NVIDIA Corporation (NASDAQ: NVDA) heads into the Dec. 21, 2025 weekend with its stock caught in a very 2025 kind of tug‑of‑war: AI demand keeps compounding, but geopolitics and export controls keep grabbing the steering wheel. Shares last traded around $180.99, after a sharp late‑week move that coincided with major policy headlines tied to China and U.S. regulators. [1]
Because it’s Sunday, U.S. markets are closed — so the “today” story is really a “where things stand now” story. And right now, NVDA’s narrative is being rewritten by three forces:
- Washington’s evolving stance on advanced AI chip exports to China (with fresh inter‑agency review steps now underway). [2]
- Wall Street’s valuation pivot — multiple analysts are calling Nvidia “cheap” relative to its own history and to the chip index, despite its enormous size. [3]
- Nvidia’s “full‑stack” expansion — software (SchedMD/Slurm), open models (Nemotron 3), and partnerships are tightening the ecosystem around its GPUs. [4]
Below is the full roundup of current news, forecasts, and analyses relevant as of Dec. 21, 2025, with what it means for NVDA stock.
NVDA stock snapshot: where NVIDIA shares stand heading into the week of Dec. 22
Nvidia shares last traded around $180.99 with heavy volume (over 300 million shares in the latest session data available) after a strong day‑over‑day gain. [5]
That move matters less as a “number on a screen” and more as a signal: investors are repricing policy risk — especially anything that changes the size of Nvidia’s accessible market in China and the intensity of export‑control enforcement.
The biggest NVDA headline: U.S. launches review that could approve H200 shipments to China
What happened
Reuters reported that the Trump administration has initiated an inter‑agency review that could lead to the first approved shipments of Nvidia’s H200 AI chips to China. According to Reuters, the Commerce Department sent license applications to the State, Energy, and Defense Departments for review; those agencies have 30 days to weigh in, but the final decision rests with President Trump under the relevant process. [6]
Why it moved the stock
The market reads this as a potential re‑opening of a major revenue lane — not for Nvidia’s newest Blackwell chips, but for a still‑powerful prior generation product that remains widely used in AI infrastructure.
Reuters also emphasized the political backlash: critics argue that exporting large volumes of high‑end AI chips could strengthen China’s military capabilities and narrow the U.S. advantage in AI. [7]
What the H200 actually is in this context
In Reuters’ reporting, the H200 is described as slower than Nvidia’s Blackwell chips at many AI tasks, but still significant — and it has not previously been allowed for sale in China. [8]
The practical investor takeaway: even “one generation back” for Nvidia can still be years ahead of what buyers can source domestically — which is why policy decisions around “second‑best” chips can still swing billions of dollars.
The earlier trigger: Trump said the U.S. would allow H200 exports to China with a 25% fee
Before the new review process news hit, Reuters reported that President Trump said the U.S. would allow exports of Nvidia’s H200 processors to China and collect a 25% fee on such sales. Reuters noted the arrangement was being finalized by the Commerce Department, and it framed the move as a compromise between allowing no advanced U.S. chips into China versus allowing the very latest Blackwell parts. [9]
This matters for NVDA investors because it puts a big, previously binary question (“China: closed or open?”) into a new, more complex box:
- Licensing + fee structure
- Security review steps
- Ongoing political pushback
- China’s own approval requirements
That’s a lot of moving parts — which is why the story is volatile even when the “direction” sounds positive.
Reuters analysis: H200 chips are already showing up in China via the grey market
Here’s the part that should make any policy analyst’s eye twitch: Reuters also published a deep dive reporting that, even before any official policy opening, H200 chips were already being acquired and used in China via grey‑market channels.
In that report, Reuters said its review of tenders and academic papers showed H200 chips were already being supplied to domestic buyers; it also described tenders and procurement activity involving universities, research institutes, and entities tied to China’s military ecosystem. [10]
Reuters highlighted examples including:
- University labs advertising access to H200s, and academic work using H200 processors for model training. [11]
- Tenders linked to military‑affiliated entities seeking H200 chips or access to H200‑level compute. [12]
- Data‑center project tenders planning large clusters incorporating thousands of accelerators, including H200‑class parts. [13]
For NVDA stock, this creates a paradox: restrictions can suppress official sales while still failing to fully prevent usage, which fuels both (a) enforcement pressure and (b) arguments that regulated legal sales might be preferable to uncontrolled leakage.
