by Er. Suhaib Bakshi
Explores how AI’s future depends increasingly on semiconductor manufacturing, infrastructure, and energy systems, highlighting initiatives like Terafab and the shift from code to industrial capacity.
Artificial intelligence is often described through advances in code. Increasingly, it is being defined by capacity. On March 21, 2026, Elon Musk outlined Terafab, a proposed semiconductor manufacturing initiative involving Tesla and SpaceX, with plans to establish advanced chip fabrication facilities in Austin, Texas. Widely reported coverage indicates that the project is intended to support chips for Tesla’s vehicles and humanoid robotics efforts such as Optimus, alongside a second line of chips designed for artificial intelligence systems, including those potentially deployed in space. The proposal remains at an early stage, and no detailed timeline has been publicly confirmed.
At its core, the initiative reflects a practical constraint. The rapid expansion of artificial intelligence has increased demand for high-performance computing hardware. Training advanced models, operating intelligent systems, and integrating AI into machines all depend on specialised semiconductor chips. Industry reporting indicates that demand for advanced chips has risen sharply, while supply remains concentrated among a limited number of highly specialised manufacturers, including Taiwan Semiconductor Manufacturing Company, Samsung Electronics, and Micron Technology. This concentration reflects the complexity of semiconductor fabrication, which requires highly specialised equipment, advanced research capabilities, and tightly integrated supply chains.
The pressure on semiconductor supply has been building for several years. Global demand for advanced chips has accelerated alongside the expansion of artificial intelligence, while manufacturing capacity remains concentrated in a small number of facilities. Even leading producers are reported to be operating near capacity, contributing to longer lead times and increased competition for access to high-performance chips. In this environment, access to computation increasingly shapes the pace at which new technologies can be developed and deployed.
Terafab can be understood as one response to this situation, reflecting an effort to bring design, manufacturing, and application into closer alignment, to improve access to computing capacity over time. This approach is consistent with broader trends in the technology sector, where companies are increasingly seeking greater control over key components of their infrastructure. Vertical integration, once associated primarily with traditional manufacturing industries, is now being explored in the context of digital systems and artificial intelligence.
At the same time, Musk has indicated that Tesla and associated artificial intelligence efforts are likely to continue working with established suppliers such as Nvidia, suggesting that any transition toward greater in-house production would be gradual. This reflects the reality that existing semiconductor leaders have developed capabilities over decades, making immediate substitution unlikely. Instead, new initiatives tend to complement rather than replace established supply chains.
The scale of the ambition has been described in broad terms, with Musk suggesting that the project could, over time, aim for very large levels of computing output, including what he described as “terawatt-level” capacity. Such figures are best understood as indicative of long-term aspiration rather than near-term capability, and remain subject to technical, financial, and practical constraints. They nonetheless illustrate the scale at which future computing infrastructure may be envisioned.
Efforts to expand semiconductor production at this scale involve considerable complexity. Advanced fabrication facilities can require tens of billions of dollars in investment and several years to become fully operational. They depend on precision engineering, global supply networks, and a highly skilled workforce. Projects of this nature are therefore long-term undertakings rather than immediate solutions, unfolding over extended time horizons rather than through rapid change.
More broadly, developments of this kind reflect a shift in how artificial intelligence is understood. For many years, progress in AI was associated primarily with improvements in algorithms and data. Increasingly, attention is turning to the material systems that support these advances, including chips, data centres, and energy infrastructure. Artificial intelligence is becoming not only a matter of software, but also of industrial capacity.
Across the industry, similar patterns are visible. Companies are investing in custom-designed chips, expanding data centre capacity, and exploring new approaches to supply chains. Major technology firms are investing heavily in artificial intelligence infrastructure, including large-scale data centres and specialised hardware, reflecting a wider recognition that access to computation is becoming a central factor in technological development. These investments are also reshaping the geography of technology, concentrating infrastructure in locations with access to capital, expertise, and energy.
The expansion of computing is closely linked to energy demand. Modern data centres require substantial and continuous power, and their growth has prompted increased attention to how energy systems can support computing infrastructure. This has led to greater interaction between technology development and energy planning, with companies exploring ways to secure stable and scalable sources of power. In this respect, computation is no longer only a question of processing information, but also of sustaining it.
There is also increasing interest in how computing systems might evolve beyond traditional terrestrial environments. Some proposals have explored whether elements of digital infrastructure could extend into low Earth orbit, where conditions such as continuous solar exposure are sometimes cited as potential advantages. These ideas remain at an early stage and are subject to technical and regulatory considerations, yet they reflect a willingness to reconsider where and how computation might take place.
This shift is not confined to a single company. Jeff Bezos’s Blue Origin has proposed a large-scale satellite network intended to function as data centres in orbit, reflecting growing interest in low Earth orbit as a potential extension of digital infrastructure. While still in the proposal stage, such efforts suggest that multiple organisations are exploring how computing capacity might expand beyond Earth-based systems, even if practical implementation remains uncertain.
Hardware companies are also adapting to these trends, developing increasingly powerful processors to support advanced artificial intelligence workloads, while newer infrastructure providers are emerging to meet rising demand for computing capacity. This broader ecosystem reflects the growing interdependence between hardware, software, and infrastructure in shaping technological progress.
For many readers, particularly in regions distant from major industrial centres, these developments may appear abstract. Yet their implications are increasingly tangible. As artificial intelligence becomes more integrated into education, communication, healthcare, and economic activity, access to reliable and affordable computing will play a growing role in shaping opportunity. In this sense, infrastructure decisions made in one part of the world can have far-reaching effects elsewhere.
At the same time, it is important to maintain perspective. Semiconductor manufacturing is among the most complex forms of industrial production. It requires sustained investment, specialised expertise, and long development timelines. Established producers have built their capabilities over many years, and new initiatives of this scale typically unfold gradually, shaped by both technical feasibility and economic realities.
For now, Terafab remains a proposal, and its eventual scope and outcomes will depend on a range of technical and practical factors. As with many large-scale initiatives, progress is likely to be incremental rather than immediate. What can be observed with greater certainty is a shift in emphasis, with artificial intelligence no longer defined solely by breakthroughs in code, but by the systems that sustain those breakthroughs over time.
In this light, projects such as Terafab are less about a single factory and more about a broader transition. They reflect an effort to align technological ambition with the material foundations required to support it. The implications of this transition will take time to unfold, yet the direction is becoming clearer.
The future of artificial intelligence will depend not only on what can be imagined, but also on what can be built, maintained, and responsibly sustained. In the end, the significance of such efforts may not lie in the machines they produce, but in the conditions they create. The future of intelligence will depend not only on invention, but on endurance, and in that quiet shift from ideas to infrastructure, the real limits of progress will be set.
(The author writes on digital infrastructure and the industrial foundations of artificial intelligence. Ideas are personal.)
