Written by: Conflux

On May 31, 2026, the U.S. Department of Commerce issued new export control guidance, officially closing the channel for Chinese companies to purchase NVIDIA advanced chips through overseas subsidiaries in Malaysia and other locations.

In the same month, the Kenyan president halted a $1 billion geothermal data center project involving Microsoft—because it would consume a third of the country's electricity once completed. President Ruto's words were: "It's like shutting down half the country."

Meanwhile, Huawei announced last week that its Ascend 950PR chip has entered mass production, and it expects its AI chip revenue to reach $12 billion for the whole year.

Three things, three continents, three completely different news items. But they all point to the same emerging reality: the competition for computing power is no longer just a matter within the tech industry.

A new era of oligopolies is emerging.

Over the past two years, there has been a reality in the AI ​​industry that is often overlooked: although there seems to be a diverse range of developments on the surface, underlying resources are becoming increasingly concentrated.

The current AI industry chain can be roughly divided into four layers: GPU chips, cloud computing platforms, basic models, and application ecosystems. At each layer, control is concentrating in the hands of a few players: in the GPU field, NVIDIA is almost the only option; in the cloud computing field, AWS, Microsoft Azure, and Google Cloud dominate; and at the model layer, OpenAI and Anthropic have captured the vast majority of the high-end model market.

In other words, the same group of companies is simultaneously controlling chips, cloud platforms, models, and distribution channels. Eric Posner, a law professor at the University of Chicago, calls this phenomenon "AI Octopus," meaning that these companies' tentacles cover the entire AI industry chain.

This differs from platform monopolies in the internet age—internet platforms control traffic, while AI platforms control intelligence itself. This "oligopoly" brings profound systemic risks:

• Centralized control and pricing hegemony: A few companies control AI pricing, API access, and content moderation standards. Developers and businesses face serious "platform lock-in" risks, as giants can change rules or cut off access at any time.
• Infrastructure vulnerability: Highly centralized computing power can easily lead to single-point failures that have a ripple effect (such as large-scale cloud service outages) and put unbearable pressure on the power grid and energy of a single region.
• Geopolitics and Computing Hegemony: Computing power is transforming from a neutral infrastructure into a strategic asset. Due to export controls, countries without independent computing capabilities (especially those in the Global South) risk marginalization and a widening technology gap in this technological wave.

In the future, more and more companies will rely on AI for development, operations, customer service, marketing, and even decision-making. Once intelligence becomes a production tool, its importance in terms of control will far surpass that of search engines and social media.

The ever-deepening "AI Iron Curtain"

Over the past two years, the US has become increasingly fragmented in its handling of chip export controls. During Biden's presidency, an "AI proliferation rule" was introduced, dividing global cooperation into three levels; Trump withdrew this rule after taking office, shifting to case-by-case approvals and temporary bans. Faced with this iron curtain, different countries have reacted very differently.

Saudi Arabia has designated 2026 as the "Year of Artificial Intelligence": through its sovereign wealth fund HUMAIN, Saudi Arabia invested $3 billion in Musk's xAI, on the condition that an AI data center of more than 500 megawatts be built in Saudi Arabia; the UAE is building a 5-gigawatt AI park in Abu Dhabi—claimed to be the largest outside the United States—with the first phase going online this year; in May, the UAE received its first batch of the latest NVIDIA chips exported from the United States.

The logic of the Gulf countries is quite straightforward: in the previous era they relied on selling oil, and in this era they rely on buying computing power.

The EU's anxiety stems from another direction: official data shows that over 80% of Europe's digital services run on non-EU infrastructure. The ongoing Cloud Computing and Artificial Intelligence Development Act (CADA) aims to triple Europe's computing power by 2030. In April of this year, France's Mistral directly released a strategic document titled "European AI: A Playbook to Own It," which translates to "European AI: Take it back."

The most difficult situation is faced by economies that barely qualify to compete: Kenya's $1 billion data center project was halted; Malaysia allocated approximately $490 million to build its sovereign AI cloud; India is subsidizing researchers' GPU usage fees; Indonesia is preparing its own large-scale model—these investments are already substantial within the scope of their respective economies.

However, this year alone, the combined AI capital expenditures of just four companies—Microsoft, Google, Amazon, and Meta—reach approximately $750 billion. This disparity in scale is itself part of the problem.

The race for computing power is increasingly pointing to a more fundamental variable: electricity. A single AI inference task can consume up to 1000 times the power of a traditional web search. To cope with the projected global data center energy consumption of 1050 terawatt-hours by 2026, tech companies have even begun directly purchasing nuclear power plants.

Is there a possibility of "not taking sides"?

It is against this backdrop that decentralized AI (DeAI) has begun to attract attention. It attempts to answer the question: besides entrusting the future to a few tech giants or a few countries, is there a third possibility?

If the internet can connect to the global network through open protocols, can AI also connect to global computing power through open networks? Can idle GPUs, independent developers, research institutions, and enterprise data centers around the world form an open AI infrastructure network?

The core idea of ​​DeAI is not complex: to coordinate independent participants through open protocols to achieve an AI system without a single central authority. Furthermore, by combining blockchain technology, cryptoeconomic incentives, and cryptographic verification mechanisms, it solves the trust problem in anonymous networks, directly addressing the pain points of centralized AI.

• Break down market concentration: Establish a distributed network of computing power, data, and model providers to create a free and competitive market pricing mechanism.
• Alleviating physical constraints: Distributing massive energy demand across power grids around the world.
• Breaking free from geopolitical dependence: Building an infrastructure layer that transcends a single jurisdiction to enable "sovereign AI".
• Increase verification transparency: Replace blind trust in the reputation of tech giants with provable technical means.

Supporters argue that this model can reduce reliance on a single supplier, improve system resilience, and provide opportunities for participation for small and medium-sized countries and businesses.

Meanwhile, institutional investors are shifting from curiosity to substantial investment. Venture capital firms (such as DCG and a16z) are injecting hundreds of millions of dollars into the DeAI protocol; traditional companies (such as Deutsche Telekom) are beginning to participate in the network as validators; moreover, governments in some countries (such as Kazakhstan) are also exploring connecting their idle national supercomputing resources to the decentralized computing market.

Conclusion

As stated in the "State of DeAI 2026" report, DeAI's core value proposition is not that it can completely outperform centralized systems in terms of performance today, but that it provides an underlying architecture that resists monopolies, rejects censorship, and decentralizes power.

With the decreasing cost of dedicated AI hardware (ASIC) and the continued flourishing of open-source models, the window of opportunity for DeAI to address operational challenges has opened. The work of building the foundation for DeAI is only just beginning.

Of course, DeAI is still a long way from becoming mainstream. Its performance, stability, and business model are all still in their early stages. But its significance may not lie in immediately challenging OpenAI, but in providing an alternative.

Historical experience tells us that when an industry has only one option, the problem is often not whether power will be abused, but when it will be abused.

The existence of competition itself is a form of checks and balances.