Global race to build AI chips and memory capacity, and its impact on supply, pricing, and capex

The global race to develop next-generation AI hardware—particularly AI chips and memory capacity—is accelerating rapidly, driven by massive investments, geopolitical strategies, and technological ambitions. This surge is reshaping supply chains, affecting pricing, and prompting significant capital expenditure (capex) across nations and private companies alike.

Major Investments and Capacity Expansions in AI Chips and Memory

In response to soaring demand for AI training and inference, leading countries and corporations are making unprecedented infrastructure commitments:

• Micron's $200 billion expansion plan across Idaho, New York, and Virginia aims to address critical AI memory shortages. As Micron CEO Sanjay Mehrotra emphasized, this massive investment seeks to reduce reliance on Chinese and Southeast Asian supply chains amidst export restrictions, ensuring resilient supply for AI applications.
• Japan’s Rapidus has secured additional ¥267.6 billion in funding, with the Japanese government now controlling up to 11.5% of the chipmaker—potentially rising to 40% if nonvoting shares are converted. This move underscores Japan’s focus on domestic memory fabrication to bolster technological independence.
• South Korea’s SK Hynix has pledged to ramp up production of AI memory chips, responding to surging demand driven by AI training and inference. Their strategic push aims to secure a dominant position in AI memory supply.
• European and Asian initiatives are expanding domestic memory fabrication capacity, partly driven by export restrictions and the desire for technological sovereignty. Japan’s Rapidus and European programs are actively investing to build resilient, onshore manufacturing.

Meanwhile, private sector startups are mobilizing capital to challenge existing giants:

• SambaNova secured $350 million to develop advanced inference hardware, while MatX raised $500 million with the goal of developing more efficient AI training chips, led by former Google TPU engineers. These startups aim to innovate in energy efficiency and performance, addressing the hardware bottlenecks that currently hamper large-scale AI deployment.
• BOSS Semiconductor and Humand are also attracting funding, focusing on energy-efficient AI chips and workforce platforms, respectively, signaling a broader industry push towards specialized AI hardware.

How Hardware Bottlenecks and Export Controls Shape the AI Buildout

A key challenge in the AI infrastructure race is the ongoing shortage of high-performance memory and chips, which are critical for training large models and inference scalability. Micron’s ambitious expansion exemplifies efforts to bridge this gap, but supply constraints persist.

Export controls further complicate the landscape. The United States has imposed restrictions on sales of advanced Nvidia chips, such as the H200 AI chip, to China, delaying deployment and prompting China to accelerate domestic memory and chip manufacturing efforts. This has led to increased investments in onshore manufacturing in Japan, the Netherlands, and other regions, aiming to mitigate geopolitical risks.

These restrictions are also fostering regional strategies for technological sovereignty. Countries are fostering local AI startups and establishing collaborations with global technology giants like Microsoft, OpenAI, and AMD to develop ecosystems less reliant on foreign hardware. For example, Meta's recent multi-billion dollar AMD chip deal reflects the push for more autonomous hardware supply chains.

The Critical Role of Energy and Mineral Resources

The AI infrastructure expansion is fundamentally intertwined with energy capacity and critical mineral supplies:

• Data centers powering AI workloads are highly energy-intensive, prompting substantial investments in renewable and green energy solutions to ensure sustainability.
• Securing supplies of lithium, cobalt, and rare earth elements is crucial for energy storage and chip manufacturing. Countries like India are forming regional alliances to diversify mineral sources and reduce dependence on geopolitically unstable regions.

Geopolitical and Regulatory Impacts

Rising geopolitical tensions are driving a strategic shift:

• Export restrictions are not only delaying hardware deployment but also incentivizing nations to onshore manufacturing and develop technological ecosystems tailored to regional needs.
• Governments are increasingly emphasizing model sovereignty and ecosystem autonomy, fostering local AI startups and partnerships to build resilient, regionally controlled AI infrastructure.
• The U.S. Department of Defense's efforts to classify and mitigate supply chain risks highlight the intersection of AI hardware development and national security, further influencing industry dynamics.

Future Outlook

The ongoing influx of billions into infrastructure projects, hyperscalers’ increased capital expenditure, and national initiatives signal a long-term transformation in AI hardware supply chains. As nations compete for technological sovereignty, secure critical resources, and build resilient supply chains, the AI infrastructure race is evolving into a strategic geopolitical contest.

In summary, the global push to expand AI chip and memory capacity is reshaping the industry landscape—highlighting investments in domestic manufacturing, overcoming hardware bottlenecks, and navigating export controls—all aimed at securing regional dominance in AI capabilities. This era of technological and geopolitical competition will likely define AI leadership for decades to come.