China demand story: Reuters says Nvidia is weighing increased H200 output due to interest from Chinese customers
Another Reuters “tell” for investors: Nvidia has told Chinese clients it is evaluating adding production capacity for H200 chips after demand exceeded current output, according to sources cited by Reuters. Reuters reported that major Chinese firms (including Alibaba and ByteDance) had reached out about purchasing, while also noting that China’s government had not yet greenlit purchases and officials discussed whether to allow imports. [14]
Reuters also reported a fascinating policy lever under discussion: a proposal to require bundling H200 purchases with a certain ratio of domestic Chinese chips, aiming to avoid stunting China’s local AI chip industry. [15]
From a stock perspective, this is the heart of the China optionality trade:
- If licenses are approved and Beijing allows imports, Nvidia could see meaningful incremental demand.
- But both governments retain veto power — and the terms (fees, bundling, compliance) can change the economics quickly.
The “offshore compute” loophole: Tencent reportedly accesses restricted Nvidia chips via cloud outside China
One of the most important structural stories this week isn’t about a single license — it’s about how Chinese firms may access top‑tier Nvidia compute without physically importing the chips.
- The Financial Times reported that Tencent secured access to advanced Nvidia AI chips through arrangements involving data centers outside China (including Japan), reflecting a broader trend of using foreign “neocloud” providers to maintain AI capacity amid export controls. [16]
- Barron’s similarly reported that Tencent has reportedly gained access to highly restricted Blackwell chips through cloud services operated by Tokyo‑based Datasection, highlighting a loophole where ownership is restricted but remote usage may still be possible under current rules. [17]
Why NVDA investors care: if policy tightens around “compute access” rather than “chip shipment,” it could reshape Nvidia’s China exposure even if direct exports remain limited.
Export controls meet enforcement: DOJ says it shut down a China‑linked AI tech smuggling network
While export policy may be loosening in some areas, enforcement is also visibly tightening elsewhere.
The U.S. Department of Justice announced it shut down a major China‑linked AI tech smuggling network and said the U.S. seized over $50 million in Nvidia technologies and cash, tied to alleged export‑control and smuggling violations. [18]
For NVDA stock, this is a reminder that:
- Compliance and tracking remain central to Nvidia’s ability to sell globally.
- Enforcement headlines can increase political pressure, which can in turn affect licensing decisions and future restrictions.
Another regulatory catalyst: Nvidia’s $5 billion Intel investment cleared by U.S. antitrust agencies
In a separate Reuters development, U.S. antitrust agencies cleared Nvidia’s planned investment in Intel, according to an FTC notice referenced by Reuters. Nvidia had said earlier it would invest $5 billion in Intel. [19]
Even without full deal terms in the notice, the market tends to interpret “clearance” as removing a near‑term uncertainty — one less regulatory “maybe” hanging over the stock.
Strategically, it also reinforces a 2025 theme: the AI supply chain is getting more interlocked, and Nvidia is increasingly willing to shape the ecosystem through capital as well as chips.
Legal overhang update: Nvidia settles Valeo trade‑secret lawsuit tied to driving assistance tech
Reuters also reported that Nvidia settled a trade‑secret lawsuit filed by automotive supplier Valeo. The case alleged a former Valeo engineer hired by Nvidia took confidential source code related to parking assistance technology and that it was inadvertently revealed during a video call; terms were not disclosed. [20]
This is not the core driver of NVDA’s AI valuation — but for a mega‑cap stock, removing litigation uncertainty is rarely a bad thing.
