Goldman Sachs raised its target price for AMD to $640: Valuation revaluation and market logic in the AI chip race.

On July 6, 2026, the U.S. semiconductor sector saw a strong rebound. The Philadelphia Semiconductor Index surged more than 4% in a single day, ending a two-day losing streak. As of the close on July 7 (Beijing time), the Nasdaq Composite rose 288.49 points, or 1.12%, to close at 26,121.16; the S&P 500 rose 0.72% to close at 7,537.43; and the Dow Jones Industrial Average rose 0.29% to close at 53,055.91—crossing above the 53,000-point threshold for the first time and setting a record high.

In this rebound, Advanced Micro Devices (AMD) gained 6.61%, closing at $552.05, with a year-to-date cumulative gain of 157.77%. The core catalyst behind this rally was a research report released by Goldman Sachs on the evening of July 5—significantly raising AMD’s 12-month target price from $450 to $640 and maintaining a “Buy” rating.

The increase of this target price amounts to 42%, making it one of the most aggressive revisions among major Wall Street investment banks in the current cycle. Meanwhile, NVIDIA closed at $195.55 on July 7, up 0.37%. The valuation gap and differences in product narratives between these two AI chip giants in the capital markets are sparking broad discussion.

What exactly does Goldman Sachs’ $640 target price mean? How far can AMD’s AI chip narrative go? This article will present a structured analysis across four dimensions: the logic behind the target price increase, an objective review of Goldman Sachs’ historical accuracy, the product competitive landscape between AMD and NVIDIA, and four dimensions of industry-chain beneficiaries.

AMD Gets a Major Target Price Increase from Goldman Sachs: Tracing the Logic

In the report, Goldman Sachs analyst James Schneider noted that the core basis for raising the target price comes from three areas.

First, demand for CPUs from agentic AI remains strong. As AI applications expand from the training side to the inference side, the role of CPUs in AI infrastructure is being reassessed. Goldman Sachs believes that AMD’s expansion of share in the server CPU market will translate into revenue contributions that exceed expectations.

Second, overall AI infrastructure spending continues to expand. The capital expenditure cycle of global cloud service providers remains in an upward trajectory, providing structural support for AMD’s data center business.

Third, expectations for Q2 earnings are strong. Goldman Sachs expects AMD to deliver a “strong” second-quarter earnings report and solid guidance in August. Schneider specifically pointed out that AMD’s comments on engagement with new customers could lift the estimated valuations for the 2027 data center segment, and the market will be highly focused on the progress of AMD’s cooperation with Meta and OpenAI.

It is worth noting that this is already the second time Goldman Sachs has significantly raised AMD’s target price within about two months. Before that, AMD’s stock price had fallen from the 52-week high of $584.73 to $517.82, and Goldman Sachs’ current stance effectively recharacterizes this pullback as a “buying opportunity” rather than a signal of a trend reversal.

Judging from the market reaction, this assessment has been validated in the short term. On July 6, AMD rose as much as 10.5% intraday and ultimately closed up 6.61%.

Goldman Sachs’ Historical Accuracy on Target Prices: A Variable That Requires Careful Scrutiny

When assessing the credibility of the $640 target price, it is necessary to look back at Goldman Sachs’ historical prediction record as a sell-side institution. Objectively speaking, its prediction accuracy has not always been consistent.

In macro forecasts, Goldman Sachs was accurate in only 3 of its 8 predictions on China’s stock market, for a hit rate of about 37%. For example, in 2018 it predicted that A-shares would rise 14% to 17%, but they ended up falling by more than 20%.

In commodities, in 2008 Goldman Sachs recommended “selling gold,” and the gold price rose 5.66% that year and then surged 24.57% the following year. In December 2012, it predicted that the gold price would rise to $1,800, but six months later it fell to $1,180. At the time, the market even joked that it was a “reverse indicator” for gold price trends.

In individual stock ratings, Goldman Sachs’ long-term “Sell” recommendation for Bank of Communications A-shares had a target price below 4 yuan, but Bank of Communications A-shares have repeatedly broken above 6 yuan since 2022, with a March 2026 close at 6.88 yuan.

