The AI infrastructure buildout isn’t slowing down, and Wall Street knows it.
Citi analysts went through the semiconductor landscape after the latest earnings cycle and came back with four names they think still have room to run.
The picks aren’t random as they reflect a clear-eyed view that demand for AI-grade chips stays strong even as investors get pickier about valuations.
If you are trying to cut through the noise, this list is worth your attention.
Nvidia (NVDA)
No brainer, as everyone is starting from here.
Nvidia printed $68.1 billion in Q4 revenue, up 73% from the prior year and data center sales alone came in at $62.3 billion, up 75%.
That one segment now drives nearly 92 cents of every dollar Nvidia earns.
The forward story is just as compelling as the company guided Q1 to roughly $78 billion, clearing Wall Street’s bar by a wide margin.
Hyperscaler spending on AI data centers is on track to exceed $630 billion in 2026, and Nvidia remains the first call for anyone building at scale.
Broadcom (AVGO)
Broadcom doesn’t get Nvidia’s headlines, but it’s becoming just as essential.
The difference is approach: while Nvidia sells general-purpose GPUs to anyone who wants them, Broadcom builds custom AI chips, called XPUs, designed specifically for individual hyperscalers.
Google, Meta, OpenAI, Anthropic. These companies want silicon built around their own models, not someone else’s blueprint.
Q1 fiscal 2026 backed the thesis hard as revenue came in at $19.31 billion, up 29% year over year.
EPS beat consensus. And here’s the bigger picture: AI inference, the ongoing work of running AI models in real products every day, not just training them, is expected to account for up to 70% of all AI computing by 2027.
Broadcom’s custom chip strategy is built for exactly that world.
Texas Instruments (TXN)
Texas Instruments won’t trend on financial circles, but Citi including it sends a signal: the analog chip recovery is real.
Analog chips aren’t glamorous as they manage power, heat, and signals across industrial machinery and automotive systems, but they are everywhere, and they matter.
TI posted Q4 revenue of $4.42 billion, up 10% year over year, and guided Q1 to a range of $4.32 billion to $4.68 billion, beating forecasts.
Industrial and automotive now make up about 75% of revenue, compared to just 43% back in 2013.
Capital expenditure is being cut to $2–3 billion in 2026 after years of heavy investment, which means free cash flow is set to improve materially.
Monolithic Power Systems (MPWR)
This one flies under the radar for most retail investors, but the pivot it made toward AI infrastructure is real, and the numbers are showing it.
MPS reported Q4 revenue of $751.2 million, up nearly 21% year over year, beating estimates. EPS of $4.79 cleared the bar.
The key metric is mix shift. Enterprise data, storage, and computing: the segment tied to AI servers and high-performance computing, now accounts for over 25% of total revenue, a sharp jump from where it stood two years ago.
Management guided Q1 to $770 million to $790 million, pointing to continued momentum.
MPS makes power management chips, essentially the systems that keep AI servers running efficiently at a massive scale.
As data centers keep growing, that stops being a niche and starts being critical infrastructure.
The post These are 4 AI chip stocks Citi wants you buying now appeared first on Invezz
