US private AI investments reached $159 billion in 2025 — more than thirteen times China's ~$12 billion. The US controls 74% of the world's high-performance AI compute and has 4,049 data centers.
What does this mean for small countries?
Competing for the largest models is not realistic
Niche expertise and unique data can create asymmetric advantages
International collaborations amplify small countries' voice
Sweden's Digital Paradox
Sweden ranks #2 globally on GII but the picture is complicated. In the EU's DESI, Sweden has fallen to 6th place. For young people (16–24), digital skills rank 20th in the EU.
Sweden invests 0.11‰ of GDP in AI, compared to Finland's 0.18‰. Meanwhile, AI startups have tripled their funding to €454M in 2025. 35% of companies use AI but 74.7% cite skills shortage as the biggest barrier.
Nordic-Baltic AI Comparison
Finland was first (2017) and invested €250M via FCAI. Elements of AI reached hundreds of thousands globally. Highest in the Nordics on Tortoise (#11).
Denmark focuses on public sector AI. Estonia is the digital champion with e-residency and X-Road.
Together, the region forms a potential AI power with 33 million inhabitants.
Nordic AI Cooperation
NordForsk: NOK 300M in 17 projects. Nordic AI Center: DKK 30M. Nordic AIR: Research environment network. New Nordics AI: Joint international profile.
Nordic side: R&D and compute. Baltic side: government-as-platform and AI governance.
The Skills Gap
WASP with SEK 6.5 billion has 80+ research groups and partners like Stanford, MIT, UC Berkeley. But expertise doesn't reach broadly — 74.7% of companies cite shortage. Finland's Elements of AI reached hundreds of thousands globally.
Sweden's Unique Data Environments
MASAI: 29% more cancer cases detected, 44% less workload. 107 quality registries with clinical data on millions of patients. Biobanks: 150M+ samples. SciLifeLab & MAX IV: Datasets requiring AI.
The EU AI Act
The AI Act classifies systems by risk. Finland implemented nationally Jan 2026, ahead of Sweden. Main requirements apply Aug 2026. GPAI rules give open models some special treatment.
AI Infrastructure
Berzelius at NSC in Linköping. Nordic green compute via renewable energy. The US has 4,049 data centers and 74% of global AI compute.
AI and Defense
NORDSEC/NORDEFCO has a long tradition. AI applications: command support, simulation, logistics, cyber defense. Can regional cooperation expand to joint defense AI?
Possible Futures
AI could increase Sweden's GDP ~9% over ten years. More importantly: what type of AI society do we want to be?
Research nation: Unique data + WASP
Regulatory leader: AI Act as export product
Nordic-Baltic hub: Regional AI power
Trust society: Data sharing via societal trust
New Perspectives
The Nobel Problems: X Prize model for AI. Cities as AI labs: Linköping and Stockholm as smart city hubs. Industrial AI: China's strategy vs our consumer focus — Volvo, ABB, Ericsson as springboard.
China sees AI as productive force. We see it as conversation partner. Both are needed.
Sources: China's State Council; X Prize Foundation; Smart City Index 2025
Where is the technology heading?
Four shifts changing the playing field
From generative to functional AI
From generating to doing: controlling systems, making decisions, executing workflows. AI as executor.
Judgment-focused architectures
Systems that determine when they know enough, should ask a human, or should abstain. Critical for healthcare and law.
Open vs closed AI
Llama, Mistral, DeepSeek vs GPT, Claude, Gemini. Open AI = technological sovereignty for small countries.
The autonomy explosion
METR: AI agents' time horizon ×2 every 7th month. From minutes to days — see the curve below.
Sources: METR Time Horizons 2026; Stanford HAI 2025; EU AI Act
×2 every 7th month
The pace of AI's autonomy development. From minutes to days in three years.
METR Time Horizons, January 2026
Possible Priorities
Possible priorities for Sweden
01
Move from Access to Use
The real challenge is meaningful use — across demographics, geographies, and sectors — so AI lifts communities rather than bypasses them.
02
Build Collaboration
No country wins in AI alone. Anchor Nordic-Baltic cooperation on data, research, and governance — and engage globally on standards and safety.
03
Focus the Efforts
Sweden's edge: high-quality data, world-class research, and a capable defense sector. Concentrate investment where advantages are hard to replicate.
04
Move from B2C to B2B/C2G
Sweden's opportunity is in industrial, professional, and institutional contexts — where domain expertise, long relationships, and trust define who wins.
05
Build Resilience and Sovereignty
Capacity to act independently when it matters, while remaining strategically open. Sovereignty should be functional and reciprocal.
These questions have no simple answers.
That is also the point. Sweden's AI future is shaped by many conversations, experiments and choices over time.
This presentation is intended as an opening, not a conclusion.
Data from February 2026. Sources: Stanford HAI 2025, OECD, Tortoise Global AI Index 2025, WIPO GII 2025, SOU 2025:12, METR Time Horizons 2026.
