Why Is Nvidia Stock Up Only 5% While AMD Is Up 300%? What's Going On With NVDA

Nvidia stock is supposed to be the AI trade. Nvidia makes the GPUs that power virtually every major AI training cluster. Nvidia has approximately 80% market share in AI accelerators. Nvidia's revenue grew 65% in fiscal 2026 to $216 billion. Nvidia has a $5 trillion market capitalization that makes it the most valuable company in the world by that measure.

And Nvidia stock is up approximately 5% year to date while AMD is up 300%, Micron is up 667%, and SanDisk is up over 600%.

The divergence is so stark and so counterintuitive that it has become one of the most discussed anomalies in technology investing in 2026. Understanding why it happened, whether it reflects something fundamentally wrong with Nvidia's position, or whether it represents exactly the kind of underperformance that precedes the next leg higher, is more useful than simply noting the performance gap.

The Expectations Problem That Has Weighed on Nvidia All Year

The starting point for understanding Nvidia stock's underperformance relative to its AI chip peers is the same expectations mechanism that caused Samsung stock to fall on a 19 fold profit increase last week.

When a stock has appreciated 900% in two years, as Nvidia did between late 2022 and early 2025, it prices in an extraordinary amount of future performance before that performance is delivered. Every institutional investor who wanted Nvidia exposure because of AI has already bought it. The $5 trillion market capitalization reflects a level of expectation-setting that requires continuous positive surprises rather than merely strong results to generate further appreciation.

AMD's 300% return this year happened from a dramatically lower starting valuation. AMD entered 2026 at a forward multiple that reflected skepticism about whether it could meaningfully compete with Nvidia in AI GPUs. Every positive development, every Helios deployment announcement, every Meta and OpenAI commitment, was new information to a market that had not priced those outcomes. The distance between where AMD was priced and where the business delivered was large enough to produce extraordinary returns.

Nvidia entered 2026 already priced for the best case AI scenario. Its revenue beats, while impressive in absolute terms, have been modest relative to what the market was already assuming. A company growing at 65% annually on $216 billion in revenue is an extraordinary business, but the market's expectations were close enough to the actual delivery that the positive surprise required to move the stock was smaller than analysts anticipated.

Why AMD, Micron, and SanDisk Caught Up Rather Than Nvidia Falling Behind

The performance divergence is better understood as AMD, Micron, and SanDisk catching up to Nvidia's AI infrastructure narrative rather than as Nvidia losing something.

AMD's EPYC server CPU market share gains against Intel were not priced into the stock at the start of the year. The Helios GPU platform winning Meta's six-gigawatt commitment and OpenAI's first-gigawatt deployment was not priced into the stock. Lisa Su's revision of the server CPU market growth from 18% to 35% annually was not priced into the stock. Each of those developments was genuinely new information that arrived into a stock trading at a fraction of the multiple Nvidia commanded.

Micron's HBM trajectory was similarly unpriced at the start of the year relative to what the actual demand environment delivered. Q3 gross margins of 84.9% were not in any consensus model at the start of the year. Q4 guidance of $50 billion in revenue exceeded the most optimistic pre-year forecasts by a substantial margin. A stock trading below $200 at the start of a year that delivers $50 billion quarterly guidance has mathematical room for the kind of appreciation that Micron's 667% return represents.

Nvidia's extraordinary performance was already visible and already priced. The company that went from $10 billion in quarterly revenue to $81 billion in eight quarters had already demonstrated what AI demand looked like at scale. The stock that had already gone up 900% had already captured most of the value creation from that revenue trajectory. The remaining upside required either acceleration beyond the extraordinary pace already demonstrated or multiple expansion beyond the already extraordinary multiple already assigned.

The Specific Headwinds That 2026 Added

Beyond the starting valuation problem, 2026 introduced specific headwinds for Nvidia stock that were not present in 2024 or 2025 and that partially explain why the underperformance extended through the year rather than resolving quickly.

The customer chip development narrative has been the most persistent headwind. Reports throughout 2026 have documented the acceleration of custom AI chip development by Nvidia's largest customers. Google's TPU expansion to external cloud providers announced today is the most recent example. Meta's development of its own AI chips from September is another. Microsoft's development of MAI models that reduce OpenAI API calls touches the software layer but reflects the same dynamic of reducing vendor dependency.

The concern is not that any of these customer chip programs replaces Nvidia immediately. It is that the long-term demand trajectory for Nvidia GPUs flattens as hyperscalers route an increasing share of inference workloads to cheaper proprietary chips while continuing to rely on Nvidia for training at the frontier. If inference is where the volume growth is and inference is where customers substitute away from Nvidia first, the demand curve for Nvidia's products looks different in 2028 and 2029 than the current trajectory implies.

