AMD Just Had Its Best Quarter Ever. Nvidia Still Added 11 Times More Data-Center Revenue.

Eight and a half billion dollars. That is what AMD's data-center business took in during the third quarter of 2026, up 97 percent from a year earlier, the biggest quarter in the company's history by almost every measure that counts. Total revenue hit a record $13.4 billion while earnings per share jumped 74 percent. Two days after the report, chief executive Lisa Su told Yahoo Finance that AMD is now, in her words, a data center first company, and every headline number went the right way. And the gap with Nvidia got wider anyway.

Here is the number that did not make the earnings headlines. In the same twelve months, Nvidia's data-center business added $48 billion of quarterly revenue, while AMD added $4.2 billion. Nvidia's gain was 11.5 times AMD's, in a year when AMD grew faster than it ever has. Percentages flatter the challenger, but absolute dollars decide the market.

What AMD actually reported

AMD reported third-quarter 2026 results on November 3, and the release reads like a company being remade in real time. Data Center revenue reached a record $8.5 billion, up 97 percent year over year and 27 percent from the prior quarter, and the segment now supplies 64 percent of AMD's total revenue, up from 47 percent a year ago. Operating income in the segment hit $2.9 billion, up 168 percent, with the operating margin widening to 34 percent from 25 percent. Two engines drove it: record server revenue from fifth-generation EPYC processors, now the most widely deployed EPYC generation in the public cloud with more than 540 instance types, and the launch of the Instinct MI450 Series GPUs, with the MI455X shipping in volume and the first Helios rack systems coming online at Anthropic, Meta, Microsoft, OpenAI, and Oracle.

Every other segment tells the same story from the opposite direction. Gaming revenue fell 40 percent to $773 million as console chip sales faded late in the cycle. Client revenue rose 7 percent to $2.9 billion on Ryzen mobile processors, while memory costs weighed on PC demand late in the quarter, and Embedded grew 29 percent to $1.1 billion. Gross margin reached 56 percent, operating income rose 75 percent to a record $3.9 billion, and inventory jumped about $3 billion in a single quarter to $11.5 billion, stockpiled to feed data-center demand. Investors noticed: AMD shares closed at $552.90 on November 11, up 7.3 percent on the day, with the stock up 158 percent for the year and the company's market value at $837 billion.

AMD Q3 2026 by segment, from the company's November 3, 2026 prepared remarks.
Segment	Revenue	Year over year	Share of total
Data Center	$8.5 billion	+97%	64%
Client	$2.9 billion	+7%	22%
Gaming	$0.77 billion	-40%	6%
Embedded	$1.1 billion	+29%	8%
Total	$13.4 billion	+45%	100%

The math nobody ran

Start with the additions, because growth rates hide the scale of the contest. AMD's Data Center revenue a year ago was $8.5 billion divided by 1.97, or about $4.3 billion, so the celebrated 97 percent growth added roughly $4.2 billion of quarterly revenue. Nvidia's Data Center revenue a year ago was $89.0 billion divided by 2.17, or about $41.0 billion, so its 117 percent growth added roughly $48.0 billion. One company's record year of growth is the other company's rounding error, and the arithmetic cuts the other way too: AMD's entire quarterly revenue, all four segments combined, is 15 percent of Nvidia's data-center revenue alone.

The ratio is stark: Nvidia's data-center business, at $89.0 billion in its latest reported quarter, is 10.5 times AMD's $8.5 billion. Annualized, the run rates are roughly $356 billion against $34 billion. If both companies held their current growth rates for another year, the gap would stretch from $80.5 billion to about $176 billion, because Nvidia is growing faster in percentage terms as well as in dollars. That is the detail that breaks the comeback story: AMD is not closing on a slowing leader. It is chasing a leader that is accelerating away.

To answer how long the chase would take, run the generous scenario, the one AMD's bulls would choose. Suppose Nvidia's data-center growth collapses from 117 percent to 30 percent a year, while AMD sustains 97 percent indefinitely. AMD still needs about 5.7 years to draw level. Five years and eight months of flawless execution, against a rival that has to stumble, just to tie. And nothing in the record suggests Nvidia is stumbling: its third-quarter guidance is $108 billion, its fiscal 2028 revenue growth is guided at roughly 70 percent, and it reports its next quarter on November 17, five days from now, with every analyst on Wall Street watching whether the 117 percent holds.

AMD's own guidance concedes the scale of the task even as it promises the fastest growth in the company's history. Fourth-quarter revenue is guided to about $17 billion, with Data Center up roughly 50 percent sequentially, which implies about $12.75 billion for the segment. If Nvidia's guided $108 billion quarter carries data-center revenue at its current 92.5 percent share, that is about $99.9 billion. Even after AMD's biggest sequential jump ever, the gap stays near $87 billion. Revenue growth above 80 percent in 2027, which AMD now expects, with Data Center passing 80 percent of total revenue, is a genuine transformation. It is also a transformation that leaves AMD roughly where Nvidia was two years ago.

