The Whole Society Is Anxious About AI Risks, Why Is Jensen Huang Betting on Anthropic’s Trillion-Dollar Revenue Goal?

TradingKey - On March 19, NVIDIA ( NVDA) CEO Jensen Huang shared two major viewpoints during a podcast discussion. First, he is bullish on the commercial potential of AI startup Anthropic, stating that reaching the $1 trillion revenue threshold by 2030 is by no means far-fetched. Second, he called on tech industry leaders to be cautious about the framing of AI risk discussions, warning against letting reasonable risk warnings devolve into baseless panic.

When discussing AI risks, Huang stated bluntly that the industry currently has a tendency toward sensationalism. He emphasized that while highlighting technical risks is necessary, sliding into groundless alarmist hype will instead cause serious harm to the industry's development.

"AI is not a conscious organism, nor is it some alien species; it is essentially a piece of computer software," Huang pointed out, noting that extreme catastrophic claims lacking supporting evidence pose a far greater danger than the risks inherent in the technology itself.

In his view, the greatest AI threat facing the United States is not technology spinning out of control, but rather cognitive bias at the societal level. "The real fatal risk is when society as a whole proactively slows down AI adoption due to fear or paranoia," Huang said. He cited the public discourse during the early days of the automobile as an example, noting that over-exaggerating the "risk of hitting people" only hinders technological proliferation, and AI development similarly requires a rational perspective.

Meanwhile, Huang remains optimistic about Anthropic's business prospects, projecting that the company's revenue could surpass $1 trillion by 2030 and suggesting that its CEO Dario Amodei's own forecasts are too conservative.

As the developer of the Claude large language model, Anthropic drew the ire of the Trump administration for insisting on two strict contractual restrictions: prohibiting AI products from being used to monitor U.S. citizens domestically and banning the development of fully autonomous lethal weapon systems.

The U.S. Department of Defense has currently designated it a "national security supply chain risk" and is pushing to divest it from government projects, while Anthropic is fighting back vigorously through legal channels.

Even amidst such intense political and legal confrontations, Huang still expressed the highest level of optimism regarding Anthropic's commercial potential. During the discussion, he stated clearly that it is not a pipe dream for Anthropic's annual revenue to exceed $1 trillion by 2030, even suggesting that CEO Dario Amodei's self-projections were too conservative.

Based on the data, this goal is indeed extremely challenging.

Anthropic's current annualized revenue is approximately $19 billion, a doubling from last year's $9 billion; however, crossing the $1 trillion threshold within five years would require maintaining a multi-fold annual growth rate.

OpenAI was also previously expected to reach trillion-dollar revenue levels, but its current annualized revenue is only about $25 billion, with costs exceeding $35 billion. Commercial monetization in the AI space still faces massive uncertainty.

NVIDIA's Ambitions

Interestingly, Huang's optimistic assessment of the AI industry is also reflected in NVIDIA's own strategic positioning. At the recent GTC conference, NVIDIA CEO Jensen Huang delivered a signal to the market that exceeded expectations: NVIDIA's Blackwell and Rubin AI chip architectures are projected to generate at least $1 trillion in cumulative order demand by the end of 2027. This forecast is a direct doubling of the $500 billion target announced during the same period last year.

During his keynote speech, Huang emphasized: "The demand curve for AI computing is just beginning to steepen; we are standing at the starting point of a once-in-a-decade technological transformation." NVIDIA's trillion-dollar forecast serves as a clear signal of this impending change.

On March 20, NVIDIA and Amazon ( AMZN) Web Services (AWS) reached a large-scale chip procurement agreement that happens to coincide with the timeline for the $1 trillion market opportunity for the Blackwell and Rubin chip series proposed by Huang.

Under the agreement, AWS will purchase 1 million NVIDIA GPUs. Ian Buck, NVIDIA's Vice President of Hyperscale and High-Performance Computing, revealed on Thursday that shipments will officially begin this year and continue through 2027.

Notably, the scope of this partnership extends far beyond 1 million GPUs, also covering Spectrum networking chips and the newly launched Groq chips, among other products. The Groq chip is a new product launched after NVIDIA's $17 billion acquisition of an AI chip startup late last year. AWS plans to use it in combination with six other NVIDIA chips to optimize AI inference performance.

"Inference is hard—outrageously hard," Buck emphasized. "Being the best at inference isn't something you can solve with just one chip. We are actually using all seven chips."

Another major breakthrough of this partnership is that NVIDIA's networking products will enter AWS data centers for the first time. For a long time, AWS has used its own custom-developed networking equipment; this collaboration signifies that NVIDIA has made key progress in the high-end cloud data center infrastructure sector.