The AI arms race will end centralized cloud services

The news that China is launching its own AI program - DeepSeek - which is reportedly cheaper and uses less data than OpenAI’s ChatGPT, is testament to how fast the AI industry is moving and how aggressive competition is going to get. 

Last month DeepSeek reportedly ran a test that showed its version three language model outperformed those of OpenAI for a relatively minuscule development budget of just $10 million. Additionally, DeepSeek will charge users significantly less - just $2.20 per million tokens, compared to $198 at OpenAI for the same usage. 

Considering the recent commitments we've just sent to the AI industry from the US, it is likely that the powers that be were well aware of this development. This is no doubt why ex-President Joe Biden issued an executive order to fast track investment into AI and use government sites to build infrastructure, while new president Donald Trump has recently signed a $500 billion deal to fund the industry.

This has not been enough to soothe the market, however, which has responded aggressively to the news. On Monday, the NASDAQ fell over 3%, led by AI poster stock NVIDIA, which is nursing losses of over 16% while Microsoft (which owns 49% of OpenAI) fell 2% and Google 4%. Ultimately, with the news that the Chinese can develop AI that is better and significantly cheaper than US companies, investors are now questioning whether these expensive US tech stocks are overvalued.

What is now clear is that we are in an aggressive AI arms race. We are going to see competition heat up significantly, with something of a Cold War erupting between the US and China when it comes to leading this sector. The production of semiconductors is going to be the front line, which is why Joe Biden made US production of these a priority and his last executive order. For China, this only makes the invasion of Taiwan more appealing.

What is not being considered early enough, though, is the cloud storage capacity we are going to need to facilitate this AI technology. Right now, the entire Internet is effectively stored between a handful of companies - and these centralised cloud solutions are simply not fit for purpose when it comes to the AI future. 

Already we see the centralised cloud solutions overburdened, putting extensive pressure on local energy supplies. In comparison, decentralised cloud solutions, which distribute the operation of these services across the world in enterprise-grade data centres will increase capacity exponentially. Meanwhile, decentralised cloud solutions can reduce pressure on local infrastructure and reduce energy consumption significantly.

This is important because the AI future we are clearly and inevitably heading into requires unlimited scalable, globally distributed cloud services with computational resources nearby. AI inference with private or real time data is already one of the short term use cases that require more decentralized networks to be efficient. 

Moreover, from a deeply fundamental and scientific level, "first principle thinking" leads to the conclusion that humankind will inevitably need fully decentralized networks, globally and at massive scale. Unless anybody believes that we're going to do less digitally in the future rather than more, decentralised cloud services are as inevitable as AI.

This is because it is only decentralised cloud solutions that will be able to handle the enormous data processing that the AI digital revolution is going to require. The idea that we continue to rely on AWS, Google and one or two other centralised cloud providers to process all of this information is ludicrous, at best. 

There is a lot of ideology behind decentralisation. Some vehemently believe it is essential for democracy, some simply don't trust centralised cloud providers. 

However, all of this is merely a side note to the central issue: without decentralized cloud solutions, we cannot scale our infrastructure to meet the demands of the current AI expansion.