AI Stocks Today (22.12.2025): Nvidia’s China Chip Pivot, Micron’s AI Memory Boom, Broadcom Margin Jitters, and Oracle’s TikTok Surprise

Monday, 22 December 2025 (22.12.2025) opens the holiday-shortened trading week with AI stocks back in the driver’s seat—but with the kind of “two steps forward, one geopolitical step sideways” energy that has defined much of 2025’s tech tape.

Over the past few sessions, investors have rotated back into the AI trade after a bout of valuation angst, helped by a tech rebound and renewed confidence that enterprise and hyperscaler AI spending isn’t evaporating—it’s just getting more selective. Global markets reflected that shift: Asian shares climbed on the back of tech-led gains, while U.S. futures nudged higher early Monday as traders positioned into the final full week of the year. ^[1]

The fresh AI-stock narrative for today is being shaped by four headline themes:

Nvidia and U.S.–China export policy (a potential revenue unlock wrapped in political risk)
Micron and the AI memory squeeze (high-bandwidth memory is becoming a strategic chokepoint)
Broadcom and the “profitability of AI infrastructure” debate (backlog strength vs. margin pressure)
Oracle’s surprise lift from a TikTok joint venture (plus ongoing scrutiny of massive AI capex)

Below is what’s making news—and why it matters if you’re tracking AI stocks today.

Nvidia stock: H200 sales to China move from theory to process

Nvidia’s AI-chip dominance has always been a story about demand. In late 2025, it’s also a story about permission.

U.S. officials have launched an inter-agency review that could enable first shipments of Nvidia’s H200 AI chips to China, according to Reuters. The Commerce Department has reportedly sent license applications to agencies including State, Energy, and Defense—part of a review process where agencies typically have a set window to weigh in, while the final decision rests with the president. ^[2]

This follows President Donald Trump’s earlier statement that the U.S. would allow H200 sales under conditions that include a 25% fee on such exports, a sharp reversal from the Biden-era approach that restricted advanced AI chip exports to China on national security grounds. ^[3]

Why investors care:

Revenue upside: China is a major buyer of accelerated compute. Even a constrained reopening can be material for a company with Nvidia’s scale.
Policy volatility risk: “Allowed” can still mean “fragile,” especially if congressional pressure grows or enforcement tightens. A senior House lawmaker requested details from the Commerce Secretary on the decision to permit H200 sales, underscoring the political scrutiny around advanced AI hardware. ^[4]
Competitive strategy: Some administration voices argue that selling advanced chips discourages Chinese alternatives from accelerating faster; critics argue it risks strengthening China’s military and AI capabilities. ^[5]

Bottom line: Nvidia’s near-term stock narrative today includes a rare catalyst that is not about next-gen architecture alone. It’s about export gates—and the market is trying to price both the opportunity and the backlash.

Nvidia’s AI “stack” expands: the SchedMD/Slurm acquisition

While the export story is loud, Nvidia also made a quieter—but strategically meaningful—move down the AI software stack: it announced it has acquired SchedMD, the company behind Slurm, a widely used open-source workload manager for HPC and AI clusters.

Nvidia says it will continue to develop and distribute Slurm as open-source, vendor-neutral software, positioning the deal as a way to strengthen the ecosystem that schedules and orchestrates large-scale AI workloads. ^[6]

Why it matters: AI infrastructure isn’t just GPUs. It’s also the scheduling, networking, and orchestration that keeps expensive compute busy. In plain terms: owning more of the plumbing can help Nvidia stay sticky even as competitors fight for silicon share.

Micron stock: AI memory becomes the bottleneck—and pricing power follows

If Nvidia is the “engine,” memory is increasingly the “fuel line.” And right now, that line is tight.

Micron forecast second-quarter adjusted profit at nearly double what Wall Street expected, driven by tight memory supply and booming AI data center demand. Micron also said it increased its 2026 capex plans to $20 billion, reflecting the scale of investment needed to meet AI-driven memory demand. ^[7]

Micron’s importance in the AI trade is tied to HBM (high-bandwidth memory)—specialized memory used in training and deploying generative AI models. Reuters notes Micron is one of only three major HBM suppliers (alongside SK Hynix and Samsung). ^[8]

The market impact showed up fast: Micron continued its rally into Friday, hitting a record closing high and ending that session up 7%, according to Reuters. ^[9]

Why this is a big deal for “AI stocks today”:

HBM is scarce, not optional. The frontier-model buildout doesn’t work the same way without it.
Pricing power shifts. Tight supply tends to mean higher prices and stronger margins—until new capacity catches up.
Capex is the signal. When a memory maker lifts spending plans, it’s effectively saying: demand is real, durable, and worth betting billions on. ^[10]

This is one reason the AI-stock conversation in late 2025 is broadening from GPUs to the entire “compute stack”: memory, networking, power, cooling, and orchestration.