The export control uncertainty has been a second headwind. The US government's restrictions on selling advanced AI chips to China have evolved throughout 2026, with Nvidia creating H20 and other modified chips for the Chinese market that carry lower average selling prices and lower margins than the H100 and H200 products they replaced. The regulatory environment for China sales has created a specific revenue uncertainty that the total company numbers obscure.

The Kyber NVL144 delay report from last week, even though Nvidia denied it and confirmed the schedule, illustrated how sensitive the stock has become to any suggestion that the product roadmap cadence might slow. A single report that Nvidia quickly denied caused enough concern to generate significant coverage and search volume, which tells you that market confidence in Nvidia's roadmap execution is not as unquestioning as it was at the peak.

What TSMC's Record Quarter Actually Means for Nvidia

Today's most important positive development for Nvidia stock was not anything Nvidia itself announced. It was TSMC reporting record Q2 revenue and confirming that AI demand shows no signs of slowing.

TSMC is Nvidia's primary manufacturing partner for its most advanced chips. Every Blackwell GPU, every H100, every upcoming Vera Rubin and Kyber chip runs through TSMC's fabs. When TSMC reports record Q2 revenue driven by AI demand, the most direct interpretation is that Nvidia's orders to TSMC are sustaining or accelerating rather than declining.

TSMC's CEO confirming that AI demand remains strong and that the company sees no signs of slowing is the most credible available real-time signal about Nvidia's demand pipeline because TSMC sees Nvidia's actual production orders rather than analyst estimates or management commentary. A manufacturing partner reporting record AI-driven demand while making public statements about demand strength is the closest thing to a demand verification that exists outside of Nvidia's own earnings report.

The $145 billion supply commitment that Nvidia has disclosed represents the company's own assessment of how much production capacity it needs to meet committed customer demand. Supply commitments of that scale are made against contracted or highly visible customer orders rather than speculative demand forecasts. The combination of TSMC's record quarter, Nvidia's $145 billion supply commitment, and the AI executive statements that AI executives told CNBC they see no signs of overcapacity in the AI buildout collectively paint a demand picture that contradicts the underperformance narrative.

The Goldman and Morgan Stanley Read That Matters Today

Goldman Sachs and Morgan Stanley both addressed Nvidia stock today in ways that are worth taking seriously rather than treating as routine analyst commentary.

Goldman Sachs highlighted that Nvidia's forward price-to-earnings ratio of approximately 21.7 times is close to the average S&P 500 multiple and dramatically below Nvidia's five-year average of 72 times. That compression makes Nvidia, on a valuation basis, less expensive relative to the market than at essentially any point since before the AI boom began. A company that grew revenue 65% in fiscal 2026 and is guiding for continued strong growth trading at the same multiple as the average S&P 500 company is a specific valuation anomaly that Goldman is pointing investors toward.

Morgan Stanley's note today said investors should jump on the bandwagon for Nvidia, framing the current price as a window that is unlikely to remain open indefinitely. Morgan Stanley's AI demand analysis has been consistently tracking hyperscaler capital expenditure commitments that point toward continued extraordinary GPU demand through 2027 and into 2028.

Both firms are making essentially the same argument from different analytical angles. The stock that has underperformed every AI chip peer in 2026 is the same stock that sells 80% of AI accelerators, reports the strongest margins in the semiconductor industry, and is trading at a forward multiple that implies the market expects the AI boom to end immediately rather than continue at any pace.

Why Nvidia Stock's Underperformance May Be Setting Up the Next Move

Market history provides several examples of dominant companies that underperformed during periods when smaller, cheaper peers were catching up to a theme, and then dramatically outperformed when the theme's benefits began flowing primarily to the established leader rather than to the catch-up plays.

Apple underperformed the smartphone sector during the period when Android manufacturers were closing the feature gap. Once the ecosystem advantages became undeniable, Apple dramatically outperformed. Amazon underperformed during the period when cloud computing competitors were closing the gap with AWS. Once the enterprise commitment to AWS infrastructure became too embedded to reverse, Amazon dramatically outperformed.

Nvidia's current situation contains elements of that pattern. The AI chip theme has been benefiting AMD, Micron, and other companies whose positioning was not fully priced relative to the opportunity. The investors who bought those companies when they were cheap captured the performance that came from the market recognizing what was already true about the business. Nvidia, which was already recognized and already priced, has seen flat performance while the recognition process played out for its peers.

The question is whether Nvidia is at a similar inflection point where the catch-up plays have largely captured their available upside and the dominant established leader is about to reassert its return advantage. The August 25 to 26 earnings report is the nearest-term test of whether the demand narrative that Goldman and Morgan Stanley are articulating translates into financial results that move the stock.

