The Bottom of AMD's Guidance Beats the Top of Intel's. That Has Not Happened in 69 Quarters.

For 69 consecutive quarters, one number in the chip business was close to a law: Intel's revenue was bigger than AMD's. That count runs from the second quarter of 2009, which is as far back as an independent tally of the two companies' SEC filings goes. In the last three months of 2023 the ratio was about 2.5 to 1. A year ago it was 1.5 to 1. On Tuesday afternoon it stopped being a safe assumption and became a forecast.

AMD guided its December quarter to approximately $17 billion, plus or minus 3 percent. Intel, five days earlier, guided $15.0 billion to $16.0 billion. Run the ranges and AMD's floor, about $16.49 billion, sits roughly $490 million above Intel's ceiling of $16.0 billion. At the midpoints AMD is $1.5 billion ahead, about 10 percent.

Both figures are forecasts. In the quarter just reported, AMD was $3.1 billion behind. It has not out-earned Intel yet. It has told investors it will in three months, and Intel has told investors it cannot fill the orders it already has.

The quarter behind the forecast

AMD's third quarter came in at $13.377 billion, up 45 percent from a year earlier and 16 percent from the June quarter. Data Center was $8.5 billion, up 97 percent, and it is now 64 percent of the company against 47 percent a year ago. Client was $2.9 billion, up 7 percent. Gaming was $773 million, down 40 percent. Embedded was $1.1 billion, up 29 percent.

GAAP operating income was $2.832 billion, a 21 percent margin, against $1.270 billion a year earlier. Net income was $2.428 billion and diluted earnings per share $1.46. On a non-GAAP basis earnings per share were $2.09, up 74 percent, against a consensus of $1.95. Revenue beat the $13.09 billion analysts had modelled and Data Center beat the $8.20 billion they had modelled, both according to LSEG figures carried by Reuters.

The line that missed was client and gaming at $3.72 billion, against an estimate of $3.79 billion. The gaming half is the reason. Semi-custom console silicon is a business with a console cycle in it, and the cycle is on the wrong side of the curve.

Where the growth actually came from

Two facts sit inside AMD's client number that the headline skips. The first is that PC shipments were collapsing around it. IDC counted 62.7 million PCs shipped in the third quarter, down 20.1 percent from 78.5 million a year earlier, the sharpest third-quarter fall it has recorded. AMD grew client revenue 7 percent into that. You do not do that with price. You do it by taking sockets.

The second is that the growth is not really client at all. Data Center added about $4.2 billion of year-on-year revenue while client and gaming gave back roughly $330 million. AMD is now a data center company that also sells PCs. Lisa Su said as much on the call, telling analysts that 80 percent of next year's sales will come from data center chips, according to Reuters.

That mix shift explains why a revenue crossover is arriving before any unit-share crossover. AMD is not selling more processors than Intel. Mercury Research's most recent published count, for the second quarter, put AMD at a record 30.7 percent of x86 unit shipments against Intel's 69.3 percent. Narrow the comparison to EPYC against Xeon SP and AMD's adjusted share reaches 46.4 percent, but that is a slice of a slice. The company about to be larger by revenue still ships fewer than a third of the CPUs.

Profitability crossed first

Revenue is the number the headline writes. Operating income is the one that says which business is healthier right now, and that line crossed in the September quarter.

AMD reported $2.832 billion of GAAP operating income. Intel reported $2.522 billion on $3.1 billion more revenue. AMD's margin was 21 percent, Intel's 15.3 percent. On a non-GAAP basis AMD's net income was about $3.5 billion against Intel's $2.756 billion.

Intel still reports more GAAP net income, $3.32 billion against AMD's $2.428 billion, and that gap is mostly tax and one-offs rather than operations. The operating line is the cleaner comparison, and it has already changed hands.

