The present contest over artificial intelligence between the United States and China is no mere commercial rivalry; it is a struggle over strategic infrastructure, prosecuted through the instrument of export controls. The foundational question is not what the technology can do, but what the government should permit in the interest of national security and long-term technological supremacy. For a company such as Alphabet Inc., whose research and cloud services sit astride the entire AI value chain, the evolving regulatory architecture demands careful navigation, for the ground is shifting from tangible hardware to intangible models and cloud-based access with unprecedented speed.
The Expanding Architecture of US Export Controls
The United States has made export controls the centerpiece of its technology competition with China. Beginning in 2022, the unlicensed export of advanced AI chips to China was forbidden, and the relevant licenses have, in practice, been consistently refused 18. The controls have tightened steadily: further restrictions were layered on in 2023 and 2024 12,49, and by mid-2026 the Commerce Department was taking measures to close the loopholes that had allowed Chinese entities to acquire restricted semiconductors through offshore subsidiaries in Singapore, Malaysia, and elsewhere 7,11,17. The explicit objective of these efforts is to deny China the compute capacity necessary to train frontier AI models 6,57.
Yet the intended effects have been far from uniform. Multiple accounts indicate that the controls have, paradoxically, accelerated Chinese domestic innovation, compelling indigenous firms to build independent technology stacks and optimize models for less powerful hardware 2,10,50. The emergence of competitive Chinese AI models, such as DeepSeek, directly challenged the assumption that export controls could secure a durable technological lead 29. Chinese AI laboratories continue to develop advanced capabilities 15,20, and demand for restricted Nvidia hardware remains robust on black and gray markets 26,42,43. The burden of proof, it seems, falls on those who assert that controls are achieving their strategic purpose.
China's Countermeasures and the Unintended Acceleration of Indigenous Innovation
Beijing has not been a passive observer. In response, it has drafted restrictions covering sixty-three technology sectors, including AI and quantum computing 3, prohibited state-funded data centers from using foreign chips 16, and employed trade secret laws to limit the outflow of AI technology 9. Measures have also been taken to prevent the emigration of skilled AI researchers 16. These reciprocal actions, in concert with US controls, are forging a bifurcated global AI ecosystem in which compute capacity is overwhelmingly concentrated in the United States and China 1,22,30,33. Allies in Europe, and elsewhere, find themselves vulnerable to supply shocks and over-reliance on either bloc 24,46.
The competitive narrative has been weaponized in domestic policy: the “AI race” with China is invoked to justify the relaxation of regulatory oversight for US companies 23,64 and to mobilize massive investments in energy and infrastructure 48. At the same time, internal US policy debates reveal a tension between maintaining technological leadership and imposing stringent security controls 47,61.
The Widening Scope: From Chips to Models and the Cloud
The scope of export controls has expanded dramatically beyond graphics processing units and chipmaking equipment 21,29,37,41. By mid-2026, the US government was attempting to restrict central processing units with AI capabilities—an effort industry leaders such as Arm Holdings’ CEO Rene Haas warned would be nearly impossible to enforce given the ubiquity of CPUs 5,14,19. Regulators have also taken aim at the “cloud loophole,” advancing legislation such as the Cloud Security Act and the Chip Security Act to prevent Chinese entities from accessing advanced compute via cloud services 40,45,59.
Most strikingly, the United States has for the first time applied export controls directly to AI models, classifying models like Anthropic’s Fable 5 and Mythos as restricted items under the Export Administration Regulations 28,36,51,65. Under this new interpretation, accessing a controlled model, even via an API query, is treated as an export 44,62. The Commerce Department’s directive focuses on AI capabilities relevant to cybersecurity, bioengineering, and chemical synthesis 38, and it applies to foreign nationals both abroad and within the United States 38. The speed of implementation—bypassing ordinary legislative processes—signals an urgency that concentrates unilateral executive power 52. It is a settled principle that such power should be wielded with caution, but also with dispatch when the national interest requires.
Strategic Implications for Alphabet Inc.
Alphabet, through Google DeepMind and Google Cloud, occupies a position at the nexus of these dynamics. The company is a primary developer of frontier AI models that could themselves become targets of export controls if they meet capability thresholds in cybersecurity, bioengineering, or other national-security domains 38. The sudden restriction of Anthropic’s models sets a precedent under which any US-based AI platform—including Google’s Gemini—could be ordered to suspend access for foreign nationals overnight 32,36,63. This introduces not merely compliance risk but strategic uncertainty: enterprise and government customers abroad may increasingly question the reliability of US-controlled AI services and turn to domestic alternatives 24,27.
At the same time, export controls on advanced chips do not directly constrain Alphabet’s domestic training infrastructure. Indeed, the US government’s prioritization of chip supply for American firms and the incentives of the CHIPS Act help ensure Alphabet’s access to cutting-edge hardware 13,31,53. The concentration of AI compute in the United States reinforces Alphabet’s competitive advantage vis-à-vis non-Chinese foreign rivals 1,33. However, if controls accelerate Chinese innovation and independent chip production, Alphabet may face a more potent Chinese competitor in global markets 10,18,54.
The persistent demand for restricted chips and the emergence of sophisticated smuggling networks indicate that no regulatory framework is watertight 40,56,59,60. Alphabet may find itself entangled in controversy if its technology ends up in prohibited hands through third parties. Moreover, the expansion of controls to model access via APIs blurs the line between cybersecurity and trade policy; Alphabet will need to invest in robust “know-your-customer” and geo-fencing capabilities to avoid inadvertent deemed exports 32,62.
Efforts to enlist allies through legislation such as the MATCH Act, the STRIDE Act, and the A.I. OVERWATCH Act signal a desire to create a multilateral regime and to remove executive sole authority over controls 8,34. A multilateral framework could level the playing field by preventing Chinese firms from bypassing US controls through third countries, but it also risks entangling Google’s international operations in a web of conflicting national regulations. The “arms race” framing is not mere hyperbole: both nations treat AI as critical strategic infrastructure 9, and Alphabet’s own reliance on Taiwan for advanced chips and its investments in supply chain diversification underscore the fragility of the current architecture 4,35. The tail risk that export controls could disrupt Western AI infrastructure cannot be dismissed 58.
Key Takeaways and Recommendations
Careful review warrants the following conclusions:
- The United States is escalating export controls from tangible chips to intangible software and cloud access with extraordinary speed, creating a regulatory environment in which Alphabet’s frontier models could be subjected to sudden, unilateral restriction orders 28,38,39,51.
- The unintended acceleration of Chinese domestic AI innovation and the persistent circumvention of controls suggest that the strategic moat provided by export restrictions may prove temporary, potentially eroding Alphabet’s competitive lead in international markets over the medium term 25,40,55,59.
- Alphabet must prepare for a fracturing global AI landscape in which some customers and governments will increasingly prefer non-US technology stacks to avoid geopolitical dependency, calling into question the long-term growth of its cloud and AI platform revenues outside allied nations 24.
- In the near term, robust enforcement and domestic chip prioritization benefit Alphabet’s training infrastructure and protect its US market share, but the company should proactively engage in sovereign-cloud partnerships and open-weight model strategies to retain relevance in a multipolar AI world.
Nothing in this approach precludes Alphabet from advocating for a balanced policy that protects national security without unduly disadvantaging American enterprises. We must proceed with caution, but also with dispatch, for the technological order of the coming decade is being forged in the crucible of export controls.