Citi’s reported Taiwan semiconductor analysis says the AI infrastructure opportunity is spreading beyond GPUs into networking, advanced packaging, custom chips and power systems. It names TSMC, ASE Technology, MediaTek and Hon Hai as Buy-rated stocks, while raising a more consequential question: can AI spending and model revenues sustain the build-out beyond 2028?
Watch Desk analysis
What happened
BigGo Finance says Citi analyst Chen Jiayi’s report draws on investor feedback from a recent US roadshow. The account describes a broader investment thesis: TSMC could benefit from advanced-node production and packaging, ASE from outsourced packaging work, MediaTek from its role in Google’s TPU ecosystem, and Hon Hai from assembling AI server racks.
The report also points to co-packaged optics, networking chips and data-centre power as parts of the expanding hardware picture. These are Citi’s expectations and analysis as reported by BigGo, not confirmed orders or independently measured forecasts.
Why it matters
The AI hardware story is becoming less about one coveted accelerator and more about the systems needed to connect, package, power and deploy compute. That matters to the supply chain: spending on a wider set of components could spread opportunity beyond GPU makers, though it also makes the investment case depend on more companies and more assumptions.
Citi’s central test, as described in the report, is whether cloud providers’ capital spending and AI monetisation can support infrastructure investment beyond 2028. A long list of potential beneficiaries does not answer that question. It does, however, show where one major analyst thinks the next bottlenecks and revenue pools may be.
Our read
This is a useful map of the AI supply chain’s expanding ambitions, not a shopping list for readers to follow. Citi’s reported ratings and forecasts are investment analysis, not proof that every named segment will grow as expected. The most interesting shift is the move from “who makes the GPU?” to “who builds everything around it?” The latter is a much bigger question, and rather less tidy than a single-chip scoreboard.
What to watch
- Whether cloud providers sustain AI infrastructure spending beyond 2028.
- Whether custom-chip, networking and packaging demand turns into disclosed orders.
- Whether co-packaged optics and new power architectures move from plans into deployments.
Discussion spark: If AI infrastructure spending broadens beyond GPUs, which matters more for the next phase: the chips themselves, or the networking, packaging and power that make them useful?
Sources and evidence
- Citi Expands AI Hardware Map, Names TSMC and Three Other Taiwan Stocks as Buys – finance.biggo.com (22 September 2026, 03:36 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.