00Opening read
The most important part of an AI system may be the part nobody puts on the cover.
TSMC is approaching another potentially record-breaking earnings report, supported by demand for advanced chips and the packaging that makes them usable. GlobalFoundries has agreed to manufacture a crucial packaging component in New York. That agreement matters because a processor and its memory still need a physical way to work together.
Electricity has the same problem. India generated renewable power it could not move through its grid, even while parts of the country faced shortages. South Korea wants to accelerate new semiconductor factories, but those plants will need water and electricity before they can produce anything.
And in Europe, Airbus has flown an uncrewed combat aircraft that still awaits integration with its European autonomous mission system.
Across chips, grids and machines, the moving constraint is increasingly the handoff between one system and the next.
01What moved
TSMC's next test is packaging, not just profit
Analysts expect another record quarter when TSMC reports earnings on Thursday. Its already reported revenue growth confirms that demand for advanced chips remains strong.
But the question for the physical AI economy is how that demand becomes finished, usable hardware. Chips need memory, packaging and qualification before they can enter a rack. Strong foundry results do not eliminate those other constraints.
New York gets a piece of the advanced-packaging chain
GlobalFoundries has signed a $2 billion agreement to make silicon interposers for TSMC at its Malta, New York, facility.
These components connect chips inside advanced AI packages. Production is expected to ramp in 2028, so the announcement is a commitment to future capacity rather than an immediate increase in supply.
It also gives the United States a more specific role in a manufacturing chain still heavily concentrated in Asia.
India shows why installed power is not delivered power
A new analysis found that India curtailed nearly 1.8 billion kilowatt-hours of renewable electricity during July and August.
The country was dealing with high demand and pressure on conventional power supplies at the same time. The problem was not simply a shortage of generation. It was the grid's inability to absorb and move all the electricity available.
That distinction matters for factories and data centers choosing where to build.
Europe's autonomous aircraft still needs its European brain
Airbus has completed the first European flight of the Valkyrie drone, built with U.S. defense company Kratos.
The aircraft flew, but the European mission software intended to control autonomous operations will be tested later. The next integrated flight is expected in early 2027.
For defense-industrial AI, an airframe that can fly is only one milestone. Software control, certification and reliable coordination with other aircraft determine whether the system becomes operational.
China and Europe are asking harder industrial questions
China's new industrial AI guidance supports factory deployment while warning against speculative technology investment.
Europe, meanwhile, is preparing measures to reduce dependence on single suppliers for critical materials and other strategic inputs. Both developments point toward a more selective phase of industrial policy.
The emphasis is shifting toward what can be deployed, sourced and sustained.
02The throughline
Physical AI is an economy of interfaces.
A chip connects to memory. A factory connects to a grid. A robot connects its sensors and motors to software. A data center connects long-term capital to reliable electricity. Each connection introduces its own engineering, commercial and regulatory work.
That makes headline capacity a poor measure of progress on its own. A wafer-start target is not a qualified package. A solar plant is not guaranteed electricity at the point of use. A successful test flight is not an operational autonomous fleet.
The companies that can manage these handoffs may have a more durable advantage than those that merely announce the largest projects. The next useful measure of progress is not how much has been promised, but how much has become productive.
03What to watch
- October 15: TSMC's earnings and updated guidance, especially advanced-packaging capacity.
- October 15-16: EU leaders' discussion of China trade and critical-supply exposure.
- Early 2027: Airbus' planned Valkyrie flight with its European mission system.
- 2028: The expected interposer ramp in New York and new physical-AI chip production in Dresden.
04Source trail
Reuters, TSMC earnings preview, October 12:
https://www.reuters.com/world/china/tsmc-expected-report-sixth-straight-quarter-record-profit-thanks-ai-boom-2026-10-12/
GlobalFoundries, interposer manufacturing agreement, October 8:
https://gf.com/news-and-events/news/globalfoundries-reaches-agreement-to-establish-us-based-supply-of-silicon-interposers-for-advanced-ai-packaging/
Reuters, Indian solar curtailment, October 12:
https://www.reuters.com/business/energy/india-wastes-4-solar-power-amid-power-shortages-july-august-report-shows-2026-10-12/
Airbus, European Valkyrie test, October 12:
https://www.airbus.com/en/newsroom/press-releases/2026-10-airbus-u740-valkyrie-uncrewed-combat-aircraft-takes-off-for-its-maiden-flight-in-europe
Reuters, China industrial AI guidance, October 9:
https://www.reuters.com/world/asia-pacific/china-issues-guidelines-new-productive-forces-including-ai-2026-10-09/
Reuters, EU-China trade measures, October 12:
https://www.reuters.com/business/autos-transportation/eu-eyes-december-trade-initiatives-after-china-car-import-deal-2026-10-12/
GlobalFoundries, Dresden expansion, October 9:
https://gf.com/news-and-events/news/globalfoundries-marks-dresden-expansion-milestone-announces-next-generation-platform-for-physical-ai/