This AI chipmaker is a double in three months. UBS sees more momentum ahead

UBS reiterated a Buy on AMD and raised its 12-month price target to $700 from $670, implying 28% upside versus Tuesday’s close. The note highlights potential new AI-chip customers (Amazon as a major MI450x customer and possible Anthropic involvement), plus upside from possible fast-inference partnerships with Cerebras and a broader push into custom data-center ASICs. UBS also flagged easing capacity constraints at Taiwan Semiconductor’s packaging services and “very bullish” supply-chain allocation trends (up to C2027), as AMD shares have already surged 112% over the past three months.

Analysis

The real implication is not "more AI demand" but a higher probability that AMD can convert design wins into shipped revenue faster than the market has assumed. If packaging constraints are easing, the bottleneck shifts from supply to execution, which usually benefits the second-tier share-gainers first because they are more levered to incremental allocation than the category leader. That argues for upside in AMD’s data-center mix and for some normalization in its customer concentration risk, while also supporting incremental demand for TSM’s advanced packaging stack.

The next leg is likely less about the headline and more about validation: management commentary on backlog conversion, gross-margin stability, and whether hyperscaler/custom ASIC wins are broad or still one-off. Over 1-3 months, the key catalyst is whether AMD can show that AI revenue is scaling without a severe margin penalty; over 6-18 months, the structural question is whether it becomes a credible second source for accelerators and inference, which would expand its addressable base and support a higher multiple. If those signals fail to appear, the stock can de-rate quickly because the easy narrative already looks crowded.

The contrarian miss is crowding: a stock that has already rerated sharply can still go higher, but the bar for incremental upside gets much higher once the sell-side converges. The market may also be underestimating how much of any new customer news is already being discounted; if the disclosed wins are smaller than expected or too back-end weighted, the first reaction could be a sell-the-news event. The clearest falsifier is any sign that packaging allocation or AI order conversion stalls into the next earnings cycle, or that AMD has to sacrifice gross margin to buy share.

Key Decisions for Investors

• Prefer buying AMD on pullbacks rather than chasing strength; use a 3-6 month horizon and require evidence of data-center revenue acceleration plus stable gross margin. If the next print fails to show order conversion or margin resilience, de-risk immediately.
• Relative-value idea: long AMD / short NVDA or long AMD / short SMH in equal dollars for 1-3 months. Thesis is that AMD has the cleaner incremental share-gain and supply-unlock story from here, while the market has already awarded the dominant player most of the AI premium.
• Use AMD call spreads instead of outright calls for a 3-6 month catalyst window. The move has already been large, so defined-risk upside exposure is preferable to paying full premium into a crowded consensus long.
• Watch TSM as the supply-chain enabler rather than a direct short; any evidence that advanced packaging allocation is still improving is a positive read-through for AMD. Falsify the thesis if packaging lead times re-widen or if AMD cites supply as the reason for delayed revenue conversion.