However, this does not mean Goldman Sachs’ research reports have no reference value. Changes to sell-side target prices often reflect the institution’s stage-based judgment of industry trends and company fundamentals, rather than precise stock price forecasts. The $640 target price implies about 16% upside from the current stock price of $552.05, and that implied return already includes the market’s pricing of uncertainty.

From another angle, Cantor Fitzgerald’s target price was even more aggressive during the same period—raised from $500 to $700, making it the most bullish forecast on Wall Street. Goldman Sachs’ $640 sits in the mid-to-upper range among many institutional forecasts, not an isolated aggressive call.

AMD MI-Series vs. NVIDIA Blackwell: Latest Progress in the AI Chip Battle

Ultimately, the underlying logic behind Goldman Sachs’ target price increase traces back to AMD’s product competitiveness in the AI chip market. Today, the competitive landscape in the AI chip market is evolving from “one dominant player” to “two leaders moving forward together,” but the gap remains significant.

Market share: NVIDIA still has absolute dominance. According to TrendForce, NVIDIA’s global AI chip market share in 2026 is expected to be about 64%, while AMD’s is about 8.6%. A UBS analyst noted that NVIDIA currently controls roughly 80% of the data center AI performance-enhancing hardware market, while AMD accounts for only about 5% to 7%. In the AI training chip market, SemiAnalysis data shows NVIDIA’s market share is 92% in 2026 Q1, and 78% in the inference chip market, nearly unchanged from a year earlier.

Product generations: NVIDIA leads, but the gap is narrowing. NVIDIA’s Blackwell platform (B200/GB200 systems) occupies the lead in 2026 shipments, while Rubin Ultra is expected to begin gradually joining starting in Q4 2026. In the AA-AgentPerf DeepSeek V4 Pro benchmark, the Blackwell architecture delivers better performance per watt than AMD’s MI355X for agent workloads.

But AMD is closing the gap. After system-level tuning, AMD’s Instinct MI355X can substantially match the B200 based on the Blackwell architecture across multiple key metrics: the MI355X comes with 288GB of high-bandwidth memory, which is more than 60% higher than B200’s 180GB.

Next-generation products: AMD’s Helios platform is a key variable. AMD is accelerating the development of its next-generation Helios rack-scale AI platform, featuring the Instinct MI450X GPU, sixth-generation EPYC CPUs, and advanced networking solutions, with volume deployment expected in the second half of 2026. Compared with the MI355X, the MI455X series AMD launched at CES 2026 offers 10x performance improvement, based on 2/3nm process technology, with 320 billion transistors. The flagship model includes 432GB of HBM4 memory.

UBS expects AMD’s Helios platform motherboards to ship in Q4 2026. In addition, AMD has announced it will invest more than $10 billion in the AI ecosystem on Taiwan, expanding strategic cooperation with the supply chain.

Core conclusion on the competitive landscape: NVIDIA still maintains significant advantages in the ecosystem, the software stack (CUDA), and customer stickiness, but AMD is gaining share on the inference side and in some training scenarios by leveraging cost-effectiveness, open standards, and innovations in memory architecture. The core logic behind Goldman Sachs’ target price increase this time is based on an optimistic view that the market has underestimated AMD’s potential to increase its share during the AI infrastructure spending expansion cycle.

Mapping AMD Supply-Chain Beneficiaries

As AMD’s AI chips scale up for shipment, it directly drives benefits across the entire industry chain—from wafer foundry to server assembly. Below is a breakdown of key beneficiary links and targets:

Wafer foundry: TSMC is the most core beneficiary. AMD’s high-end CPUs, GPUs, and AI chips heavily rely on TSMC’s advanced process nodes and advanced packaging. As AMD and ARM continue to expand their market share, more orders for high-end CPUs will flow to advanced process suppliers such as TSMC. The next-generation AMD EPYC central processing unit code-named “Venice” has already moved into mass production using TSMC’s 2nm process, making it the first 2nm product in high-performance computing to enter volume production. On July 7, TSMC’s ADR rose 4.06% and closed at $451.79.

Advanced packaging: ASE Technology Holding is the main beneficiary. The widespread adoption of chiplet architecture and increasing complexity are driving ongoing benefits for advanced packaging manufacturers such as ASE. Each time AMD sells one MI450X GPU, it involves TSMC’s N2/N3 process capacity and ASE’s advanced packaging yield.

