Bessent says US will command 80% of global compute power

Treasury Secretary Scott Bessent said the US will soon command 80% of global computing power, up from 50% to 60% today, positioning AI infrastructure as the central battleground of American economic strategy against China.

"The United States is an AI superpower, and we cannot afford to lose this race," Bessent said during a July 20 appearance on the Mike Rowe Show. He described the computing contest as a strategic competition the US "must not lose," warning that ceding the lead would give an adversary "unacceptable" strategic leverage.

Bessent placed artificial intelligence alongside semiconductors and quantum computing as the three pillars of US economic power and national security. He cited historical parallels — the automobile, Google Search — to argue that technological disruption creates more opportunities than it destroys, adding that he has yet to observe net job losses attributable to AI. The Treasury Secretary's remarks echo those of David Sacks, the White House AI and cryptocurrency czar, who has publicly argued that the US must maintain its edge in both compute capacity and model capability at any cost.

The statement from the nation's top economic official provides a de facto policy endorsement for sustained AI infrastructure spending, directly supporting the capital expenditure cycles of Nvidia Corp., Advanced Micro Devices Inc., and the major US hyperscale cloud operators — Amazon Web Services, Microsoft Azure and Google Cloud. The three cloud providers collectively plan to spend more than $200 billion on infrastructure in 2026, according to industry estimates.

Bessent first outlined this policy direction during a June 3 Senate hearing, where he voiced support for a Trump-era executive order on AI cybersecurity cooperation between the federal government and the private sector. He followed with a June 23 speech spotlighting US leadership in foundational technologies, naming AI, semiconductors and quantum computing as pillars of national economic power.

The 80% figure, while striking, lacks an independent statistical basis. No publicly available third-party research firm currently publishes a "global compute share" metric, making Bessent's claim a policy aspiration rather than a verifiable data point. Investors should watch for corroboration from semiconductor industry trackers and cloud infrastructure reports in coming quarters, according to analysts cited by Podcast Alpha.

China has been racing to close the gap, with companies such as Huawei Technologies Co. developing its Ascend 910B AI chips as a domestic alternative to Nvidia's products, despite US export controls restricting access to advanced semiconductors. The Biden-era chip export curbs, maintained by the current administration, have slowed but not halted China's AI compute expansion, according to semiconductor supply chain analysts. Chinese tech giants Baidu Inc., Alibaba Group Holding Ltd. and Tencent Holdings Ltd. have stockpiled Nvidia chips ahead of potential further restrictions, highlighting the strategic importance of compute access.

For market participants, the policy direction is unambiguous: the US government intends to support domestic AI infrastructure buildouts for the foreseeable future. Nvidia, whose data center revenue reached $47.5 billion in its most recent fiscal year, remains the primary beneficiary of hyperscale capital spending. The company's H100 GPUs, delivering 990 TFLOPS of FP16 performance, and its forthcoming Blackwell architecture are the backbone of most large-scale AI training clusters deployed by US cloud providers. AMD's MI300X accelerators, with 192 GB of HBM3 memory, offer an alternative for inference workloads at a lower price point, though they have yet to meaningfully erode Nvidia's estimated 80% share of the AI accelerator market.

However, the compute dominance thesis carries risks. If the 80% share projection proves overstated, the current valuation premium on AI-exposed equities could face a correction. Nvidia trades at approximately 35 times forward earnings, a multiple that embeds aggressive growth assumptions. Independent verification from firms such as Gartner Inc. or IDC will be critical to distinguishing policy ambition from measurable reality.

The next data points to watch include quarterly capital expenditure disclosures from the three major US cloud providers and capacity utilization reports from Taiwan Semiconductor Manufacturing Co., which manufactures the advanced chips powering AI training clusters. Semiconductor Industry Association sales data will also offer tangible measures of compute expansion. Bessent's claim will either gain credibility as these numbers accumulate or fade as a rhetorical flourish.

This article is for informational purposes only and does not constitute investment advice.