NEW YORK — Dec. 12, 2025 (6:00 p.m. ET) — A sharp pullback hit many of the U.S. stock market’s most closely watched AI stocks on Friday, as investors digested fresh concerns about profitability, margins, and the pace of payback from the biggest AI infrastructure buildout in modern tech.
By the closing bell, the Nasdaq Composite fell 1.69%, the S&P 500 dropped 1.07%, and the Dow slipped 0.51%, with the Nasdaq marking its lowest close since Nov. 25, according to Reuters’ market wrap. [1]
At the center of the selloff: Broadcom and Oracle. Broadcom’s margin commentary reignited doubts about the economics of custom AI silicon, while Oracle’s AI-heavy cloud expansion and OpenAI-linked data-center plans drew a new wave of scrutiny. [2]
Why AI stocks fell on Friday: “show me the profits” replaces “show me the capex”
The market’s message on Dec. 12 was blunt: AI is still the dominant theme — but investors want clearer evidence of durable earnings power, not just large capital expenditure plans.
Reuters reported that investors rotated away from technology as Treasury yields rose after some policymakers pushed back against easier monetary policy, adding pressure to high-valuation growth stocks. [3]
That macro backdrop amplified a sector-specific debate already raging across Wall Street:
- How profitable are AI chips and AI systems — really?
- How fast will hyperscaler and enterprise AI spending convert into recurring cash flow?
- Are data-center timelines slipping due to labor, materials, and power constraints?
The result was a broad hit to semiconductors: Reuters noted the Philadelphia Semiconductor Index sank 5.1%, its weakest session since Oct. 10, and “every stock” in the index lost ground. [4]
Broadcom (AVGO): strong AI demand, but margin pressure triggers a reality check
Broadcom’s selloff was a headline event for AI investors because the company sits at the intersection of two of the market’s hottest areas:
- custom AI accelerators (ASICs) designed with hyperscalers, and
- the high-speed networking that stitches AI data centers together.
On Friday, Reuters reported Broadcom shares fell more than 11% after the company warned that growing sales of lower-margin custom AI processors were squeezing profitability, feeding fears that the AI boom may be less lucrative than bulls assume. [5]
The core issue: revenue is rising — but investors fear profit per dollar is falling
In its earnings coverage, Reuters highlighted that Broadcom projected first-quarter revenue of about $19.1 billion, above analysts’ estimates, and said AI semiconductor revenue is expected to double to $8.2 billion in the fiscal first quarter. [6]
But Broadcom’s CFO also warned gross margin would fall about 100 basis points sequentially, “primarily reflecting a higher mix of AI revenue.” [7]
Broadcom’s CEO also cited a $73 billion backlog expected to ship over the next 18 months — yet Reuters noted investor concern about customer concentration (the backlog coming largely from just five customers) and a rising mix of systems sales, which can carry lower gross margins. [8]
A big tell for the AI supply chain: customer spending is huge — and concentrated
Reuters also reported Broadcom secured massive contracts this year, including $21 billion from Anthropic in the past two quarters tied to Google’s custom Ironwood chips. [9]
That detail cuts both ways for investors:
- It reinforces that AI demand remains enormous.
- It also underscores that AI capex is increasingly concentrated among a handful of buyers, which can heighten volatility when sentiment shifts.
Oracle (ORCL): OpenAI-linked data centers, higher capex, and a debt spotlight
Oracle’s stock weakness has become a proxy for a broader market question: What happens to the AI trade when the financing and construction realities of mega data centers move into the foreground?
Reuters reported Oracle shares sank sharply after the company warned its fiscal 2026 capital expenditures were expected to be $15 billion higher than previously estimated, stoking doubts about the pace of AI payoff. [10]
The OpenAI infrastructure bet — and the market’s unease
Reuters described Oracle’s leap into the AI infrastructure race as being driven by a $300 billion OpenAI deal, while also noting investor concerns around the company’s debt-fueled buildout. [11]
Those concerns flared again Friday after a Bloomberg report claimed some Oracle OpenAI-related data centers were pushed out — but Reuters reported Oracle denied delays, saying “all milestones remain on track.” [12]
Crucially for AI stock investors, Reuters also pointed to a theme that’s increasingly moving markets: bottlenecks beyond chips. One analyst told Reuters that worries about building data centers — including construction delays and power availability — are becoming a bigger factor as investors scrutinize returns on AI spending. [13]
Nvidia (NVDA): stock drops with the group — but China demand signals remain powerful
Even as many AI chip stocks sold off Friday, Nvidia’s fundamentals remained a major focus — especially after a fast-moving policy shift around China.
Reuters reported Nvidia told Chinese clients it is evaluating adding production capacity for its H200 AI chips after orders exceeded current output levels. [14]
That demand surge followed Reuters’ report that President Donald Trump said the U.S. would allow Nvidia to export H200 processors to China and collect a 25% fee on such sales. [15]
However, Reuters also reported uncertainties remain because Chinese authorities had not yet greenlit purchases, and officials held emergency meetings to discuss conditions — including a proposal that H200 purchases be bundled with a ratio of domestic chips. [16]
From a market standpoint, this creates a two-track story:
- Near-term volatility tied to policy conditions, approvals, and supply allocation (with Reuters noting H200 quantities are limited as Nvidia focuses on newer lines). [17]
- A longer-term demand signal: customers still appear eager to buy high-end accelerators wherever they can legally and operationally do so.
The selloff spread: AI “infrastructure” and adjacent plays get hit hard
Friday’s move wasn’t confined to mega-cap AI chip names.
