Meta's Muse Sent Chip Stocks Soaring. What the Filings Say About META, AMD, INTC, QCOM and NVDA

Meta Platforms rose 11.4% on Monday, September 21, 2026 — its biggest one-day gain since April 2025, worth roughly $190 billion of market value. The trigger was Muse, the personal AI agent Meta launched on September 8, which reached the top of the U.S. iOS free-app chart within two weeks. The bigger moves were elsewhere: Arm up 17%, Intel up 12%, AMD up 10% to close above a $1 trillion market value for the first time, Qualcomm up 6–9%. Nvidia, the company the AI trade is usually about, gained 2.3%. The Nasdaq set a record close. This article does something the price chart cannot: it puts the five companies most affected next to their own financial statements and asks what the filings say about the businesses that just got repriced.

What the market is actually betting on

The thesis behind Monday’s rally is specific and worth stating precisely, because you cannot evaluate a bet you have not written down. AI agents — software that carries out tasks such as booking an appointment or completing a purchase — run inference continuously, on the user’s behalf, for hundreds of millions of people. Training a model is a one-time, GPU-heavy job; running an agent for a billion users is an always-on workload that leans on CPUs and on-device chips as much as on data-centre accelerators. If Muse is the first mass-market agent, then demand for CPUs and inference silicon (Arm, Intel, AMD, Qualcomm) rises alongside the demand for training hardware (Nvidia), rather than being displaced by it. Arm has said it sees the server CPU market growing about 35% a year to $120 billion by 2030; AMD’s own figure is $220 billion.

That is a reasonable thesis. It is also a thesis about 2028–2030 revenue, and the stocks moved on a two-week app-store ranking. The gap between those two things is where the financial statements come in.

Five companies, five verdicts, one table

Investor Sam scores every company on five pillars — quality, financial health, growth, value and momentum — using only the figures in its SEC filings, and combines them into a single verdict. The pillar scores below are as of the latest annual filing for each company. What matters here is not the overall rating so much as the spread between the pillars, because that spread is the difference between a business that is strong and a stock that is merely moving.

Two things jump out. First, the quality column runs from 100 to 7: this is not a group of similar businesses that happen to sell chips, it is a group with almost nothing in common except the word “AI” in the headline. Second, the value column is uniformly weak. Even before Monday, none of the five was cheap on our sector-relative and intrinsic measures, and the pillar scores were computed at prices well below where the shares trade now.

Meta: the buyer, not the seller

Meta is the one company on this list that spends on chips rather than selling them, and its filings show what that costs. Its FY2025 numbers on the income statement are excellent by any standard — a 41% operating margin, a 30% net margin, a 28% return on equity. But the year-over-year direction is down: net margin was 38% the year before and return on equity was 34%. The balance sheet tells you why. Meta ended 2024 with about $15 billion more cash than debt and ended 2025 with about $23 billion more debt than cash. That swing of nearly $38 billion is the AI build-out, financed. Our verdict flags return on assets falling 27% year over year for the same reason: the asset base ballooned with data centres before the revenue from them arrived.

Muse is the first evidence that the revenue side might show up. If a free agent used by hundreds of millions of people becomes a new surface for the advertising business, the capex was worth it. If it does not, Meta is a company that took on debt to build capacity it will have to write down slowly. The 11% move says the market now leans toward the first reading. The filings say the question is still open — and that the answer will arrive on the income statement, quarter by quarter, in the form of revenue growth that exceeds depreciation growth.

Nvidia: the incumbent that barely moved

Nvidia’s trailing figures remain the best in the group and among the best of any large company in the database: a 56% net margin, a 76% return on equity, a 62% return on invested capital, net cash on the balance sheet. Its P/E on our ratios page actually fell from 48× to 38× over the past fiscal year because earnings grew faster than the share price. The 2.3% gain on Monday is a market saying “this thesis helps everyone else more than it helps you,” which is probably right in relative terms and irrelevant in absolute ones. Nvidia’s own flag in our system — return on assets down 11% year over year — is the flag every hyper-profitable company eventually gets as it accumulates cash. The number to watch is the growth pillar, currently 23, because it measures how much last year’s growth rate has slowed.

