Over the span of a few weeks, five seemingly unrelated stories from Finland, Texas, Hong Kong, Beijing, and New Delhi traced the same map: the global AI infrastructure race is no longer fought purely in research labs, between teams building the smartest model. The real contest is playing out on three far more concrete fronts: who controls the energy supply, who can raise capital fastest, and who wins over the countries that haven't yet picked a side. Here's the full picture.
Before diving into each episode, the core argument is worth stating plainly. If the first phase of the AI race was defined by a handful of labs competing for the most powerful model, this second phase is shifting decisively toward an infrastructure contest, one built on power plants, billion-dollar fundraises, and diplomatic alliances.
The West's Energy Bet
In the United States and Europe, energy has become the most talked-about bottleneck. Nothing illustrates this more clearly than Google's thirteen-billion-euro investment in Finland, locking the company into a twenty-two-year contract with a nuclear plant that, without the deal, was set to shut down before 2030. For the first time in Europe, AI compute demand has become the financial mechanism keeping a strategically important national energy asset alive.
Around the same time, a signal arrived from an unexpected corner: Bitcoin mining. As we reported when covering IREN, a former Bitcoin miner that reinvented itself as an AI data center giant, the scarce skill today isn't “doing Bitcoin” but managing energy, land, and grid connections at industrial scale. That expertise, built during the mining boom, turns out to be extraordinarily valuable in the AI infrastructure race.
The Map at a Glance
Three fronts, five episodes. Source: SpazioCrypto.
- Energy: Google in Finland, IREN from mining to data centers.
- Capital: Z.AI raises $9.5 billion in under a year.
- Influence: China proposes an open-source AI zone for the Global South.
China's Capital Sprint
On the other side, China's primary challenge isn't energy itself but the speed of raising enough capital to compete, in a landscape made harder by U.S. export restrictions on advanced chips. The clearest example is Z.AI, which raised nearly $9.5 billion in under nine months, across three successive funding rounds that came progressively closer together. Cut off from the most advanced American chips since January, the company has no choice but to build an alternative computational infrastructure, which explains why its appetite for capital has become almost insatiable.
Alongside this, a research report from China's largest telecom operator sketched out what may be the next phase of strategy: according to that forecast, by 2029 inference could account for eighty percent of China's AI compute market, surpassing training for the first time. The shift points away from training a few giant models toward sustaining millions of AI agents running continuously, an infrastructure demand that is even more distributed and costly than today's.
The Third Front: Winning the Global South
There is a third dimension to this competition, less technical and more geopolitical. At the BRICS summit in New Delhi, China's president proposed the creation of a China-led open-source AI zone, offering models, training, and technological cooperation to the bloc's eleven member countries. The timing was deliberate: the announcement came just one day after leading American AI figures publicly debated whether to slow down the pace of AI development, an appeal the White House rejected precisely to avoid ceding ground to Beijing.
This may be the most telling detail of the entire picture. While the U.S. openly debates caution and restraint, China is choosing to accelerate and to share, working to lock in emerging markets before those countries commit to a different technology ecosystem.
The Bigger Picture
Taken together, these five stories mark a phase shift in the global AI race. The challenge is no longer only to build the best model, but to secure the energy to run it at industrial scale, the capital to finance the necessary infrastructure, and the international goodwill to spread its adoption far beyond national borders. Three separate fronts, all converging on the same underlying question: who will succeed in building, first and at sufficient scale, the physical and diplomatic infrastructure on which the next generation of AI will rest?
The practical lesson for anyone tracking this space is that following model announcements alone tells only part of the story. The more consequential contest, and arguably the more durable one, is being decided elsewhere: in power plants, on the trading floors of Asian exchanges, and at diplomatic summits where countries are choosing which technology ecosystem to join. Keeping an eye on these three fronts deserves the same attention that, until now, has been reserved almost entirely for model benchmarks.
