Advanced Micro Devices, Inc. (NASDAQ: AMD) heads into the final week of 2025 with two narratives pulling investor attention in opposite directions: accelerating demand for AI infrastructure outside the hyperscalers—and tightening, politically charged rules around what high-end compute can be shipped into China.
On a quiet Christmas Day for markets, the AMD story is still moving: export licensing and “fee” frameworks are evolving, large China-based buyers are reportedly evaluating significant orders, and AMD’s own roadmap messaging continues to steer toward rack-scale systems and multi-year AI deployments with marquee partners.
AMD stock: where shares stand heading into year-end
U.S. markets were closed on December 25, 2025, so the latest official trading data reflects Christmas Eve (Dec. 24). AMD shares last traded at $215.04, with an intraday range of roughly $214.00–$216.48 and volume near 8.0 million shares.
Recent price action shows why AMD remains on traders’ radar: the stock saw a sharp move up late last week (including a notable jump on Dec. 19) before settling into a tighter band ahead of the holiday.
That sets the stage for what many investors are really trying to handicap into 2026: how much of AMD’s AI upside converts into durable revenue and margins—and how much gets constrained by geopolitics.
The biggest swing factor: China export licensing and the new “fee” reality for AI GPUs
One of the most consequential developments for AMD’s near-term AI accelerator business is the reopening—under conditions—of a path to ship certain AI chips into China.
Reuters reported that AMD CEO Lisa Su said the company has licenses to ship some MI308 chips to China and is prepared to pay a 15% “tax”/fee to the U.S. government when those shipments occur. Reuters also described the MI308 as a downgraded variant designed to comply with U.S. export controls, and noted it was swept into restrictions alongside Nvidia’s H20 earlier in 2025. Reuters
The Associated Press framed the same framework more bluntly: Nvidia and AMD agreed to share 15% of revenue from certain chip sales to China as part of an unusual arrangement tied to export licenses. AP also highlighted pushback from lawmakers and legal experts who questioned the legal basis and durability of the structure.
Why it matters: even a limited China channel can materially affect accelerator supply/demand balance and inventory planning—especially for a company like AMD that has been scaling Instinct shipments and trying to broaden its AI customer mix beyond a handful of mega-buyers.
But the China opportunity is not simply “open” or “closed.” Reuters also pointed to a headwind: Chinese guidance encouraging state-funded data center projects to use domestically produced AI chips, a policy tilt that could cap the addressable market for U.S. suppliers even when licenses exist. Reuters
Alibaba and MI308: reports of a potentially massive order
Into that export-policy backdrop, the market’s attention snapped to reports that Alibaba is evaluating a significant purchase of AMD’s China-tailored accelerator.
TechNode reported—citing people familiar with the matter—that Alibaba is considering buying 40,000 to 50,000 MI308 AI accelerators from AMD. The report also said the MI308 received U.S. export approval but requires AMD to pay a 15% licensing fee to U.S. authorities, and described the chip as having 192GB of HBM3 memory aimed at long-context inference workloads.
If that scale were confirmed, it would be meaningful for two reasons:
- Unit volume: Tens of thousands of accelerators is “headline” size even in an era of multi-gigawatt AI buildouts.
- Strategic signaling: It would suggest that despite domestic chip policy pressure, China’s largest AI stakeholders still see value in importing U.S. accelerators where permitted.
Still, it remains a report—not a confirmed purchase order—and would likely depend on licensing, delivery schedules, and political conditions on both sides.
AMD’s China engagement: Lisa Su meets China’s commerce minister in Beijing
AMD’s executive-level engagement in China also made headlines earlier in the holiday week.
Reuters reported that China’s Commerce Minister Wang Wentao met with AMD CEO Lisa Su in Beijing, and that both sides discussed AMD’s business development in China and “strengthening cooperation,” without further detail. Reuters
In isolation, such meetings are not unusual for major multinational suppliers. In the current environment, however, they are watched closely as potential signals about regulatory temperature, customer access, and future licensing posture.
AI infrastructure momentum: Vultr’s $1B Ohio cluster using AMD Instinct MI355X
While export policy dominates many trading conversations, AMD’s operational story in AI is also about where demand is coming from—and increasingly, that includes cloud providers trying to position themselves as alternatives to hyperscalers.