Nvidia’s “full‑stack” strategy: Slurm (SchedMD) + Nemotron 3 open models
Nvidia acquires SchedMD (the company behind Slurm)
Reuters reported Nvidia acquired SchedMD, tied to Slurm, an open‑source workload manager widely used to schedule large computing jobs in high‑performance computing and AI clusters. Nvidia said the software would remain open‑source. [21]
Nvidia’s own blog post emphasized it has collaborated with SchedMD for over a decade and would keep investing in Slurm’s development while supporting a broad ecosystem. [22]
Why this matters for NVDA stock: Investors don’t pay Nvidia’s multiple just for silicon — they pay it for stickiness. The more Nvidia becomes “the operating layer” for AI factories (schedulers, libraries, systems, models), the harder it is to swap out.
Nvidia launches Nemotron 3 open models
Reuters reported Nvidia launched the Nemotron 3 family of open‑source AI models, with the smallest version available now and larger versions planned for early 2026. [23]
Nvidia’s own release positions Nemotron 3 as optimized for agentic AI applications and highlights performance/throughput claims for the Nano model. [24]
This is also part of a competitive narrative: Reuters noted the rising influence of open‑source AI models from Chinese companies and framed Nvidia’s open strategy as landing at a moment when trust, security, and transparency debates are intensifying. [25]
OpenAI partnership headline — but with a caveat: Nvidia CFO says the $100B deal isn’t finalized
One more Reuters item investors are still digesting: Nvidia CFO Colette Kress said Nvidia’s agreement with OpenAI to invest up to $100 billion is not yet finalized, and she described ongoing work toward a definitive agreement. [26]
That comment matters because the Nvidia–OpenAI relationship sits at the center of broader market concerns about “circular” AI ecosystem deals — where investment, infrastructure build‑outs, and chip purchases can become tightly linked. [27]
Nvidia’s own Q3 FY2026 release had previously described a strategic partnership with OpenAI tied to deploying at least 10 gigawatts of Nvidia systems for OpenAI’s infrastructure. [28]
Nvidia fundamentals: what the company last reported and what it guided next
In its third quarter fiscal 2026 results, Nvidia reported (GAAP) revenue of $57.006 billion, net income of $31.910 billion, and diluted EPS of $1.30. Data Center revenue was $51.2 billion (record), reflecting how dominant the AI infrastructure segment has become in Nvidia’s mix. [29]
For Q4 fiscal 2026, Nvidia guided:
- Revenue of $65.0 billion ±2%
- GAAP / non‑GAAP gross margin of 74.8% / 75.0% ±50 bps
- GAAP / non‑GAAP operating expenses of ~$6.7 billion / $5.0 billion [30]
Nvidia also noted substantial capital return: $37.0 billion returned to shareholders in the first nine months of FY2026 via buybacks and dividends, with $62.2 billion remaining under repurchase authorization at quarter end. [31]
Dividend note for near‑term calendars: Nvidia said it would pay a $0.01 per share quarterly cash dividend on Dec. 26, 2025 to shareholders of record on Dec. 4, 2025. [32]
Next major date for NVDA stock: Feb. 25, 2026 earnings (confirmed on Nvidia’s IR site)
Nvidia’s investor relations events calendar lists “NVIDIA 4th Quarter FY26 Financial Results” on Feb. 25, 2026. [33]
That date is the next major “compression point” for NVDA’s story, because it’s where policy headlines meet numbers: investors will be listening for commentary on supply, demand, China exposure, and the pace of platform transitions.
Wall Street’s NVDA forecast: “cheap,” $275 targets, and what’s behind the call
A big chunk of this week’s NVDA discussion can be summarized as: “Yes, it’s enormous — but the multiple isn’t acting like it.”
- Investors.com reported that Truist’s William Stein raised his Nvidia price target from $255 to $275 (buy rating), and that Bernstein’s Stacy Rasgon reiterated an outperform with a $275 target as well. [34]
- MarketWatch summarized Rasgon’s argument that Nvidia was trading at less than ~25x forward earnings, placing it in a low historical percentile, and highlighted the claim that NVDA’s relative valuation to the chip index had rarely looked cheaper. [35]
- Bloomberg likewise reported the “historic discount” framing versus the SOX (Philadelphia Semiconductor Index). [36]
The bull case embedded in those targets
Even among bullish analysts, the logic tends to rely on three assumptions:
- AI data center capex continues through 2026 (not a one‑year bubble).