Server ODMs: Wistron, Quanta, Inventec, and others. System-level deployment of the AMD Helios platform requires cooperation from server manufacturers. Wistron, Wistron NeWeb, and Inventec are working to build systems based on AMD Helios. The mass production of the three AI platforms (NVIDIA Vera Rubin, AWS Trainium 3, and AMD Helios) will trigger a new wave of shipments across the supply chain, including Wistron, Quanta, Hon Hai, and others.

Other key links: These include substrate supplier Ibiden, thermal solution provider chin-??, server management chip supplier ASpeed, and testing equipment suppliers Chroma and Hongjing, among others.

Bank of America simultaneously raised valuation expectations for supply chain leaders such as TSMC and ASE, believing that advanced process nodes and advanced packaging remain the most defensible segments in the industry chain.

Conclusion

Goldman Sachs raising AMD’s target price to $640 is a snapshot of the broader reassessment of valuations in the AI chip space. Achieving this target price depends on the convergence of multiple factors: the production ramp schedule of the AMD Helios platform and the speed of customer onboarding, the sustained pull from agentic AI demand for CPUs, and a tangible increase in AMD’s market share in the AI training chip market.

Based on current data, NVIDIA’s dominant position in the AI chip market is unlikely to be shaken in the short term—there is still an order-of-magnitude gap between its 64% global market share and AMD’s 8.6%. However, AMD is seeking a larger share in the incremental market by accelerating product iteration, adopting an open ecosystem strategy, and deeply integrating with the supply chain.

For investors, the significance of the $640 target price is not about predicting the exact endpoint of the stock price, but about understanding Goldman Sachs’ analytical framework for judging AMD’s potential to increase share during the AI infrastructure spending expansion cycle. As Goldman Sachs analysts have said, the earnings call in August will be an important near-term stock catalyst, while the “Advancing AI” annual event on July 22 to 23 offers an early window to observe technological progress.

The AI chip race is far from over. Whether AMD’s narrative can evolve from a “follower” into a “fast follower” will be tested in the product cycle over the next 12 to 18 months.

FAQ

1. What is the basis for Goldman Sachs raising its AMD target price to $640?

Goldman Sachs analyst James Schneider said the basis includes sustained strong demand for CPUs from agentic AI, an expansion in overall AI infrastructure spending, and strong expectations for AMD’s second-quarter earnings. Goldman Sachs expects AMD to deliver strong earnings in August and believes the market will focus heavily on AMD’s collaboration progress with Meta and OpenAI.

2. What does the competitive landscape between AMD and NVIDIA in the AI chip market look like?

NVIDIA remains in an absolute dominant position. The expected global AI chip market shares in 2026 are about 64% for NVIDIA and about 8.6% for AMD. In the AI training chip market, NVIDIA has a market share of 92%. However, AMD is narrowing the gap through product iterations such as the MI355X and MI450X, as well as the Helios platform, building some competitiveness on the inference side and on cost-performance.

3. How accurate has Goldman Sachs been historically with its target price predictions?

Goldman Sachs’ prediction accuracy is not stable. Statistics show it was correct in only 3 of its 8 predictions for China’s stock market, for a hit rate of about 37%. It has also made multiple directional misjudgments in commodities and individual stock ratings. Sell-side target prices more often reflect stage-based industry judgments rather than precise stock price predictions.

4. Which industry-chain beneficiaries will be boosted by the ramp-up of AMD AI chips?

Core beneficiaries include wafer foundry TSMC (advanced process nodes), the packaging and testing company ASE Technology Holding (advanced packaging), server ODM firms such as Wistron, Quanta, Inventec, and others, as well as substrate supplier Ibiden, thermal solution provider chin-??, and server management chip supplier ASpeed.

5. How much upside remains for AMD’s stock in the future?

Goldman Sachs’ $640 target price implies about 16% upside from the $552.05 closing price on July 7. Cantor Fitzgerald’s target price is even more aggressive at $700. But actual performance will depend on the production ramp pace of the Helios platform, the AI infrastructure capital expenditure cycle, and the real progress of AMD’s market share gains.