GM. This is Milk Road PRO, your geopolitical GPS for the trillion-dollar AI buildout, recalculating route while the rest of the market is still reading the old map.
One number to anchor this report:
$600B.
That's how big the sovereign AI market could be by 2030, according to McKinsey. Not total AI spending. Just the slice driven by sovereignty requirements: governments and enterprises demanding that AI infrastructure, data, and models stay inside national borders.
To put that in perspective: that's roughly 2x the entire global semiconductor industry revenue in 2024.
This isn't a policy debate anymore. It's a capital deployment cycle, and it's moving fast.
Why this matters to your portfolio:
In previous reports, we've covered the AI infrastructure squeeze and the race to decouple growth from human labor.
Sovereign AI sits at the intersection of both.
Countries aren't just buying GPUs. They're building entire national AI ecosystems - data centers, energy infrastructure, cloud platforms, and homegrown models - because relying on Silicon Valley for their intelligence layer is a strategic vulnerability.
That creates a second wave of AI infrastructure demand that most investors haven't priced in.
The first wave was hyperscalers spending +$600B on CapEx in 2026. The second wave is governments doing the same thing, on their own soil, with their own money, for their own strategic reasons.
Hereβs what weβre covering today:
- π What sovereign AI actually means (and why "build your own ChatGPT" misses the point).
- π° The global spending map: who's writing the biggest checks.
- ποΈ The AI value chain: where governments are investing, layer by layer.
- π― The investment playbook.
What is sovereign AI?
Sovereign AI is the idea that a country (or region) should own and control its own AI capabilities, compute, data, models, and the infrastructure they run on, rather than renting them from a handful of American and Chinese tech giants.
Think of it like energy independence, but for intelligence.
And honestly, itβs a complete myth. The real goal is resilience.
We can break this down into four dimensions:
- Where data and compute physically sit (territorial).
- Who manages them (operational).
- Who owns the intellectual property (technological).
- Which jurisdiction's laws apply (legal).
In practice, most countries care about a mix of all four.
The trigger? Governments realized that relying on foreign-based cloud providers for healthcare, defense, education, and public services creates a strategic vulnerability. If your country's AI runs on someone else's servers, under someone else's laws, you've outsourced a critical function.
The CrowdStrike/Microsoft outage in July 2024 made this concrete.
A single faulty software update brought air traffic to a near standstill across multiple countries, knocking out supermarkets, banks, and hospitals.
The incident demonstrated how fragile a digital world can be when dominated by digital monopolies.
The global spending map
The numbers are staggering. And they come from everywhere.
Gartner estimates global sovereign cloud infrastructure spending will hit $80B in 2026 alone, growing 35.6% year-over-year. The fastest-growing regions? Middle East and Africa (89% growth), mature Asia-Pacific (87%), and Europe (83%).
But sovereign cloud is just one layer of the stack. When you add national GPU clusters, sovereign AI models, energy infrastructure, and workforce programs, the total is much larger.
Here's the country-by-country picture:
France committed β¬109B in total AI investment, including a partnership with Fluidstack to build one of the world's largest AI supercomputers. 500,000 next-generation chips, 1 GW of compute power, operational by 2026. Mistral AI raised β¬1.7B in Series C (led by ASML with an 11% stake) and launched Mistral Compute, a sovereign AI infrastructure stack designed to eliminate reliance on foreign cloud providers.
South Korea is deploying a $735B long-term sovereign AI initiative. Samsung alone is investing $230B. The government selected five consortia, led by Naver, SK Telecom, LG, NCSoft, and Upstage, to compete for national AI champion status, narrowing from five to two by 2027. South Korea's AI Basic Act took effect in January 2026.
The UK pledged Β£2B for AI and launched a Sovereign AI Unit backed by nearly Β£500M. But there's a sharp tension here: supply-side investment without demand-side adoption creates "managed dependence, not sovereignty." The UK signed Memorandums of Understanding with Anthropic, OpenAI, Cohere, and Nvidia, but these are non-binding agreements that provide access to technology rather than building lasting domestic capacity.
India deployed 62,000 GPUs through its IndiaAI program and pulled in $15B from Google alone for a data center hub in Visakhapatnam, alongside similar commitments from Microsoft, OpenAI, and Reliance Industries. India's data center capacity grew roughly 66% faster than the global average between 2018 and 2025.
Saudi Arabia launched HUMAIN, a sovereign wealth fund-backed initiative to build a full-stack AI ecosystem. The UAE is co-developing Stargate UAE with Nvidia, OpenAI, Oracle, and G42.
Canada announced a $2B Sovereign AI Compute Strategy.
Brazil committed roughly $4.3B through 2028, directing 65% toward business innovation and deployment, not research labs.
Germany is tackling sovereign AI from the enterprise side. 53-59% of German companies still rely on non-EU service providers for AI applications and cloud infrastructure, one of the highest dependency rates in Europe. But two of the country's biggest corporate players are building sovereign alternatives.
The Schwarz Group (parent of Lidl and Kaufland) broke ground on an β¬11B data center in LΓΌbbenau, Brandenburg, on the site of a former coal power plant. With capacity for up to 100,000 GPUs and 300 MW of power, it's set to be larger than any existing AWS, Microsoft, or Google data center in Germany. The facility will power STACKIT, the group's sovereign cloud platform.
Deutsche Telekom, meanwhile, launched its Industrial AI Cloud in Munich in early 2026, a β¬1B sovereign AI infrastructure with 10,000 Nvidia Blackwell GPUs, built together with Nvidia and SAP.
Telekom is also building out its T Cloud Public as a sovereign European alternative, promising full feature parity with U.S. hyperscalers by the end of 2026. The two companies are now in talks to jointly build an "AI Factory" and apply for EU funding together.
The pattern is clear: this isn't one or two countries experimenting. It's a global buildout, running in parallel to the hyperscaler buildout, creating a second demand curve for the same infrastructure.
The AI sovereignty value chain
Uh, Ohβ¦ π§ The rest of this report is exclusive to PRO members!
Already a PRO member? Log in here.
WHATβS LEFT INSIDE? π
- The layers of sovereign AI telling us where to look for investment opportunities.
- The sovereignty choke points: the few suppliers every country depends on.
- The investable playbook: the clearest opportunities by layer.
Upgrade your subscription today to unlock access to all of the milky insights above, PLUS:
- More where that came from! Get two PRO reports every month, that boil big-picture shifts down to specific investment plays across chips, cloud, energy infrastructure and more.
- Monthly AMAs: Your direct line to our lead analysts. Join our monthly "open mic" sessions to grill us on everything from the latest Tesla FSD data to picking the right energy stock. ποΈ
- Access to the PRO community: Where you can chat directly with our analysts and fellow PRO members in our private Discord. Ask questions, share ideas, and stay plugged in between reports. π«
- The AI portfolio: A live look at exactly where we are putting our capital to work across the AI ecosystem. π
Already a PRO member? Log in here.
