AI’s Next Battle: Capital, Power and Control

The artificial intelligence boom is entering a harder, costlier phase

The race to dominate artificial intelligence is no longer being defined only by the release of ever more powerful models. It is increasingly being shaped by giant pools of capital, the scramble for electricity and data-center capacity, abrupt shifts in trade policy and intensifying political fights over who should bear the costs of the technology’s expansion.

A set of developments across the Middle East, Asia, the United States and Australia this week underscored how quickly the AI surge is broadening beyond Silicon Valley software into a global contest over financing, infrastructure and rules.

In Abu Dhabi, MGX, the investment firm backed by Mubadala and G42, closed an AI-focused fund at about $49 billion, making it one of the largest war chests assembled for the sector. The firm has already backed OpenAI and Anthropic, and also supported xAI before its merger with SpaceX, placing it at the center of a widening Gulf effort to convert sovereign wealth into influence over the future of computing.

In South Korea, KKR said it would take control of a $1.3 billion renewables platform with SK as the country positions itself for a rise in electricity demand tied to semiconductors and AI data centers. The deal came just after Seoul announced three large investment projects spanning chips, so-called physical AI and data-center development, underscoring the extent to which governments now see AI as an industrial policy project as much as a technological one.

And in Washington, Anthropic said the Trump administration had lifted export controls imposed only weeks ago on its Claude Fable 5 and Mythos 5 models, a rapid reversal that highlighted how quickly official decisions can alter the competitive landscape for leading AI companies.

Meanwhile in Australia, a different kind of battle is intensifying. Creatives, advocacy groups and some politicians are pushing back against a reported proposal that would grant AI companies broader rights to mine copyrighted works in exchange for more than $50 billion in data-center investment and a compensation fund for artists. Critics have described the idea as a bargain that would shift too much power to technology companies while asking creators to accept uncertain protections.

Taken together, the episodes suggest that the next phase of the AI boom will be fought on three fronts: who finances it, who powers it and who governs it.

Sovereign money moves to the center

The scale of MGX’s new fund is a sign of how much AI has matured, in financial terms, from a speculative software theme into what major investors increasingly view as a long-duration asset class requiring enormous and patient capital.

That shift matters because building the AI economy now means paying not just for startups and model research but also for chips, networking equipment, specialized facilities, energy procurement and global partnerships. Gulf states, flush with sovereign wealth and eager to diversify beyond hydrocarbons, have become especially aggressive in trying to secure a strategic position in that stack.

For Abu Dhabi, AI investing offers more than financial return. It offers leverage in a technology that many governments now regard as fundamental to economic competitiveness and national power. MGX’s links to companies including OpenAI, Anthropic and xAI illustrate how capital from the region is no longer peripheral to the industry. It is becoming part of the architecture of who gets to build it.

Still, important questions remain. It is not yet clear where most of the new capital will be deployed, whether into model developers, chip and cloud infrastructure, international data centers or adjacent sectors like robotics. But the size of the fund alone signals that investors expect AI spending to remain vast and sustained.

The power problem becomes impossible to ignore

If the first phase of the AI frenzy was about computational breakthroughs, the second is about the physical limits that stand in the way of scaling them. Chief among those limits is electricity.

That was the significance of the South Korean renewables deal. KKR’s move to control the platform with SK reflects growing investor confidence that the rise of AI will create enduring demand not only for processors and servers but also for the power generation needed to run them. Data centers devoted to advanced AI workloads are far more energy-intensive than traditional computing facilities, and governments from the United States to Asia have been grappling with how to secure enough supply without worsening grid strain or climate goals.

South Korea has been explicit in tying AI to its broader industrial ambitions. Officials have framed semiconductors, physical AI and data centers as part of a single national growth strategy. But the challenge is not merely attracting projects. It is building enough generation, transmission, water access and permitting capacity to support them.

That tension is emerging globally. As governments compete for AI investment, they are discovering that the prize comes with expensive practical demands. Power, land and infrastructure are becoming as strategically important as algorithms.

Policy can redraw the map overnight

The lifting of export controls on Anthropic’s Claude Fable 5 and Mythos 5 models offered a reminder that government policy can reshape AI competition with startling speed.

The restrictions had been imposed in mid-June, making the reversal unusually swift for an export-control dispute involving advanced technology. The move raised immediate questions about whether the administration’s decision reflected a narrow adjustment specific to Anthropic or the beginning of a broader recalibration in how the United States intends to regulate AI model access abroad.

Export controls have become one of the main tools through which governments try to manage strategic technologies, particularly where concerns about national security and geopolitical rivalry intersect. In semiconductors, such controls have already become a central feature of U.S. policy. Applying similar pressure to frontier AI models has long been discussed, but doing so remains difficult because the technology is less tangible, more rapidly changing and often delivered through cloud-based services.

For AI companies, the stakes are high. A restriction can cut off markets, complicate partnerships and shape perceptions of who is favored by Washington. A sudden removal, as in Anthropic’s case, can just as quickly restore commercial flexibility and alter rivalries among leading firms.

In Australia, the backlash comes into view

If capital and power define one side of the AI boom, public legitimacy may define the other.

In Australia, debate has been building over whether the government should relax copyright rules or create exemptions that would make it easier for AI companies to train systems on creative works, in return for major investment in data centers and related infrastructure. Reports of a proposal involving more than $50 billion in investment and a $350 million annual compensation fund for artists have sharpened those concerns.

For many writers, musicians and other creators, the issue is not simply payment. It is whether governments should rewrite long-established protections in order to subsidize a new industry’s growth. Opponents argue that such a deal would socialize the cultural costs of AI while privatizing the commercial upside. Supporters counter that countries that fail to accommodate AI development risk losing investment, technical capacity and jobs.

That conflict is becoming familiar across democracies. The more AI moves from abstract promise to concrete physical development, the more visible its trade-offs become: heavier electricity use, bigger demands on land and water, and sharper disputes over how training data is acquired. What once looked like a mostly technical debate now looks increasingly like a political one.

The Australian fight is therefore notable not only for its local consequences but also because it may preview arguments that other governments will face. As AI companies seek access to content, energy and public support, they are likely to encounter more organized resistance from groups that feel they are being asked to give up too much in return for uncertain benefits.

A new stage of the contest

The common thread in these developments is that AI is becoming embedded in the machinery of states and markets. It is being financed by sovereign and private capital on a vast scale, tied to national energy planning and industrial strategy, and contested in legislatures and regulatory systems.

That represents a change from the earlier period of the boom, when attention centered on chatbots, benchmark performance and corporate alliances. Those questions remain important, but they no longer capture the whole story. The deeper contest now is over the foundations that make AI expansion possible.

Who controls the money. Who supplies the power. Who writes the rules.

Those questions, more than the next model release, may determine where the industry goes next.