/Tech2h ago

Nations Ramp Up Sovereign AI Spending on Local Chips and Data Centers

36323176565.4K

Original post

Zephyr@zephyr_z9#1695inTech

"Sovereign AI" enjoooyers should realize that nation states will be using Nvidia/AMD GPUs to train their sovereign models (which will also get export controlled after a certain threshold)

mon@moninvestor

Invest in - "sovereign AI".

As the US keeps tightening export controls on advanced AI chips, technology and recently new frontier models (Fable 5), governments outside the US are realizing they can't depend on AI they don't control.

So they're racing to build their own. That means their own chips, data centers, and models, on their own soil. That's "sovereign AI," and it's become one of the biggest spending trends in tech.

1. The EU is going all in. Through its InvestAI plan, the EU committed €20 billion to build AI gigafactories, each with around 100,000 AI chips, to train Europe's own very large models instead of relying on US companies. They even have a Commissioner for Tech Sovereignty. The whole goal is for Europe to stand on its own as an "AI continent."

2. Australia too. Its National AI Plan pushes sovereign capability, meaning Australian data on Australian controlled infrastructure. Data center investment there is forecast to reach about AU$26 billion by 2030, and giants are piling in, with Microsoft investing A$25 billion and AWS A$20 billion to build local capacity. The same story is forming across the Middle East, India, and Japan.

3. So how does a beginner actually invest in this? The key insight: you don't need to guess which AI model wins. You invest in the "picks and shovels," the companies that build the chips, data centers, and power that every sovereign AI project runs on. They get paid regardless of whose model wins.

4. The simplest, lower risk way is an ETF. Instead of betting on one company, an ETF holds a whole basket.

For this theme, something like the Global X Data Center & Digital Infrastructure ETF $DTCR, which holds data center operators and digital infrastructure names, packages it into one ticker.

VanEck also just launched a data center supply chain ETF $RACK. This is usually the smartest starting point for a beginner.

5. If you want individual names (higher risk, higher reward), think in layers. For chips, there's Nvidia $NVDA, which sells the GPUs to nearly everyone, including the EU, since it has no chip production of its own yet. For the data center and "neocloud" layer, there's the big 3, $IREN Nebius $NBIS, and CoreWeave $CRWV, which build and rent out the actual compute.

6. Why IREN stands out on the sovereign angle. It has power and data centers across the US, Europe (Spain), and Asia Pacific, the exact regions building sovereign AI. And its recent Mirantis acquisition added the software and Kubernetes orchestration layer that turns raw GPUs into a complete, standalone AI cloud a government or bank can run on its own terms.

> Hardware plus software plus global footprint.

NFA. DYOR.

5:17 AM · Jun 13, 2026 · 45.3K Views

Sentiment

Many users dismissed sovereign AI spending on local chips and data centers as futile and ironic, citing heavy reliance on foreign chips, tiny scale, and the pointlessness of replicating paid APIs.

Pos

8.4%

Neg

91.6%

10 comments with sentiment.

Cluster Engagement

Posts from X

Most Activity

VIEWS20.1KBOOKMARKS27LIKES153RETWEETS5REPLIES15

Zephyr@zephyr_z9

U are either on Chinese Open model/Huawei stack or are dependent on USG for model/compute access

USG won't allow non American entities to have cyber capabilities that could threaten American national security

(tbh they won't even allow private American entities with that kind of capability)

Zephyr@zephyr_z9

"Sovereign AI" enjoooyers should realize that nation states will be using Nvidia/AMD GPUs to train their sovereign models (which will also get export controlled after a certain threshold)

2h20.1K15327

Daniel Koss@daniel_koss

@zephyr_z9 It seems very unlikely NVDIA and AMD chips will not get sold to EU and APAC. That would be hella retarded for US and a crazy push toward the entire world running on Chinese infra.

1h770101

NorthFace@PandaDaytona

@zephyr_z9 The real bottleneck is commodities (copper, uranium, rare earths...).

1h1141

Kinez@_Kinez_

@zephyr_z9 Invest 0$ into AI Just pay nerds to write software, as you always did Problem solved, like no really EUs, Russias, Indias of the world don't really have to participate in the china-america LLLm gorillian$ datacenter build out jerking, just use human brains mfer

1h864

Trilbo Swaggins@Tril1boswagginz

@zephyr_z9 Not really true. GPT 5.5 and Opus 4.8 are still readily available and outperform any chinese models.

