Analysis Alibaba this week opened an AI war chest containing tens of billions of dollars, a revamped LLM lineup, and plans for AI datacenters in Europe. But it also prompted a flurry of questions over how it will achieve all this in an increasingly fragmented IT landscape, when critical resources are in short supply.
Proximity to the demand is key – in order for Alibaba to compete globally, it will need to position itself as close to its users as possible
At the mega vendor's Apsara conference in Hangzhou, China, it detailed the latest iteration of its Qwen3-Omni LLM, which can process text, images, audio, and video while generating text and speech.
Critically, it is available under an Apache 2.0 license, making embedding it more tempting for companies loath to tie themselves into the ecosystems of ChatGPT et al.
But the Chinese giant also sketched out plans for a chain of AI datacenters straddling the Middle East, Europe, Southeast Asia, and Latin America. According to Reuters, it plans to invest $53 billion in AI infrastructure over three years.
Alibaba already operates 91 availability zones across 29 regions worldwide, including existing facilities in London and Germany.
The new blueprint would include its first datacenters in the Netherlands, France, and Brazil.
Alibaba database researcher Dr Feifei Li said in a statement: "Our strategic expansion of global infrastructure is designed to cater for the accelerating demand from forward-thinking customers."
Neal Riley, AI Innovation Lead at The Adaptavist Group, told us: "The international competition for dominance in the GenAI space has been heating up in the model arena, with strong contenders emerging from China like Qwen and DeepSeek. The move by Alibaba is going to see a similar competition across providers of infrastructure for AI tooling and systems.
"Proximity to the demand is key – in order for Alibaba to compete globally, it will need to position itself as close to its users as possible."
It's understood that Alibaba plans to offer a full suite of services from the new European datacenters, from standard cloud and elastic compute services up to big data analytics, machine learning, and AI.
To tempt companies into using its facilities to "design, launch, and scale groundbreaking AI agents and applications," it will offer a range of resources "including up to 2 billion free tokens on Model Studio, Alibaba Cloud's one-stop generative AI development platform, and up to $120,000 cloud credits from Alibaba Cloud."
One thing it's unlikely to be offering, though, is access to Nvidia's GPUs, which have become synonymous with AI infrastructure.
As a Chinese entity, Alibaba is blocked by Washington from buying Nvidia's top-end kit. At the same time, the US wants Nvidia to sell its cutdown H20 parts to China – with Washington taking a cut of those sales.
Beijing meanwhile has reportedly told Chinese firms not to use those Nvidia chips.
So what alternatives are there for Chinese firms – other than sourcing full-fat Nvidia kit on the gray market?
Well, there is Alibaba's own homegrown alternative, the T-Head, which is now considered able to go toe to toe with the H20, without giving a kickback to DC.
So would Alibaba be offering European consumers its own chips rather than the Nvidia parts European governments seem desperate to bring into Europe to boost the continent's AI capacity? Alibaba isn't saying.
But it's also worth noting that while Nvidia is being held back from doing chip business with Alibaba, the two announced a tie-up on physical AI – or robots depending on your point of view.
Added to chip sovereignty, there is the broader question of AI and data sovereignty.
- Alibaba Cloud plans expansion into Europe and South America
- Alibaba Cloud reveals its uptime and efficiency secrets developed by in-house network boffins
- Alibaba looks to end reliance on Nvidia for AI inference
- Alibaba admits Qwen3's hybrid-thinking mode was dumb
European governments and customers are increasingly nervous about the likelihood of Washington leaning on cloud providers and potentially gaining access to European facilities – and the data in them. In June, Microsoft France execs confirmed that they could not guarantee that data on French citizens could never be transmitted to the US government without the explicit agreement of French authorities.
But the Chinese authorities have their own ways of leaning on cloud operators.
So will governments in the EU and beyond see Alibaba's plans as a quick boost to their own autonomy? Or would they see them as a threat and look to stymie them?
The EU's Foreign Direct Investment (FDI) protocols can be deployed when it has concerns over overseas entities investing in critical national infrastructure – and datacenters are undoubtedly seen as such.
That would be just one more hurdle among the power, land, and water shortages that complicate datacenter plans. Not to mention local opposition. But it seems likely that Alibaba will be taking on datacenter capacity from specialist operators. Its first wave of German datacenters, for example, are located in Vodafone facilities in Frankfurt.
This should mean it isn't subject to the EU FDI regime. And the datacenter providers will have already done the heavy lifting around land, power, and planning. If there was any doubt about the datacenters getting beyond the drawing board, would the firm have stuck its neck out and announced them? One notable exception from Alibaba's blueprint was the UK, the site of one of Alibaba's existing datacenters.
A DSIT spokesperson told us: "Whilst we would not speculate on possible investments, datacenters are now regulated as critical national infrastructure. Government can step in to review any investment deal from any country if it might affect national security."
Given the recent $42 billion trade pact between the UK and US, spearheaded by the likes of Microsoft and Google, the UK has enough to chew on for now. After all, it does include around 120,000 GPUs courtesy of Nvidia. ®