Reuters reported that cloud infrastructure company Vultr plans to invest more than $1 billion in a 50-megawatt, 24,000-chip AI cluster in Springfield, Ohio, deploying AMD Instinct MI355X GPUs interconnected with an ethernet-based fabric. Reuters also reported Vultr expects the cluster to be online by early 2026, with the CEO arguing Vultr’s services are typically priced well below hyperscalers.
For AMD, deals like this are strategically important:
- They expand the “AI customer surface area” beyond a few top-tier model builders.
- They help validate AMD’s ability to deliver at scale into cloud deployment patterns.
- They reinforce AMD’s broader pitch: performance plus openness across systems, networking, and software stacks.
The long game: OpenAI’s 6‑gigawatt partnership and the MI450 ramp
AMD’s most ambitious public AI commitment remains its multi-year partnership with OpenAI.
In an AMD press release, the company said OpenAI will deploy 6 gigawatts of AMD GPUs across multiple generations. AMD also stated that the initial 1 gigawatt deployment of AMD Instinct MI450 Series GPUs is expected to begin in the second half of 2026.
One detail that stood out to markets: AMD disclosed it issued OpenAI a warrant for up to 160 million shares of AMD common stock, structured to vest as milestones are achieved—beginning with the initial 1‑gigawatt deployment and scaling as purchases ramp toward 6 gigawatts.
This is not near-term revenue by itself; it’s a multi-year execution test. But it does anchor AMD’s narrative that its Instinct roadmap is not merely “competitive,” but positioned for deployments measured in gigawatts rather than racks.
Rack-scale strategy: HPE, “Helios,” and sovereign/HPC systems
AMD has also continued to frame its AI future as rack-scale infrastructure, not standalone chips.
In a December 2 AMD press release, AMD and HPE announced an expanded collaboration around open rack-scale AI infrastructure. AMD said HPE will be among the first OEMs to adopt AMD’s “Helios” architecture, and the companies also referenced a supercomputer (“Herder”) built on an HPE Cray platform using AMD Instinct MI430X GPUs and next-generation AMD EPYC “Venice” CPUs. Advanced Micro Devices, Inc.
Meanwhile, in its Q3 2025 financial results release, AMD reiterated that customer momentum for its AI platforms is accelerating and pointed to rack-scale deployments, including an Oracle plan for a publicly available AI supercluster powered by the “Helios” rack design and an initial deployment of 50,000 GPUs starting in Q3 2026. Advanced Micro Devices, Inc.
The common theme: AMD is trying to compete at the system level—where switching costs, software integration, and networking architecture can matter as much as raw chip FLOPS.
AMD’s latest fundamentals: Q3 results, Q4 guidance, and the China revenue gap
AMD’s most recent official financial snapshot remains its third quarter 2025 report.
AMD reported Q3 2025 revenue of $9.246 billion, up 36% year over year, and said its Data Center segment revenue was $4.3 billion, up 22% year over year, driven by demand for 5th Gen EPYC processors and Instinct MI350 Series GPUs.
Two lines in AMD’s own earnings materials are particularly relevant to the China story:
- AMD said Q3 results did not include revenue from MI308 shipments to China.
- For Q4 2025 guidance, AMD projected revenue of about $9.6 billion ± $300 million and non-GAAP gross margin around 54.5%, again noting its outlook does not include MI308 revenue from China shipments.
This gap matters because it implies that any restart of MI308 shipments into China—while constrained—could create a “delta” relative to AMD’s baseline guidance framework. At the same time, it’s also a reminder that AMD is trying to grow even without relying on China as a primary data-center AI growth engine.
Washington risk is rising again: the SAFE Chips Act and shifting export-control debate
Export rules aren’t only an executive-branch story—they’re increasingly a legislative one.
Tom’s Hardware reported on a proposed SAFE Chips Act, described as an effort to lock current export rules into law for roughly 30 months, with the practical effect of limiting AMD and Nvidia to MI308/H20-class products for China through that window.
At the same time, broader policy news around competitor Nvidia shows how fluid U.S. posture remains. Reuters reported the Trump administration launched a review that could result in allowing sales of Nvidia’s H200 chips to China (with a government fee), prompting debate about national security and enforcement.