Broadcom stock: massive AI demand, but margin anxiety won’t leave the chat

Broadcom sits at a key intersection of the AI buildout: custom accelerators and the networking infrastructure that feeds them. But Broadcom also became a case study in a new investor question:

“Even if AI demand is huge… how profitable is it?”

Reuters reported Broadcom shares fell sharply earlier this month after the company warned that growing sales of lower-margin custom AI processors were squeezing profitability. Broadcom also disclosed a $73 billion backlog it expects to ship over the next 18 months—proof that demand is there, even if the margin mix is shifting. ^[11]

In other words, Broadcom is living the modern AI paradox:

Win business at enormous scale
Then explain why the economics are still good when the mix changes

This tension matters across AI infrastructure stocks because it influences how investors treat capex-heavy expansion: as a compounding engine… or as a margin trap.

Oracle stock: TikTok joint venture sparks a surprise AI-adjacent rally

Oracle has been one of 2025’s most watched AI infrastructure stories, partly because of its ambition—and partly because big AI spending raises questions about financing and timelines. Reuters has previously highlighted investor anxiety around AI capex and “circular” funding dynamics in the AI ecosystem. ^[12]

Then came an unexpected spark: TikTok.

Reuters reported ByteDance signed binding agreements to shift control of TikTok’s U.S. operations to a group of investors including Oracle, with the new U.S. entity structured so investors hold 80.1% and ByteDance retains 19.9%. Oracle, Silver Lake, and Abu Dhabi-based MGX were cited among major investors. ^[13]

In Friday’s session, Oracle jumped 6.6%, according to Reuters, as the deal boosted sentiment around Oracle’s cloud and strategic positioning. ^[14]

Why it matters for AI stocks today:

Cloud scale and data governance are increasingly central to AI, and Oracle is trying to be a heavyweight in that arena.
Non-AI catalysts can still move “AI stocks.” In 2025, investors often treat big platform companies as bundles of narratives (cloud + AI + regulation + M&A + now TikTok, apparently).

Macro backdrop: holiday-week “Santa rally” hopes meet valuation nerves

AI stocks don’t trade in a vacuum. They trade inside a market that’s trying to balance:

Easing inflation signals (supportive for rate-cut expectations and long-duration growth stocks) ^[15]
Global rate divergence (Japan’s rate hike and a weaker yen added cross-market volatility) ^[16]
High concentration risk in mega-cap tech

A Reuters commentary published today noted that U.S. equity valuations look stretched by some measures (including the Shiller P/E for the S&P 500 above 40) and highlighted how concentrated markets have become, with the largest U.S. tech companies carrying enormous combined value. ^[17]

Meanwhile, the AI boom still looks like the defining equity theme of 2025. The Financial Times’ year-end review pointed to huge gains in AI-linked chipmakers such as SK Hynix and noted Nvidia briefly reached a $5 trillion valuation milestone during the year—symbolic of how dominant AI has been in market leadership. ^[18]

Translation: investors are still “in” the AI story, but increasingly picky about price, profitability, and policy risk.

So… what’s the core narrative for AI stocks today?

Heading into 22.12.2025, the AI trade is being repriced around a more practical question than “Is AI big?”:

Who captures the profits—and who just funds the buildout?

Nvidia’s upside is tied to both product cycles and export policy gates. ^[19]
Micron’s strength is tied to a very tangible constraint: HBM scarcity, which supports pricing power and forces investment. ^[20]
Broadcom’s debate is a micro-version of the whole AI market debate: backlog is real, but margins matter. ^[21]
Oracle’s moves show how quickly sentiment can swing when a platform company gains (or loses) credibility around execution and strategic positioning. ^[22]

What investors are watching next this week

As the market heads into the year’s final stretch, the next big “AI stocks” catalysts are likely to be:

Any official decisions or signals around Nvidia H200 export licenses and broader U.S.–China AI chip policy. ^[23]
AI infrastructure spending discipline: updates that clarify whether capex translates into profitable growth (a theme that hit Oracle and Broadcom earlier this month). ^[24]
Macro data sensitivity: anything that shifts rate-cut expectations can move high-multiple AI names quickly. ^[25]