What the August 26 Earnings Need to Show

Nvidia reports fiscal Q2 2027 results on August 25 to 26. Q2 guidance of approximately $45 billion in revenue was set during the Q1 report and represents the specific bar the earnings need to clear to validate the buy case at current prices.

The key question beyond the revenue figure is what management says about the demand trajectory into fiscal Q3 and Q4. If Sundar Pichai's confirmation of Google Cloud compute constraints and the AI executives' statements to CNBC about seeing no overcapacity reflect actual demand that is reaching Nvidia's order books, Q2 results should beat the guidance with comfortable margin and Q3 guidance should imply continued sequential growth.

The Kyber platform timeline is the product roadmap question that August 26 will address in a formal earnings context for the first time since the denial of the delay report. Any management specificity about Kyber's deployment timeline, the customers who have committed to the platform, and the revenue contribution expected from Kyber in fiscal 2027 would provide the forward visibility that investors who are concerned about the roadmap are waiting for.

For those looking to participate in global financial markets, having access to the right trading platform matters. WEEX offers crypto and stock trading products, covering major global markets including US stocks and digital assets.

Conclusion

Nvidia stock's 5% year-to-date return while AMD is up 300% and Micron is up 667% is one of the most discussed anomalies in technology investing in 2026. The explanation is not that Nvidia's business has weakened or that the AI boom has bypassed the company that supplies 80% of AI accelerators. The explanation is that AMD and Micron were cheap entering the year relative to their AI opportunity while Nvidia was expensive, and cheap stocks with improving fundamentals outperform expensive stocks with strong fundamentals in environments where the AI narrative is expanding rather than contracting.

The specific question today is whether that dynamic is exhausted. AMD at 73 times forward earnings, Micron at 22 times forward earnings after a 667% run, and SanDisk at a dramatically compressed multiple after an 800% rally from its low all have less catch-up potential than they did at the start of the year. Nvidia at 21.7 times forward earnings with 61 analysts averaging a $302 target and both Goldman and Morgan Stanley explicitly recommending the stock today is the AI chip play that has the most valuation compression remaining to correct.

August 26 is where the thesis gets its next test. TSMC's record quarter today is the most recent external validation that the demand for Nvidia's chips has not slowed. Whether that demand appears in Nvidia's Q2 revenue and Q3 guidance is the financial confirmation that makes the 5% year-to-date return look like a base rather than a ceiling.

FAQ

1. Why has Nvidia stock underperformed AMD and Micron in 2026?
Nvidia entered 2026 already priced for the AI opportunity at a premium multiple reflecting its dominant market position. AMD and Micron entered the year with valuations that did not fully reflect their AI upside. As the market recognized AMD's server CPU gains and Helios GPU deployments and Micron's extraordinary HBM margins, those cheaper stocks delivered larger returns from the same AI theme that Nvidia was already fully priced for.

2. Is Nvidia stock cheap at 21.7 times forward earnings?
Goldman Sachs described the 21.7 times forward P/E as compelling, noting it is close to the S&P 500 average and dramatically below Nvidia's five-year average of 72 times. For a company growing revenue at 65% annually with 80% AI accelerator market share, the current multiple implies the market is pricing a rapid end to AI demand rather than its continuation at any pace.

3. What did TSMC's record Q2 revenue mean for Nvidia stock today?
TSMC is Nvidia's primary manufacturing partner. Record TSMC Q2 revenue driven by AI demand confirms that Nvidia's production orders are sustaining or accelerating rather than declining. TSMC's CEO confirmation that AI demand shows no signs of slowing is a real-time external validation of Nvidia's demand pipeline that is more concrete than analyst estimates or management commentary alone.

4. What is the Kyber NVL144 and why does it matter for Nvidia stock?
Kyber NVL144 is Nvidia's next rack-scale AI platform following Vera Rubin, designed to link multiple GPUs into a cluster for AI training and inference. A report suggesting Kyber would be delayed to 2028 caused concern about Nvidia's product roadmap before Nvidia denied the report and confirmed the schedule. The August 26 earnings call is the first formal opportunity to provide Kyber timeline specifics in a financial reporting context.

5. When does Nvidia next report earnings?
Nvidia reports fiscal Q2 2027 results on August 25 to 26, 2026. Q2 revenue guidance of approximately $45 billion was set during the Q1 report. Q3 guidance and any Kyber deployment timeline update will be the forward-looking signals investors use to evaluate whether the current 21.7 times forward multiple is the beginning of a re-rating or a temporary compression before further decline.