The gap math, computed from AMD's November 3, 2026 remarks and Nvidia's August 26, 2026 results.
Figure	Value	How it was computed
Nvidia Data Center, latest quarter	$89.0 billion	Nvidia Q2 FY2027 release
AMD Data Center, latest quarter	$8.5 billion	AMD Q3 2026 remarks
Size ratio	10.5x	89.02 / 8.5
Nvidia quarterly DC gain, 1 year	+$48.0 billion	89.02 - (89.02 / 2.17)
AMD quarterly DC gain, 1 year	+$4.2 billion	8.5 - (8.5 / 1.97)
Additions ratio	11.5x	48.0 / 4.19
Annualized run rates	$356B vs $34B	Quarterly revenue x 4
Gap in one year at current rates	~$176 billion	89.02 x 2.17 - 8.5 x 1.97
Catch-up time, generous case	~5.7 years	AMD +97%, Nvidia +30%
AMD total revenue vs Nvidia DC	15.1%	13.4 / 89.02

Gaming is now a rounding error

Step back from the Nvidia comparison, because the second story in this quarter is about AMD itself. Gaming, the business that defined AMD for a generation through PlayStation and Xbox chips and Radeon graphics cards, brought in $773 million. That is 5.8 percent of revenue, against 64 percent for Data Center. AMD's entire gaming quarter equals about 18 percent of what its data-center business added in a single year. Su's line about a data center first company is not branding. It is a description of a balance sheet that already flipped, and the fourth-quarter outlook makes it plainer: Client and Gaming revenue is guided down more than 20 percent sequentially, while Data Center jumps about 50 percent.

None of this makes AMD a failure, far from it. Helios racks are running at five of the most demanding AI customers on earth, EPYC is the host CPU of choice for custom accelerators (OpenAI picked Turin EPYC chips for its first custom accelerator platforms), the Venice generation has more than 240 OEM platforms committed, and AMD bought two companies in the quarter to fill gaps in low-latency inference silicon and spatial intelligence, hiring World Labs founder Fei-Fei Li as chief scientist. A company growing data-center revenue 97 percent a year with a 34 percent operating margin in the segment is a serious second source, and the AI industry wants a second source badly enough to hand AMD warrants and multi-year commitments. Being the clear number two in the biggest buildout in computing history is a very good business, but it is just not a close race.

Limitations

Several numbers in this article carry asterisks that belong in the open. AMD does not split Data Center revenue between EPYC server CPUs and Instinct AI GPUs, so nobody outside the company knows what share of the $8.5 billion is AI accelerators and what share is server processors, which means the head-to-head AI chip comparison with Nvidia is approximate by construction. The two quarters are not perfectly aligned either: AMD's quarter is calendar Q3 2026, while Nvidia's latest reported quarter ended July 26, 2026, so this is latest-reported-quarter against latest-reported-quarter, and Nvidia's November 17 report will reset the numbers within days. Growth-rate extrapolation is arithmetic, not a forecast, since neither company sustains triple-digit growth indefinitely, and the 5.7-year catch-up horizon is a what-if built on assumed rates, not a prediction. Margin comparisons mix measures: AMD's 56 percent is a company-wide gross margin and its Data Center operating margin is 34 percent, while Nvidia's 75 percent is a company-wide gross margin, and AMD discloses no Data Center gross margin to compare directly. Finally, AMD took a $1.1 billion noncash charge in the quarter from customer warrants whose dilutive mechanics are not fully public, and its stock, up 158 percent this year at a trailing price-to-earnings ratio near 110, is priced for the growth story to keep working.

Strongest case against

Steelman: the skeptic's case starts with timing and ends with structure. First, the comparison is a snapshot taken at AMD's best possible moment, days after its record quarter and days before Nvidia's report, and a strong Nvidia print on November 17 would make the gap look worse, not better, but a weak one would narrow it fast. Second, AMD's growth is partly a mix story: EPYC server CPUs are a large share of the $8.5 billion, so the AI accelerator gap is narrower or wider than the segment figures suggest, and nobody can tell which from public filings. Third, the warrants AMD issued to customers mean the largest buyers now hold equity upside in their supplier, a structure that buys demand today and dilutes shareholders tomorrow, and the $1.1 billion charge is the first installment of that bill. Fourth, AMD's own guidance admits the consumer businesses are shrinking into irrelevance, which concentrates the whole company on a single bet: that hyperscalers keep spending at a pace nobody has ever sustained. If AI infrastructure spending pauses for even two quarters, a company with $11.5 billion of inventory and Data Center at 64 percent of revenue has further to fall than a diversified one. Every clause of that case is true, and it is also compatible with AMD being the best-positioned challenger Nvidia has ever faced, which is why the November 17 report matters more than the November 3 one.

What you can do

If you hold AMD stock, the number to watch on November 17 is not Nvidia's revenue, which everyone expects to beat, but whether Nvidia's data-center growth stays above 100 percent, because that rate decides whether AMD's catch-up horizon is years or never. If you buy AI infrastructure, get AMD's MI450 pricing in writing now, while five frontier labs are validating it, since a credible second source is worth more before it becomes the default than after. If you work in semiconductors, note that EPYC's host-CPU wins at OpenAI and the custom-accelerator programs mean the CPU socket is becoming the strategic ground, not the accelerator alone. And if you are simply watching the AI buildout, treat AMD's quarter as the clearest evidence yet that the boom has a second winner, then ask the harder question it raises: what does it mean for the industry when the second winner is still ten times smaller than the first?

The Bottom Line

AMD just posted the best quarter in its history: $13.4 billion of revenue, a data-center business up 97 percent, gaming reduced to a footnote, and a chief executive willing to say out loud that AMD is a data center first company. It is a genuine transformation, and the stock market agrees, up 158 percent this year. It is also, by the arithmetic nobody ran on earnings day, not a close race: Nvidia added $48 billion of quarterly data-center revenue in the same year AMD added $4.2 billion, the leader is 10.5 times the challenger's size and growing faster in percentage terms, and even if Nvidia's growth collapsed to 30 percent while AMD sustained 97 percent, the catch-up would take nearly six years. AMD built a very good business in the biggest buildout in computing history. The numbers say it built it ten times smaller than the one that matters most.