For AMD investors, the implication is straightforward: policy volatility is now a core variable in forecasting accelerator revenue, margin mix, and inventory risk.
Product and roadmap pulse: Zen 6 signals and the EPYC “Venice” era
Even as AI accelerators dominate headlines, AMD’s CPU roadmap remains a pillar of the investment case—especially in data centers, where CPU attach and platform control can influence GPU pull-through.
Tom’s Hardware reported AMD published an early developer-facing Zen 6 document, describing a ground-up redesign on a 2nm node and pointing to an “8-wide” CPU core with strong vector capabilities. Tom’s Hardware
Separately, AMD itself continues to reference next-generation EPYC “Venice” CPUs in rack-scale and supercomputing contexts (including Helios-related deployments), underscoring that AMD sees CPUs and GPUs as an integrated platform play rather than separate product cycles. Advanced Micro Devices, Inc.
Wall Street forecasts: price targets cluster well above the latest close, but models disagree
On December 25, the most visible “forecast” content around AMD wasn’t a single new note—it was the ongoing churn of consensus target updates and valuation models.
Different aggregators show different averages (based on which firms they include and how they weight updates), but many cluster around the high-$200s:
- TipRanks showed an average AMD price target around $279.94.
- MarketBeat listed an average price target around $277.06 and a consensus “Moderate Buy.” MarketBeat
- QuiverQuant summarized recent targets with a median of $283 (based on its tracking of analyst updates).
- StockAnalysis showed a median target of $277, while its average appeared lower—illustrating how a few outlier targets can pull means around depending on methodology.
The gap between the latest close (~$215) and many consensus targets is one reason AMD remains popular in year-end “watch list” content—but it also raises the stakes for execution in 2026.
Valuation debate: “priced for perfection” vs. “platform inflection”
On Christmas Day itself, AAII published a valuation-focused piece that captured the skepticism side of the AMD debate.
AAII’s analysis argued AMD screened as expensive on multiple metrics, citing (as of Dec. 24) a price-to-sales ratio of 10.9, a P/E ratio of 112.6, and an AAII Value Score that translated to a low grade.
A separate December 21 Seeking Alpha analysis similarly leaned cautious—maintaining a Hold stance while acknowledging strong growth momentum, citing premium valuation and margin headwinds from elevated R&D spending as reasons the risk-reward looks balanced at current levels.
These perspectives don’t negate the bullish AI narrative—but they do explain why AMD stock can be hypersensitive to any signal that the AI ramp is slower, more expensive, or more policy-constrained than expected.
What’s next: CES 2026, the next earnings date, and the headlines that could move AMD
With markets reopening after the holiday, AMD has several concrete calendar catalysts that could shape January and early February positioning:
- CES 2026 keynote (Jan. 5, 2026): AMD says CEO Dr. Lisa Su will deliver the CES opening keynote on Jan. 5 at 6:30 p.m. PT, and CES listings mirror that schedule. Investors will listen for updates spanning AI PCs, client CPUs, graphics, and AMD’s broader AI platform narrative.
- Next earnings window (estimated Feb. 3, 2026): Multiple market calendars list AMD’s next earnings date as Feb. 3, 2026 (after market), though such dates can be “estimated” until formally confirmed by the company. Yahoo Finance
- China export signals: Any confirmation (or denial) around large MI308 orders, plus clarity on how durable the “15% fee” framework is, could reprice AMD’s near-term accelerator expectations. Reuters
- AI infrastructure execution: Watch for deployment milestones tied to Vultr’s Ohio cluster timeline and broader rack-scale wins that validate AMD’s system-level strategy.
Bottom line on Dec. 25, 2025
AMD ends 2025 in a familiar but intensified position: strong product and platform momentum, rapidly expanding AI “logos,” and a widening set of infrastructure pathways—paired with policy risk that can reroute revenue, reshape margins, or delay shipments with little warning.
For readers tracking AMD stock news today, the near-term question isn’t whether AI demand exists—it does. The question is whether AMD can translate its 2025 momentum into a cleaner 2026 execution story while navigating a China channel that’s reopening, but only through a narrow and politically controversial gate.