The US will probably continue publicly releasing models slightly ahead of China's and just keep the frontier locked up.

2h4375

Dual Acies@DualAcies

@zephyr_z9 So AI is the new Manhattan Project and only a handful of countries will be greenlit to finish the project.

Yup.

2h4433

Kinez@_Kinez_

@HoodMagician3 @zephyr_z9 Noooo i must use 1000$ per token SOTA LLM to center a div on app/website that 3.5 people from some niche industry company use to track inventory of their materials in warehouse you get me? This is P1 task that only some wizards could do before chatgippity and claude were invented

1h141

HoodMagician@HoodMagician3

@_Kinez_ @zephyr_z9 There is no more brains Kinez

1h111

gtsm@getsumm2

Perhaps controversial but I think this is an expected and ultimately bullish outcome. The fact that frontier models are effectively national security good was always going to happen and we’ve seen this unfold over recent months via export controls, sovereign subsidies, etc.

It’s a race which was private enterprise at first but is now sovereigns which we see via China, US, EU investment announcements.

We had hypers blowing cash, credit, now arm offerings into this because you’re done if you don’t spend and research continuously.

Now we’re going to have sovereigns yolo into the AI trade and guess how deep their pockets are? Infinite, money printer go brrr.

We need to hold assets, anything harder than cash is good. Might as well hold what’s going to get an infinite bid.

@TMTLongShort @ContrarianCurse might have good takes on this. Decoupling continues. USA really needs to do something about robotics.

2h2642

Semiconductor Insider@SemiconductorsX

Nations pouring billions into "sovereign AI" (EU, Australia, Middle East, India, etc.) remain fully dependent on Nvidia & AMD GPUs.

US export controls are tightening fast, even allies aren't safe forever.

Market exploding: $15B → $177B by 2035.

But without chip independence, it's mostly theater. This matters for everyone: • Geopolitics: US leverage stays massive • Innovation: Fragmentation slows global progress • Security: Data & defense risks rise • Economy: Higher costs, slower AI benefits for all Hardware sovereignty is the real bottleneck.

Diversify compute now.

2h1901

Sean@SomethingOnSnow

@zephyr_z9 They are going to have to undo the rule.

Imagine the development penalty of having to prevent any cyber security risks and satisfy a regulator before any model release. Open models will smoke anything that has this requirement.

2h1391

Rosy Prosperity@rosyprosperity

@zephyr_z9 In your opinion, which will be the beneficiaries (if any at all) of this phenomenon?

2h409

P.S.@PS2323206362968

@zephyr_z9

2h218

Eddie@eddiepugmire

@zephyr_z9 How is USG going to enforce it? The practical result will be disabling those models for all customers to ensure compliance. If the models remain available even for some domestically, foreign entities will find a way to access them.

2h181

Paperviewr@Paperviewr

@zephyr_z9 much more difficult to do that as the supply chain to create these nvidia/AMD data centers are too linked. If you try to ban nvidia/AMD out of europe they will stop supply ASML, so this is much more difficult to control. only the software layer is easier to control

2h149

Mr Rio@Mr_Rio_

@zephyr_z9 sovereign AI will not save the AI trade, numbers too smol

2h115

The AI Therapist@TheAIShrink

@zephyr_z9 France spends billions on GPU clusters to run the same OpenAI API they paid for in dollars. sovereign independence = 404 error

2h108

NorthFace@PandaDaytona

@zephyr_z9 It reminds of 5G setup in Europe. "Do not use Huawei equipments for 5G" vs. "Huawei is cheaper than Nokia"

1h100

Zakrea@Zakrea_312

@zephyr_z9 True, but UK has ARM, interesting to see if they make a push for it in this direction.

2h95

Asymmetric Focus@asymmetricfocus

@zephyr_z9 And Nvidia/AMD GPUs will almost fully rely on TSMC (Taiwan), which fully relies on ASML (EU/Netherlands).

There is no 100% sovereign, just moving down the stack.

1h291