The Weekly Brief

Enjoying this newsletter? Know someone who would too? Forward this email to them!
Sign up (it’s free)

Signal Takeaway: Electric vehicles are on track to become structurally cheaper than petrol cars across Africa - shifting the long-term economics of mobility in emerging markets.

A new Nature Energy analysis found that battery-electric vehicles in Africa are projected to beat petrol cars on total cost of ownership before 2040, driven by falling battery prices and increasing use of off-grid solar charging. The study also found two-wheeled EVs could reach cost parity even earlier due to lower battery requirements.

The assumption that internal combustion vehicles will remain cheaper in low-income markets is weakening. If EVs become inherently less expensive to buy and operate, adoption becomes an economic decision rather than a climate one.

However, cost parity alone does not solve infrastructure, financing, or grid reliability constraints - which remain binding.

Lower total costs unlock the possibility of large-scale EV adoption in African markets. Realizing that shift depends on charging infrastructure, financing access, and grid improvements.

Signal Takeaway: AI-driven infrastructure demand is now constraining general-purpose server CPU supply.

Intel informed major Chinese customers of up to six-month backlogs for its newest Xeon server processors and raised prices roughly 10% in China. AMD cited 8–10 week lead times for some server chips. Both attributed delays to surging AI and data-center demand.

The AI hardware bottleneck now extends beyond GPUs into core server CPUs. Infrastructure expansion is constrained not just by accelerators but by the broader semiconductor stack.

Rising prices and extended lead times signal that AI-related demand is saturating production capacity.

It constrains near-term data-center buildouts and IT upgrade cycles. It increases incentives for capacity expansion and supply-chain diversification.

Signal Takeaway: AI performance gains are increasingly coming from network optimization, not just compute chips.

Cisco announced the Silicon One G300, a 3nm networking switch chip designed to accelerate AI workloads by dynamically rerouting traffic around congestion points in large GPU clusters. The company claims certain AI tasks can run approximately 28% faster using the chip. Shipments are scheduled for the second half of 2026.

As AI clusters scale, networking becomes a bottleneck. Performance is now constrained by how efficiently GPUs communicate, not just how powerful they are individually.

This shifts attention and capital toward full-stack data-center optimization.

Improved networking unlocks higher throughput per cluster and better utilization of existing GPUs. It reduces one scaling bottleneck but does not resolve underlying energy or supply constraints.

Signal Takeaway: Europe has expanded its ability to prototype next-generation chips domestically.

Imec inaugurated “NanoIC,” a €2.5 billion pilot fabrication line in Belgium to prototype chips beyond the 2nm node. Backed by roughly €1.4 billion in public funding and €1.1 billion from industry, the facility will use advanced EUV lithography to support early-stage development under the EU Chips Act strategy.

Europe historically lacked shared infrastructure for cutting-edge semiconductor prototyping. This reduces a structural bottleneck in advanced chip R&D and strengthens Europe’s position in the semiconductor value chain.

It does not create mass production capacity - but it moves feasibility forward.

It unlocks faster experimentation and design validation for advanced chips in Europe. It lowers the barrier for startups and firms seeking to work at sub-2nm scales.

That’s all for today, please reply to this email if you have any comments or feedback, we’d love to hear from you about what we can do better!

Have you enjoyed this email? Make sure to share it with your friends and colleagues.

See you soon,

Max

Subscribe (it’s free)