AMD: the $1 trillion question

AMD deserves the most careful reading, because the gap between its momentum and its fundamentals is the widest on the list. Our verdict is STRONG BUY with a composite of 77 — but look at where the points come from. Growth scores 100 (net margin doubled from 6% to 12.5%; operating margin from 7% to 11%), financial health scores 100 (net cash, a current ratio of 2.9), momentum scores 98 (the stock was up 291% over the trailing year when the data was computed). Quality scores 53: return on equity is 6.9%, return on invested capital 4.4%, net margin 12.5%. Those are respectable numbers for an industrial company. They are not the numbers of a business worth a trillion dollars today.

Qualcomm: high quality, halved margin

Qualcomm is the quiet one. A 26% return on equity, a 27% return on invested capital, interest coverage of 19× and a valuation of 33× trailing earnings. The oddity in its FY2025 filing (fiscal year ended September 2025) is that net margin fell from 26% to 12.5% while operating margin held steady at 26–28%. When those two diverge, the cause is below the operating line — a tax charge, an impairment, a legal settlement — and it is usually non-recurring. Read the income statement below operating income before concluding the business deteriorated; our verdict’s “return on assets declining” flag is triggered by that one line. If the on-device inference thesis is right, Qualcomm is the most direct beneficiary of the five, since its chips already sit in the phones an agent runs on.

Intel: a turnaround priced as if it had already happened

Intel’s quality score of 7 is not a typo. Its FY2025 net margin was −0.5%, return on equity −0.2%, operating margin −4%. That is a dramatic improvement on FY2024 (net margin −35%), which is why the growth pillar scores 66 — but it is an improvement from catastrophic to break-even, not from good to better. The balance sheet carries about $32 billion of net debt against roughly $8.5 billion of EBITDA, or 3.8×, in a week when the Fed raised rates and the 10-year Treasury sits above 5%. Interest coverage cannot be calculated because there is no operating income to cover it with.

The stock was up 312% over the trailing year when our data was computed, and it added another 12% on Monday. The Muse thesis — agents need CPUs, Intel makes CPUs — is a reason for that. But every dollar of the current share price is a claim on a recovery that the filings show has reached zero and not yet gone positive. That can be a fine investment. It is not a fundamentals investment; it is a turnaround bet, and the honest way to hold it is to know which one you are making.

What to do with a one-day move of 11%

Three practical suggestions, none of which involve predicting next week.

1. Separate the pillars. If you own any of these, look at the verdict page and ask which pillar your thesis rests on. If it is momentum, you are trading. If it is quality and health, you are investing, and one day’s move should not change your view. Meta and Nvidia are quality-and-health stories that happen to have momentum; AMD and Intel are momentum-and-growth stories whose quality has to catch up.
2. Compare them properly. The compare tool lines up any two or three companies’ statements side by side. AMD against Nvidia is the instructive pair: similar market values, a fourfold difference in net margin.
3. Watch the next filing, not the next headline. For Meta, the tell is revenue growth versus depreciation growth. For AMD, it is operating margin. For Intel, it is operating income turning positive. For Qualcomm, it is whether the below-the-line charge repeats. For Nvidia, it is the growth rate itself. Every one of those is a number on the income statement, and none of them is an app-store ranking.

Frequently asked questions

Why did Nvidia rise so much less than Intel and Arm?

Because the Muse thesis is about inference on CPUs and edge devices, where Nvidia’s share is smaller, and because Nvidia had already been re-rated for the AI trade. A 2% move on a company with Nvidia’s margins is not a verdict on its business; it is the market allocating an incremental idea to the companies with the most to gain from it.

Is the STRONG BUY on AMD a recommendation to buy at $1 trillion?

No. The verdict is a deterministic score computed from the most recent annual filing, and the pillars that drive it — growth, health and momentum — are visible in the table so you can weigh them yourself. It was computed at a price far below the current one, which is why the value pillar deserves particular attention now. Investor Sam is an analysis tool, not an advisor.

Where do these figures come from?

From each company’s most recent 10-K on SEC EDGAR: Meta and AMD for the year ended December 2025, Qualcomm for the year ended September 2025, Nvidia for the year ended January 2026, Intel for the year ended December 2025. Monday’s share-price moves are from public market data.

Sources

• Company 10-K filings via SEC EDGAR: Meta Platforms, Advanced Micro Devices, Intel, Qualcomm, Nvidia (fiscal years as noted above); pillar scores and ratios as computed on Investor Sam.
• Market moves for September 21, 2026 as reported by Bloomberg, CNBC and Yahoo Finance.
• Arm Holdings and AMD public statements on server CPU market projections.