SOL

#GatePreIPOs首发SpaceX #Crypto market recovery #US-Iran negotiations and troop deployment game