2025 Week 14 news

This week we look at:

• Trump’s ‘Liberation Day’ tariffs and their impact on compute
• OpenAI’s PaperBench reveals potential for agentic ML researchers
• An AI timeline charting the path to superintelligence by 2027

No liberation for AI under new tariff policies

This week, President Trump’s “Liberation Day” tariff announcement sent shockwaves through global markets. The Dow saw its worst drop since 2020, plunging over 2,000 points, while the Nasdaq entered bear market territory. For the AI revolution that hinges on massive infrastructure and stable global supply chains, these actions could mean profound disruption.

Economists reacted with near-universal condemnation. Yet, the administration appears unfazed. Influential voices suggest Trump won’t back down due to market pressure, indicating the turmoil might even be part of the plan… a high-stakes gamble.

To understand the ‘why’ behind these high-risk policies, it helps to look at both President Trump’s personal convictions and the broader forces influencing his administration. Trump himself holds long-standing, deeply personal beliefs viewing trade deficits as fundamentally unfair and harmful to America. This core conviction, seeing trade imbalances almost as a form of abuse, acts as a powerful engine for policy. But other influential currents are also shaping events, often by explicitly exploiting Trump’s pre-existing obsession:

• Economic Nationalism: This faction, strong within the MAGA base, directly shares and amplifies Trump’s personal trade fixation. They provide the “America First” policy framework – like aggressive tariffs – designed to combat the perceived injustices Trump fixates on, offering him validation and policy tools that resonate deeply with his instincts.
• Radical Disruption: Other influential figures, some linked to techno-libertarian and neoreactionary ideas seeking to dismantle traditional governance, appear to leverage Trump’s willingness to take drastic, disruptive action on trade to advance their own distinct goals. Unitary executive power, DOGE, the crypto reserve, and the rumoured “Mar-a-Lago Accord” (a dollar revaluation and US debt restructuring) exemplify how Trump’s obsessions can become a vehicle for existential attacks on the established governmental and financial order.
• Geopolitical Adversaries: Intentionally or not, when Trump acts forcefully, the resulting economic strain and damaged alliances objectively benefit ideological rivals who want to see weakened US influence and Western cohesion.

These agendas – currently in a dangerous alignment and interacting with Trump’s own potent beliefs about trade – help explain why policies with such high stakes are being pursued. It appears to be a complex mix of deeply felt personal conviction being channelled and leveraged by different power centres within and around the administration.

This strategy is thus riddled with contradictions. Tariffs fuel inflation, they hit US corporate profits and sectors like construction. Furthermore, some analysts suggest that while the tariff-driven inflation might be temporary due to weak underlying money growth, the economy was already fragile. These policies, therefore, risk aggravating existing weaknesses and hitting complex supply chains much as we saw during the pandemic.

How might this impact AI on the supply-side?

• Datacentre costs will spike: Tariffs significantly inflate construction costs. While specific items like copper or semiconductors might have carve-outs, building datacentres relies on a vast ecosystem. Tariffs on imported steel, aluminium, servers, routers, switches, cooling systems, etc will see an average cost increase approaching 15% for a new facility. This highlights how ham-fisted the policy is – targeting specific inputs ignores the complex, interconnected nature of these critical infrastructure projects.
• Energy bottlenecks will worsen: A prime example of this interconnectedness is the supply of large power transformers (LPTs), essential for AI’s enormous energy appetite. The US imports over 80% of its LPTs, already facing 18-24-month lead times. Tariffs impacting steel or other components threaten to exacerbate this critical shortage, directly choking off the power required for AI expansion.
• Solutions won’t arrive quickly: Reshoring complex component manufacturing takes years (5+ for most critical datacentre components). This disruption, coupled with broader economic fears, is will chill investment – hyperscalers are likely reassessing their ambitious plans.

This US-centric disruption creates potential openings elsewhere. Nations in Asia and the Middle East may seize the opportunity to accelerate their own datacentre build-outs and AI capabilities, attracting talent and investment seeking some stability.

Furthermore, Big Tech companies (the ‘Mag 7’), despite taking market hits, may be forced to reconsider their geographic footprint. With upwards of 50% of global compute currently concentrated in the US, the instability could compel them to diversify their critical AI infrastructure globally to mitigate risk. Interestingly, the impact wasn’t uniform across the tech giants. While companies heavily reliant on hardware or consumer goods like Apple, Amazon, Meta, Nvidia, and Tesla saw stock drops of 10-15%, Microsoft and Google proved more resilient, falling only around 5%. Analysts attribute this to their lower direct exposure to tariffs on physical goods, strong cloud revenues, and the inherent difficulty in tariffing enterprise software services.

Ironically, as Treasury Secretary Bessent blames the market crash on China’s DeepSeek (“a Mag 7 problem, not a MAGA problem“), the administration’s own actions risk slowing domestic AI progress and potentially strengthening China’s relative position in the long run. While US export controls on the highest-end chips remain a barrier for China, the broader disruption and alienation caused by US policy can only weaken America’s grip on global AI leadership.

Takeaways: The current US policy trajectory, driven by a volatile mix of nationalism, neoractionary ideas, and personal conviction, poses a significant threat to the trajectory of AI. By snarling supply chains, inflating costs, and creating uncertainty, these actions will hinder the vital infrastructure build-out needed for AI progress. The consequences may extend far beyond the US, potentially accelerating a global redistribution of technological power and undermining the very dominance the administration claims to champion.