The catch is direction. AMD guides non-GAAP gross margin of about 55 percent for the December quarter, down from 56 percent, even as revenue rises 27 percent sequentially. Rack-scale systems put bought-in accelerators and high-bandwidth memory inside AMD's revenue line. Helios is a rack, not a chip, and racks carry other people's silicon in them. The revenue crown is arriving with a mix that dilutes margin. Intel's guide is a 38 percent non-GAAP gross margin, so AMD is still the more profitable manufacturer per dollar of sales. It is less profitable per dollar than it was last quarter.

Two corrections to the coverage

The Register's write-up, which framed this as a change at the top of x86, makes two claims that do not survive the filings.

The first is that Intel's guidance amounts to "almost no growth." Intel's fourth quarter of 2025 was $13.7 billion. The midpoint of the new guide, $15.5 billion, is about 13 percent above that. What the number is, precisely, is a 6 percent sequential decline from Intel's own third quarter, with client revenue guided down more than 20 percent as OEMs drain inventory. Intel is forecasting annual growth and quarterly shrinkage at the same time, which is a stranger story than flatness.

The second is that Intel "would still have higher overall revenue, due to its foundry operations." It would not, because foundry revenue is already inside the number. Intel's third-quarter segment table shows Foundry at $6.697 billion, All Other at $692 million, and intersegment eliminations of minus $6.291 billion, reconciling to $16.513 billion consolidated. The foundry line is mostly Intel selling wafers to Intel. There is no second revenue stream stacked on top of the $15 billion to $16 billion guide. That guide is the whole company.

Supply is the story underneath

Neither company is forecasting a demand problem. Both are forecasting a supply one, from opposite ends.

Su's line on the call was that AMD has "secured the wafers, substrates, advanced packaging, memory, and other key components needed to support much higher growth next year," and that it is investing alongside suppliers to hold capacity. She guided 2027 company growth above 80 percent, data center CPU revenue roughly doubling, and data center AI revenue "significantly more than double."

Lip-Bu Tan's line, in Intel's release, was narrower: "Our immediate priority is to unlock additional supply to meet extraordinary demand for our Xeon family." Intel has said the first half of 2027 will be tighter on wafers and substrates than the second half. It is guiding down into a market where it cannot ship what customers are asking for.

Read together, the crossover is being decided by who can get substrate, packaging capacity and HBM allocated for the next four quarters, not by who has the better processor. AMD's forecast assumes its supply chain executes at a scale it has never run at. Intel's assumes it stays constrained. Both assumptions are checkable, and neither company has published the purchase commitments behind them.

Memory makes it sharper. Su told analysts that "higher memory and component costs weighed on demand" in the PC market, and said AMD expects PC shipments to fall by a double-digit percentage in 2027 while AMD outperforms the market. That is the same market Intel is guiding down into. The difference is exposure. Client is about 22 percent of AMD's revenue and roughly half of Intel's.

What it changes if you buy x86

For most of the last decade the procurement default was simple. Intel was the volume platform and AMD was the value alternative you qualified as a second source, often reluctantly and often late. That default is now upside down in one specific respect. The vendor with the larger revenue run rate is the one whose roadmap depends on capacity nobody has tested at this scale, and the smaller vendor is the one that has publicly said it cannot fill orders.

None of this says switch suppliers. It says a 2027 platform decision now carries a supply assumption inside it, and it is worth knowing which one. Three questions are worth putting to an OEM or a distributor this quarter rather than next. Which of the 2027 SKUs are dual-sourced at the board level, not just at the CPU socket. What allocation, as opposed to price, has been committed in writing for the first half. And how much of the unit cost increase is memory content rather than processor price, because those two move for different reasons and only one of them is negotiable.

The strategic version of the same question is validation cost. Qualifying a second x86 vendor is not a purchase order. It is firmware, driver stacks, operating system certification and a support model, and it takes quarters. The teams that started that work while AMD was still the cheaper option are the ones who will have a real choice in 2027. The ones who start now will be buying whatever is available.

At DMC we do that work with hardware companies and the organisations buying from them: platform and vendor qualification planning, BOM cost modelling when component prices move quarterly, and sourcing strategy when allocation rather than price is the constraint. If your 2027 x86 plan still assumes one vendor can fill it, let's talk.