- Nvidia continues shipping at scale (platform execution, supply chain, networking, systems).
- The regulatory environment doesn’t remove too much addressable demand, too quickly.
Barron’s reported UBS strategists pushing back on “AI bubble” fears and forecasting global AI spending rising from $423 billion in 2025 to $571 billion in 2026, with a longer‑term AI total addressable market estimate reaching into the trillions by 2030. [37]
Key NVDA catalysts investors are tracking next
1) The H200 export review clock (and China’s response)
Reuters reported the inter‑agency review process involves a 30‑day window for departments to weigh in, with the President holding final authority. [38]
But Reuters also emphasized the second gate: whether Beijing permits Chinese firms to purchase/import the chips. [39]
2) GTC 2026: Nvidia’s next major platform narrative moment
Nvidia’s GTC page lists the conference returning to San Jose on March 16–19, 2026. [40]
Historically, GTC has become the company’s main stage for roadmap, systems, networking, and “AI factory” messaging — all of which can influence both forward estimates and valuation.
3) The product roadmap beyond Blackwell: Rubin timeline visibility
Nvidia previously announced Rubin CPX is expected to be available at the end of 2026, according to its newsroom release. [41]
For NVDA stock, the recurring question is not “does Nvidia have a roadmap?” — it’s “does Nvidia execute the cadence without slippage while everyone else is chasing?”
Risks to watch: what could break the NVDA bull narrative
Even with a strong forecast backdrop, Nvidia’s stock is unusually sensitive to a few specific risk categories right now:
- Policy whiplash / export controls: A favorable signal can reverse quickly if political pressure intensifies, or if enforcement stories (like smuggling cases) drive stricter rules. [42]
- China demand uncertainty: Reuters reporting makes clear that even if the U.S. opens the door, China may impose conditions (including bundling with domestic chips) or block imports. [43]
- “Compute access” loophole closures: If regulators move from “no physical chips” to “no remote usage,” that could reshape demand patterns and partner behavior. [44]
- Deal complexity in the AI ecosystem: Reuters’ note that the OpenAI investment agreement is not finalized underscores how quickly sentiment can shift around big infrastructure partnerships. [45]
Bottom line for NVIDIA stock on Dec. 21, 2025
Nvidia stock is entering the final stretch of 2025 with momentum — but it’s momentum driven as much by policy probability as by product performance.
The near‑term NVDA setup looks like this:
- If the H200 export review progresses toward approvals and China demand converts into licensed orders, Nvidia’s revenue “surface area” expands. [46]
- If Washington tightens rules around remote compute access (the offshore cloud path highlighted in FT/Barron’s reporting), the long‑term China picture could become more complicated again. [47]
- Meanwhile, Wall Street is increasingly framing NVDA not as “expensive AI hype,” but as “a dominant platform trading at a historically low relative valuation,” with notable firms clustering around $275 targets. [48]
And the next time Nvidia can really “settle the argument with numbers” is Feb. 25, 2026, when it reports Q4 FY2026 results. [49]
References
1. www.investors.com, 2. www.reuters.com, 3. www.investors.com, 4. www.reuters.com, 5. www.investors.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.ft.com, 17. www.barrons.com, 18. www.justice.gov, 19. www.reuters.com, 20. www.reuters.com, 21. www.reuters.com, 22. blogs.nvidia.com, 23. www.reuters.com, 24. investor.nvidia.com, 25. www.reuters.com, 26. www.reuters.com, 27. www.reuters.com, 28. nvidianews.nvidia.com, 29. nvidianews.nvidia.com, 30. nvidianews.nvidia.com, 31. nvidianews.nvidia.com, 32. nvidianews.nvidia.com, 33. investor.nvidia.com, 34. www.investors.com, 35. www.marketwatch.com, 36. www.bloomberg.com, 37. www.barrons.com, 38. www.reuters.com, 39. www.reuters.com, 40. www.nvidia.com, 41. nvidianews.nvidia.com, 42. www.reuters.com, 43. www.reuters.com, 44. www.ft.com, 45. www.reuters.com, 46. www.reuters.com, 47. www.ft.com, 48. www.investors.com, 49. investor.nvidia.com