Reuters’ U.S. market wrap singled out steep declines in several companies that have benefited from AI enthusiasm, including SanDisk (-14.7%), “AI infrastructure” firm CoreWeave (-10.1%), and power-adjacent name Oklo (-15.1%). [18]
That’s notable because it suggests investor caution is extending into:
- leveraged AI infrastructure models, and
- the broader ecosystem built around serving AI data-center growth (storage, power, hosting, financing).
In other words, when the market starts debating “AI bubble” risk, it often pressures the highest-beta parts of the AI complex first.
A rare bright spot: Rivian (RIVN) jumps on AI + autonomy strategy
While most AI-linked names slid, Reuters highlighted one exception in Friday’s tape: Rivian surged after analysts cheered its shift toward a custom chip and a more defined autonomy roadmap.
Reuters reported Rivian rose 17.9%, with the company unveiling its Rivian Autonomy Processor (to be produced by TSMC) and shifting away from Nvidia processors in its driver-assistance stack. [19]
For AI stock watchers, Rivian’s move is a reminder that the “AI trade” isn’t only chips and cloud — it’s also about edge AI in vehicles, robotics, and real-world autonomy, where differentiation can be as much about system design as raw model performance.
Forecasts for 2026: analysts point to Alphabet and Amazon as “best AI stocks next year”
One of the most investor-relevant developments on Dec. 12 was the growing shift from “AI winners by narrative” to “AI winners by measurable monetization.”
MarketWatch reported that JPMorgan highlighted Alphabet (GOOGL) and Amazon (AMZN) among its top AI ideas for 2026, pointing to their AI and cloud infrastructure spending and projecting stronger operating performance next year. [20]
Among the forecasts cited:
- Alphabet: a $385 price target from JPMorgan and expectations tied to AI integration across Google’s ecosystem. [21]
- Amazon: expectations for improved AWS growth and a jump in free cash flow as margins improve. [22]
Whether investors agree or not, the emphasis is telling: analysts are increasingly framing AI upside around cloud revenue acceleration, usage-driven pricing, and operating leverage — not simply capex headlines.
Policy and regulation: new federal AI order adds another variable for AI stocks
Alongside earnings-driven volatility, AI investors also had to digest policy headlines on Dec. 12.
Reuters reported that President Donald Trump’s executive order aimed at blocking state AI laws faces political and legal hurdles, including debates about whether federal agencies can tie AI regulation compliance to programs such as the $42 billion BEAD broadband funding initiative. [23]
The White House also published the order text describing a federal push to evaluate and challenge state AI laws deemed to obstruct national AI policy. [24]
For publicly traded AI companies, the market relevance is indirect but real:
- A more uniform national framework could reduce compliance fragmentation for Big Tech.
- But legal uncertainty and political pushback can add headline risk — which tends to matter most when valuations are already stretched.
AI data centers meet local resistance: why permitting risk is creeping into the narrative
Another Dec. 12 story with potential second-order impact on AI stocks: the physical footprint of AI infrastructure.
Politico reported that the Chandler City Council in Arizona rejected a proposed rezoning for a large AI data center campus amid local concerns (including noise, water use, and perceived limited job benefits). [25]
Pair this with the Oracle debate about construction timelines, and a key theme emerges for 2026: AI isn’t only a software story anymore. For many companies, AI growth now depends on:
- land,
- power interconnects,
- construction labor,
- cooling and water planning,
- and local permitting.
Those constraints can influence the timing of revenue recognition — and the market is increasingly sensitive to timing.
What to watch next week for AI stocks
After Friday’s pullback, investors are now looking for catalysts that could either stabilize the AI trade or deepen the reassessment.
Reuters reported markets are heading into major labor market and inflation data next week (including nonfarm payrolls and consumer inflation), which could influence yields and risk appetite after the October government shutdown disrupted official data releases. [26]
For AI stocks specifically, here are the pressure points to monitor:
- AI margins and pricing power: Can chipmakers and system vendors defend profitability as they scale custom solutions?
- Data-center execution: Any news on delays, power constraints, or financing costs (especially for levered builds) can move high-beta names quickly. [27]
- China/export policy headlines: Nvidia’s H200 demand story is constructive, but approvals and conditions remain a swing factor. [28]
- Rotation risk: If yields rise further, high-multiple AI leaders may stay under pressure even without company-specific bad news. [29]
Bottom line
Friday’s action didn’t “end” the AI bull case — but it did raise the bar. After a year where the AI narrative lifted much of the market, Dec. 12 showed investors are increasingly separating:
- AI adoption (which continues), from
- AI profitability and timing (which is being re-priced in real time).
That distinction is likely to define how AI stocks trade into 2026 — especially for companies most exposed to data-center spending cycles, margin mix changes, and execution risk.
This article is for informational purposes only and does not constitute investment advice.
References
1. www.reuters.com, 2. www.reuters.com, 3. www.reuters.com, 4. www.reuters.com, 5. www.reuters.com, 6. www.reuters.com, 7. www.reuters.com, 8. www.reuters.com, 9. www.reuters.com, 10. www.reuters.com, 11. www.reuters.com, 12. www.reuters.com, 13. www.reuters.com, 14. www.reuters.com, 15. www.reuters.com, 16. www.reuters.com, 17. www.reuters.com, 18. www.reuters.com, 19. www.reuters.com, 20. www.marketwatch.com, 21. www.marketwatch.com, 22. www.marketwatch.com, 23. www.reuters.com, 24. www.whitehouse.gov, 25. www.politico.com, 26. www.reuters.com, 27. www.reuters.com, 28. www.reuters.com, 29. www.reuters.com
